This paper develops a Bayesian optimal experimental design for robot kinematic calibration on$\mathbb{S}^{3} \times \mathbb{R}^{3}$. Our method builds upon a Gaussian process approach that incorporates a geometry-aware kernel based on Riemannian Matérn kernels over$\mathbb{S}^{3}$. To learn the forward kinematics errors via Bayesian optimization with a Gaussian process, we define a geodesic distance-based objective function. Pointwise values of this function are sampled via noisy measurements taken using fiducial markers on the end-effector using a camera and computed pose with the nominal kinematics. The corrected Denavit-Hartenberg parameters are obtained using an efficient quadratic program that operates on the collected data sets. The effectiveness of the proposed method is demonstrated via simulations and calibration experiments on NASA's ocean world lander autonomy testbed (OWLAT).

Autonomous exploration of unknown environments is fundamentally a problem of decision making under uncertainty where the agent must account for uncertainty in sensor measurements, localization, action execution, as well as many other factors. For large-scale exploration applications, autonomous systems must overcome the challenges of sequentially deciding which areas of the environment are valuable to explore while safely evaluating the risks associated with obstacles and hazardous terrain. In this work, we propose a risk-aware meta-level decision making framework to balance the tradeoffs associated with local and global exploration. Meta-level decision making builds upon classical hierarchical coverage planners by switching between local and global policies with the overall objective of selecting the policy that is most likely to maximize reward in a stochastic environment. We use information about the environment history, traversability risk, and kinodynamic constraints to reason about the probability of successful policy execution to switch between local and global policies. We have validated our solution in both simulation and on a variety of large-scale real world hardware tests. Our results show that by balancing local and global exploration we are able to significantly explore large-scale environments more efficiently.

We introduce a novel Fractal Hand robotic gripper. The hand has only 1 actuator, but (2n+1−1) joints, where a design parameter n defines the depth of the fingers’ tree structures. The hand is synergistic in its operation (because its joint movements are coupled through the hand’s interaction with the grasped object), but it is not anthropomorphic. The basic finger and hand geometry, governing kinematics, and quasi-statics mechanics of a rigid version of the hand are developed. These analyses remarkably show that under mild constraints, the grasped object is compliantly stable at an equilibrium grasp configuration. Thus, the Fractal Hand adapts to a very wide range of planar objects with a single design. Grasp planning is thus simplified. A companion paper [33] introduces a design methodology for this new class of robot hands, and multiple prototypes.

Ensuring robot safety in complex environments is a difficult task due to actuation limits, such as torque bounds. This paper presents a safety-critical control framework that leverages learning-based switching between multiple backup controllers to formally guarantee safety under bounded control inputs while satisfying driver intention. By leveraging backup controllers designed to uphold safety and input constraints, backup control barrier functions (BCBFs) construct implicitly defined control invariant sets via a feasible quadratic program (QP). However, BCBF performance largely depends on the design and conservativeness of the chosen backup controller, especially in our setting of human-driven vehicles in complex, e.g, off-road, conditions. While conservativeness can be reduced by using multiple backup controllers, determining when to switch is an open problem. Consequently, we develop a broadcast scheme that estimates driver intention and integrates BCBFs with multiple backup strategies for human-robot interaction. An LSTM classifier uses data inputs from the robot, human, and safety algorithms to continually choose a backup controller in real-time. We demonstrate our method’s efficacy on a dualtrack robot in obstacle avoidance scenarios. Our framework guarantees robot safety while adhering to driver intention.

This paper and its companion propose a new fractal robotic gripper, drawing inspiration from the centuryold Fractal Vise. The unusual synergistic properties allow it to passively conform to diverse objects using only one actuator. Designed to be easily integrated with prevailing parallel jaw grippers, it alleviates the complexities tied to perception and grasp planning, especially when dealing with unpredictable object poses and geometries. We build on the foundational principles of the Fractal Vise to a broader class of gripping mechanisms and address the limitations that had led to its obscurity. Two Fractal Fingers, coupled with a closing actuator, can form an adaptive and synergistic Fractal Hand. We articulate a design methodology for low-cost, easy-to-fabricate, large workspace, and compliant Fractal Fingers. The companion paper delves into the kinematics and grasping properties of a specific class of Fractal Fingers and Hands.

Future NASA lander missions to icy moons will require completely automated, accurate, and data efficient calibration methods for the robot manipulator arms that sample icy terrains in the lander's vicinity. To support this need, this paper presents a Gaussian Process (GP) approach to the classical manipulator kinematic calibration process. Instead of identifying a corrected set of Denavit-Hartenberg kinematic parameters, a set of GPs models the residual kinematic error of the arm over the workspace. More importantly, this modeling framework allows a Gaussian Process Upper Confident Bound (GP-UCB) algorithm to efficiently and adaptively select the calibration's measurement points so as to minimize the number of experiments, and therefore minimize the time needed for recalibration. The method is demonstrated in simulation on a simple 2-DOF arm, a 6 DOF arm whose geometry is a candidate for a future NASA mission, and a 7 DOF Barrett WAM arm.

PARSEC (Payload Anchoring Robotic System for the Exploration of Cliffs) is an autonomy-equipped aerial manipulator that can deploy self-anchoring payloads on rocky vertical surfaces. It consists of a hexacopter and a two Degrees of Freedom (2 DoF) mass balancing manipulator, which can autonomously deploy a self-anchoring payload from its custom end-effector. The payload anchors itself via an actuated microspine gripper. Payload sensor data is wirelessly transmitted to the primary vehicle during and after deployment. A novel state machine controls the four-stage PARSEC deployment process. First, the rotorcraft brings the payload into contact with the surface and applies a constant 6 N normal force through a feedback control loop to preload the payload microspine gripper. Second, while the rotorcraft maintains the constant normal force, the gripper is commanded to close until engagement with the surface is confirmed through the current feedback sensing. Then, the aerial manipulator pulls with 5 N force on the anchored payload to ensure a secure grip before releasing the package and flying away. We present experimental validation of a successful deployment of a 430 g payload on a vertical vesicular basalt surface.

Many approaches to grasp synthesis optimize analytic quality metrics that measure grasp robustness based on finger placements and local surface geometry. However, generating feasible dexterous grasps by optimizing these metrics is slow, often taking minutes. To address this issue, this paper presents FRoGGeR: a method that quickly generates robust precision grasps using the min-weight metric, a novel, almost-everywhere differentiable approximation of the classical$\epsilon$grasp metric. The min-weight metric is simple and interpretable, provides a reasonable measure of grasp robustness, and admits numerically efficient gradients for smooth optimization. We leverage these properties to rapidly synthesize collision-free robust grasps-typically in less than a second. FRoGGeR can refine the candidate grasps generated by other methods (heuristic, data-driven, etc.) and is compatible with many object representations (SDFs, meshes, etc.), We study FRoGGeR's performance on over 40 objects drawn from the YCB dataset, outperforming a competitive baseline in computation time, feasibility rate of grasp synthesis, and picking success in simulation. We conclude that FRoGGeR is fast: it has a median synthesis time of 0.834s over hundreds of experiments.

The discovery of ocean worlds such as Enceladus, Titan, and Europa motivates the development of versatile autonomous mobility systems to enable the next era of space exploration where there is large uncertainty in terrain specifications due to a lack of prior surface reconnaissance missions. To explore these environments, we propose Exobiology Extant Life Surveyor (EELS): the first large-scale (4 lm long with 400 Nm peak torque) snake robot. The large scale is achieved by using a screw-based active skin mechanism to decouple motion and shape control. Autonomous mobility for such a system remains an open problem due to its many Degrees of Freedom (DoFs), complex terrain interactions, and intermittent localization failures in GPS-denied perceptually degraded environments due to the presence of fog, dust, featureless terrains, etc. We propose NEO, an autonomy architecture that scales to large DoFs to generate a versatile set of gaits to achieve mobility in unknown extreme environments. We also discuss the resilience capabilities of NEO that achieves closed-loop tracking performance by leveraging exteroception when available but can also operate with proprioception only, leading to resiliency against localization failures via graceful degradation in performance rather than unsafe behaviors. A quantitative hardware evaluation of exteroceptive leader-follower gait is performed indoors on synthetic ice along with qualitative results of field deployment of the proprioceptive leader-follower and sidewinding gaits in extreme environments of icy and sandy terrains with mobility-stressing elements such as trenches, undulations, and steep slopes (up to 35 degrees). We present a set of lessons learned from field deployments with a summary of challenges and open research problems. Video: www.rohanthakker.in/eels-neo-autonomy.html

Nonlinear dynamical effects are crucial to the operation of many agile robotic systems. Koopman-based model learning methods can capture these nonlinear dynamical system effects in higher dimensional lifted bilinear models that are amenable to optimal control. However, standard methods that lift the system state using a fixed function dictionary before model learning result in high dimensional models that are intractable for real time control. This paper presents a novel method that jointly learns a function dictionary and lifted bilinear model purely from data by incorporating the Koopman model in a neural network architecture. Nonlinear MPC design utilizing the learned model can be performed readily. We experimentally realized this method on a multirotor drone for agile trajectory tracking at low altitudes where the aerodynamic ground effect influences the system's behavior. Experimental results demonstrate that the learning-based controller achieves similar performance as a nonlinear MPC based on a nominal dynamics model in medium altitude. However, our learning-based system can reliably track trajectories in near-ground flight regimes while the nominal controller crashes due to unmodeled dynamical effects that are captured by our method.

We present a method for solving the coverage problem with the objective of autonomously exploring an unknown environment under mission time constraints. Here, the robot is tasked with planning a path over a horizon such that the accumulated area swept out by its sensor footprint is maximized. Because this problem exhibits a diminishing returns property known as submodularity, we choose to formulate it as a tree-based sequential decision making process. This formulation allows us to evaluate the effects of the robot's actions on future world coverage states, while simultaneously accounting for traversability risk and the dynamic constraints of the robot. To quickly find near-optimal solutions, we propose an effective approximation to the coverage sensor model which adapts to the local environment. Our method was extensively tested across various complex environments and served as the local exploration algorithm for a competing entry in the DARPA Subterranean Challenge.

We present a method for autonomous exploration of large-scale unknown environments under mission time con-straints. We start by proposing the Frontloaded Information Gain Orienteering Problem (FIG-OP) - a generalization of the traditional orienteering problem where the assumption of a reliable environmental model no longer holds. The FIG-OP ad-dresses model uncertainty by frontloading expected information gain through the addition of a greedy incentive, effectively expe-diting the moment in which new area is uncovered. In order to reason across multi-kilometer environments, we solve FIG-OP over an information-efficient world representation, constructed through the aggregation of information from a topological and metric map. Our method was extensively tested and field-hardened across various complex environments, ranging from subway systems to mines. In comparative simulations, we observe that the FIG-OP solution exhibits improved coverage efficiency over solutions generated by greedy and traditional orienteering-based approaches (i.e. severe and minimal model uncertainty assumptions, respectively).

This paper considers progressively more demanding off-line shortest path sensory coverage problems in an optimization framework. In the first problem, a robot finds the shortest path to cover a set of target nodes with its sensors. Because this mixed integer nonlinear optimization problem (MINLP) is NP-hard, we develop a polynomial-time approximation algorithm with a bounded approximation ratio. The next problem shortens the coverage path when possible by viewing multiple targets from a single pose. Its polynomial-time approximation simplifies the coverage path geometry. Finally, we show how the complete sensory coverage problem can be formulated as a MINLP over a decomposition of a given region into arbitrary convex polygons. Extensions of the previously introduced algorithms provides a polynomial time solution with bounded approximation. Examples illustrate the methods.

When autonomous robots interact with humans, such as during autonomous driving, explicit safety guarantees are crucial in order to avoid potentially life-threatening accidents. Many data-driven methods have explored learning probabilistic bounds over human agents’ trajectories (i.e. confidence tubes that contain trajectories with probability δ), which can then be used to guarantee safety with probability 1− δ. However, almost all existing works consider δ ≥ 0.001. The purpose of this paper is to argue that (1) in safety-critical applications, it is necessary to provide safety guarantees with δ < 10−8, and (2) current learning-based methods are illequipped to compute accurate confidence bounds at such low δ. Using human driving data (from the highD dataset), as well as synthetically generated data, we show that current uncertainty models use inaccurate distributional assumptions to describe human behavior and/or require infeasible amounts of data to accurately learn confidence bounds for δ ≤ 10−8. These two issues result in unreliable confidence bounds, which can have dangerous implications if deployed on safety-critical systems.

Koopman-based learning methods can potentially be practical and powerful tools for dynamical robotic systems. However, common methods to construct Koopman representations seek to learn lifted linear models that cannot capture nonlinear actuation effects inherent in many robotic systems. This paper presents a learning and control methodology that is a first step towards overcoming this limitation. Using the Koopman canonical transform, control-affine dynamics can be expressed by a lifted bilinear model. The learned model is used for nonlinear model predictive control (NMPC) design where the bilinear structure can be exploited to improve computational efficiency. The benefits for control-affine dynamics compared to existing Koopman-based methods are highlighted through an example of a simulated planar quadrotor. Prediction error is greatly reduced and closed loop performance similar to NMPC with full model knowledge is achieved.

Characterizing what types of exoskeleton gaits are comfortable for users, and understanding the science of walking more generally, require recovering a user’s utility landscape. Learning these landscapes is challenging, as walking trajectories are defined by numerous gait parameters, data collection from human trials is expensive, and user safety and comfort must be ensured. This work proposes the Region of Interest Active Learning (ROIAL) framework, which actively learns each user’s underlying utility function over a region of interest that ensures safety and comfort. ROIAL learns from ordinal and preference feedback, which are more reliable feedback mechanisms than absolute numerical scores. The algorithm’s performance is evaluated both in simulation and experimentally for three non-disabled subjects walking inside of a lower-body exoskeleton. ROIAL learns Bayesian posteriors that predict each exoskeleton user’s utility landscape across four exoskeleton gait parameters. The algorithm discovers both commonalities and discrepancies across users’ gait preferences and identifies the gait parameters that most influenced user feedback. These results demonstrate the feasibility of recovering gait utility landscapes from limited human trials.

Aircraft that can launch ballistically and convert to autonomous, free-flying drones have applications in many areas such as emergency response, defense, and space exploration, where they can gather critical situational data using onboard sensors. This paper presents a ballistically-launched, autonomously-stabilizing multirotor prototype (SQUID - Streamlined Quick Unfolding Investigation Drone) with an onboard sensor suite, autonomy pipeline, and passive aerodynamic stability. We demonstrate autonomous transition from passive to vision-based, active stabilization, confirming the multirotor's ability to autonomously stabilize after a ballistic launch in a GPS-denied environment.

This paper presents a novel episodic method to learn a robot's nonlinear dynamics model and an increasingly optimal control sequence for a set of tasks. The method is based on the Koopman operator approach to nonlinear dynamical systems analysis, which models the flow of observables in a function space, rather than a flow in a state space. Practically, this method estimates a nonlinear diffeomorphism that lifts the dynamics to a higher dimensional space where they are linear. Efficient Model Predictive Control methods can then be applied to the lifted model. This approach allows for real time implementation in on-board hardware, with rigorous incorporation of both input and state constraints during learning. We demonstrate the method in a real-time implementation of fast multirotor landing, where the nonlinear ground effect is learned and used to improve landing speed and quality.

Many tasks in robot-assisted surgeries (RAS) can be represented by finite-state machines (FSMs), where each state represents either an action (such as picking up a needle) or an observation (such as bleeding). A crucial step towards the automation of such surgical tasks is the temporal perception of the current surgical scene, which requires a real-time estimation of the states in the FSMs. The objective of this work is to estimate the current state of the surgical task based on the actions performed or events occurred as the task progresses. We propose Fusion-KVE, a unified surgical state estimation model that incorporates multiple data sources including the Kinematics, Vision, and system Events. Additionally, we examine the strengths and weaknesses of different state estimation models in segmenting states with different representative features or levels of granularity. We evaluate our model on the JHU-ISI Gesture and Skill Assessment Working Set (JIGSAWS), as well as a more complex dataset involving robotic intra-operative ultrasound (RIOUS) imaging, created using the da Vinci® Xi surgical system. Our model achieves a superior frame-wise state estimation accuracy up to 89.4%, which improves the state-of-the-art surgical state estimation models in both JIGSAWS suturing dataset and our RIOUS dataset.

This paper presents a personalized gait optimization framework for lower-body exoskeletons. Rather than optimizing numerical objectives such as the mechanical cost of transport, our approach directly learns from user prefer-ences, e.g., for comfort. Building upon work in preference-based interactive learning, we present the CoSpar algorithm. CoSpar prompts the user to give pairwise preferences between trials and suggest improvements; as exoskeleton walking is a non-intuitive behavior, users can provide preferences more easily and reliably than numerical feedback. We show that CoSpar performs competitively in simulation and demonstrate a prototype implementation of CoSpar on a lower-body exoskeleton to optimize human walking trajectory features. In the experiments, CoSpar consistently found user-preferred parameters of the exoskeleton's walking gait, which suggests that it is a promising starting point for adapting and personalizing exoskeletons (or other assistive devices) to individual users.

This paper serves as one of the first efforts to enable large-scale and long-duration autonomy using the Boston Dynamics Spot robot. Motivated by exploring extreme environments, particularly those involved in the DARPA Subterranean Challenge, this paper pushes the boundaries of the state-of-practice in enabling legged robotic systems to accomplish real-world complex missions in relevant scenarios. In particular, we discuss the behaviors and capabilities which emerge from the integration of the autonomy architecture NeBula (Networked Belief-aware Perceptual Autonomy) with next-generation mobility systems. We will discuss the hardware and software challenges, and solutions in mobility, perception, autonomy, and very briefly, wireless networking, as well as lessons learned and future directions. We demonstrate the performance of the proposed solutions on physical systems in real-world scenarios.3The proposed solution contributed to winning 1st-place in the 2020 DARPA Subterranean Challenge, Urban Circuit.4

Sampling-based motion planners (SBMP) are commonly used to generate motion plans by incrementally constructing a search tree through a robot's configuration space. For high degree-of-freedom systems, sampling is often done in a lower-dimensional space, with a steering function responsible for local planning in the higher-dimensional configuration space. However, for highly-redundant systems with complex kinematics, this approach is problematic due to the high computational cost of evaluating the steering function, especially in cluttered environments. Therefore, having an efficient, informed sampler becomes critical to online robot operation. In this study, we develop a learning-based approach with policy improvement to compute an optimal sampling distribution for use in SBMPs. Motivated by the challenge of whole-body planning for a 31 degree-of-freedom mobile robot built by the Toyota Research Institute, we combine our learning-based approach with classical graph-search to obtain a constrained sampling distribution. Over multiple learning iterations, the algorithm learns a probability distribution weighting areas of low-cost and high probability of success, which a graph search algorithm then uses to obtain an optimal sampling distribution for the robot. On challenging motion planning tasks for the robot, we observe significant computational speed-up, fewer edge evaluations, and more efficient paths with minimal computational overhead. We show the efficacy of our approach with a number of experiments in whole-body motion planning.

This paper presents a technique to concurrently and jointly predict the future trajectories of surgical instruments and the future state(s) of surgical subtasks in robot-assisted surgeries (RAS) using multiple input sources. Such predictions are a necessary first step towards shared control and supervised autonomy of surgical subtasks. Minute-long surgical subtasks, such as suturing or ultrasound scanning, often have distinguishable tool kinematics and visual features, and can be described as a series of fine-grained states with transition schematics. We propose daVinciNet - an end-to-end dual-task model for robot motion and surgical state predictions. daVinciNet performs concurrent end-effector trajectory and surgical state predictions using features extracted from multiple data streams, including robot kinematics, endoscopic vision, and system events. We evaluate our proposed model on an extended Robotic Intra-Operative Ultrasound (RIOUS+) imaging dataset collected on a da Vinci® Xi surgical system and the JHU-ISI Gesture and Skill Assessment Working Set (JIGSAWS). Our model achieves up to 93.85% short-term (0.5s) and 82.11% long-term (2s) state prediction accuracy, as well as 1.07mm short-term and 5.62mm long-term trajectory prediction error.

Hybrid locomotion, which combines multiple modalities of locomotion within a single robot, enables robots to carry out complex tasks in diverse environments. This paper presents a novel method for planning multi-modal locomotion trajectories using approximate dynamic programming. We formulate this problem as a shortest-path search through a state-space graph, where the edge cost is assigned as optimal transport cost along each segment. This cost is approximated from batches of offline trajectory optimizations, which allows the complex effects of vehicle under-actuation and dynamic constraints to be approximately captured in a tractable way. Our method is illustrated on a hybrid double-integrator, an amphibious robot, and a flying-driving drone, showing the practicality of the approach.

Optimizing lower-body exoskeleton walking gaits for user comfort requires understanding users' preferences over a high-dimensional gait parameter space. However, existing preference-based learning methods have only explored low-dimensional domains due to computational limitations. To learn user preferences in high dimensions, this work presents LINECOSPAR, a human-in-the-loop preference-based framework that enables optimization over many parameters by iteratively exploring one-dimensional subspaces. Additionally, this work identifies gait attributes that characterize broader preferences across users. In simulations and human trials, we empirically verify that LINECOSPAR is a sample-efficient approach for high-dimensional preference optimization. Our analysis of the experimental data reveals a correspondence between human preferences and objective measures of dynamicity, while also highlighting differences in the utility functions underlying individual users' gait preferences. This result has implications for exoskeleton gait synthesis, an active field with applications to clinical use and patient rehabilitation.

In preference-based reinforcement learning (RL), an agent interacts with the environment while receiving preferences instead of absolute feedback. While there is increasing research activity in preference-based RL, the design of formal frameworks that admit tractable theoretical analysis remains an open challenge. Building upon ideas from preference-based bandit learning and posterior sampling in RL, we present DUELING POSTERIOR SAMPLING (DPS), which employs preference-based posterior sampling to learn both the system dynamics and the underlying utility function that governs the preference feedback. As preference feedback is provided on trajectories rather than individual state-action pairs, we develop a Bayesian approach for the credit assignment problem, translating preferences to a posterior distribution over state-action reward models. We prove an asymptotic Bayesian no-regret rate for DPS with a Bayesian linear regression credit assignment model. This is the first regret guarantee for preference-based RL to our knowledge. We also discuss possible avenues for extending the proof methodology to other credit assignment models. Finally, we evaluate the approach empirically, showing competitive performance against existing baselines.

Reinforcement Learning (RL) algorithms have found limited success beyond simulated applications, and one main reason is the absence of safety guarantees during the learning process. Real world systems would realistically fail or break before an optimal controller can be learned. To address this issue, we propose a controller architecture that combines (1) a model-free RL-based controller with (2) model-based controllers utilizing control barrier functions (CBFs) and (3) online learning of the unknown system dynamics, in order to ensure safety during learning. Our general framework leverages the success of RL algorithms to learn high-performance controllers, while the CBF-based controllers both guarantee safety and guide the learning process by constraining the set of explorable polices. We utilize Gaussian Processes (GPs) to model the system dynamics and its uncertainties. Our novel controller synthesis algorithm, RL-CBF, guarantees safety with high probability during the learning process, regardless of the RL algorithm used, and demonstrates greater policy exploration efficiency. We test our algorithm on (1) control of an inverted pendulum and (2) autonomous carfollowing with wireless vehicle-to-vehicle communication, and show that our algorithm attains much greater sample efficiency in learning than other state-of-the-art algorithms and maintains safety during the entire learning process.

Dealing with high variance is a significant challenge in model-free reinforcement learning (RL). Existing methods are unreliable, exhibiting high variance in performance from run to run using different initializations/seeds. Focusing on problems arising in continuous control, we propose a functional regularization approach to augmenting model-free RL. In particular, we regularize the behavior of the deep policy to be similar to a policy prior, i.e., we regularize in function space. We show that functional regularization yields a bias-variance trade-off, and propose an adaptive tuning strategy to optimize this trade-off. When the policy prior has control-theoretic stability guarantees, we further show that this regularization approximately preserves those stability guarantees throughout learning. We validate our approach empirically on a range of settings, and demonstrate significantly reduced variance, guaranteed dynamic stability, and more efficient learning than deep RL alone.

The operation of multirotors in crowded environments requires a highly reliable takeoff method, as failures during takeoff can damage more valuable assets nearby. The addition of a ballistic launch system imposes a deterministic path for the multirotor to prevent collisions with its environment, as well as increases the multirotor's range of operation and allows deployment from an unsteady platform. In addition, outfitting planetary rovers or entry vehicles with such deployable multirotors has the potential to greatly extend the data collection capabilities of a mission. A proof-of-concept multirotor aircraft has been developed, capable of transitioning from a ballistic launch configuration to a fully controllable flight configuration in midair after launch. The transition is accomplished via passive unfolding of the multirotor arms, triggered by a nichrome burn wire release mechanism. The design is 3D printable, launches from a three-inch diameter barrel, and has sufficient thrust to carry a significant payload. The system has been fabricated and field tested from a moving vehicle up to 50mph to successfully demonstrate the feasibility of the concept and experimentally validate the design's aerodynamic stability and deployment reliability. SUPPLEMENTARY MATERIAL Videos of the experiments: https://youtu.be/sQuKJfllyRM.

Enforcing safety is a key aspect of many problems pertaining to sequential decision making under uncertainty, which require the decisions made at every step to be both informative of the optimal decision and also safe. For example, we value both efficacy and comfort in medical therapy, and efficiency and safety in robotic control. We consider this problem of optimizing an unknown utility function with absolute feedback or preference feedback subject to unknown safety constraints. We develop an efficient safe Bayesian optimization algorithm, StageOpt, that separates safe region expansion and utility function maximization into two distinct stages. Compared to existing approaches which interleave between expansion and optimization, we show that StageOpt is more efficient and naturally applicable to a broader class of problems. We provide theoretical guarantees for both the satisfaction of safety constraints as well as convergence to the optimal utility value. We evaluate StageOpt on both a variety of synthetic experiments, as well as in clinical practice. We demonstrate that StageOpt is more effective than existing safe optimization approaches, and is able to safely and effectively optimize spinal cord stimulation therapy in our clinical experiments.

This work demonstrates dual-arm lifting of bulky objects based on inferred object properties (center of mass (COM) location, weight, and shape) using proprioception (i.e. force torque measurements). Data-driven Bayesian models describe these quantities, which enables subsequent behaviors to depend on confidence of the learned models. Experiments were conducted using the NASA Jet Propulsion Laboratory's (JPL) RoboSimian to lift a variety of cumbersome objects ranging in mass from 7kg to 25kg. The position of a supporting second manipulator was determined using a particle set and heuristics that were derived from inferred object properties. The supporting manipulator decreased the initial manipulator's load and distributed the wrench load more equitably across each manipulator, for each bulky object. Knowledge of the objects came from pure proprioception (i.e. without reliance on vision or other exteroceptive sensors) throughout the experiments.

This paper introduces a new method for inverse reinforcement learning in large state spaces, where the learned reward function can be used to control high-dimensional robot systems and analyze complex human movement. To avoid solving the computationally expensive reinforcement learning problems in reward learning, we propose a function approximation method to ensure that the Bellman Optimality Equation always holds, and then estimate a function to maximize the likelihood of the observed motion. The time complexity of the proposed method is linearly proportional to the cardinality of the action set, thus it can handle large state spaces efficiently. We test the proposed method in a simulated environment on reward learning, and show that it is more accurate than existing methods and significantly better in scalability. We also show that the proposed method can extend many existing methods to large state spaces. We then apply the method to evaluating the effect of rehabilitative stimulations on patients with spinal cord injuries based on the observed patient motions.

Spinal cord stimulation (SCS) has recently enabled humans with motor complete spinal cord injury (SCI) to independently stand and recover some lost autonomic function. However, the nature of the recovered motor activity and the interplay between SCS and motor training are not well understood. Understanding the effect of stand training and spinal stimulation on motor activity during bipedal standing is important for designing spinal rehabilitation therapies that seek to combine spinal stimulation and rehabilitative robots. In this study, we examined electromyography (EMG) data gathered from two SCI patients and six healthy subjects as they attempted standing. We analyzed the muscle activation patterns and EMG waveform shape to quantify both the changes in SCI patient motor activity with training, and the differences between healthy motor activity and SCI patient motor activity under stimulation. We also looked for correlations between the similarity in SCI patients' motor activity to healthy subjects and their overall standing ability. We found that good standing in SCI patients does not emulate healthy standing muscle activity. Furthermore, patient stand training heavily influenced motor activation patterns, but not in ways that improved standing ability. These results indicate that current training techniques do not optimally influence motor activity, and robotic rehabilitation strategies for SCI patients should target essential features of motor activity to optimize functional performance, rather than emulate healthy activity.

This paper develops new methods to design high performance rotary series elastic actuator springs for robotics applications. The approach is based on a spring arm mathematical model that was previously introduced by the authors. The key contribution is the development of an optimization-based design method which maximizes the springs' overall torque density through optimization of the arm profile. An improved analysis algorithm allows for rapid torsional loading response simulation with possible internal contacts between the spring arms. The proposed design and analysis algorithms are validated through FEA and prototype mechanical testing.

Multifinger caging offers a robust object grasping approach. However, while efficient computation of two-finger caging grasps is well developed, the computation of three-finger caging grasps has remained a challenge. This paper considers the caging of polygonal objects with three-finger hands which maintain an equilateral triangle formation during the grasping process. While the c-space of such hands is 4-D, their contact space that represents all two and three finger contacts along the grasped object's boundary forms a 2-D stratified manifold. The paper describes a caging graph that can be constructed in the hand's relatively simple contact space. Starting from a desired immobilizing grasp of the polygonal object, the caging graph can be readily searched for the largest finger opening that maintains a three-finger cage about the object to be grasped. Any equilateral finger placement within the corresponding caging regions guarantees robust object grasping.

We consider sequential decision making under uncertainty, the optimization over large decision space with noisy comparative feedback. This problem can be formulated as a K-armed Dueling Bandits problem where K is the total number of decisions. When K is very large, existing dueling bandits algorithms suffer huge cumulative regret before converging on the optimal arm. This paper studies the dueling bandits problem with a large number of dependent arms. Our problem is motivated by a clinical decision making process in large decision space. We propose an efficient algorithm CorrDuel for the problem which makes decisions to simultaneously deliver effective therapy and explore the decision space. Many sequential decision making problems with large and structured decision space could be facilitated by our algorithm. After evaluated the fast convergence of CorrDuel in analysis and simulation experiments, we applied it on a live clinical trial of therapeutic spinal cord stimulation. It is the first applied algorithm towards spinal cord injury treatments and experimental results show the effectiveness and efficiency of our algorithm.

High performance legged and mobile manipulating robotic platforms require light weight actuators with high torque density, efficiency, and accuracy. This paper introduces a new type of high reduction bearingless gearbox which is based on a gearbox with double-row planetary pinion, and can achieve high reduction ratios in a single composite stage. We discuss the practical issues which limit the applicability of the existing double row gearbox, and motivate our innovations. Several advantages of the bearingless gearbox over current approaches in terms of robustness, possible load distribution, manufacturability, and assembly are described. In the concept, all gear components float unconstrained, which is achieved by introducing an additional kinematic constraint that allows planets to be the same, in which case the planet carrier can be substituted with a secondary sun gear. The resulting planetary gearbox can be readily integrated into compact robotic joints. Its few lightweight components can be manufactured with high accuracy with standard machining techniques.

This paper is concerned with the analysis, design, and prototyping of rotary planar springs for robotics applications such as rotary series elastic actuators, or mechanical couplings. The key contribution is the development of a mathematical model, based on curved beam theory, that allows rapid design, analysis, and optimization of rotary springs that have arbitrary arm shape. The paper also introduces methods to reduce the spring mass via composite arm structures, or arm cutouts. A prototype is designed, analyzed and tested to demonstrate the validity of the model.

Spinal cord stimulation has enabled humans with motor complete spinal cord injury (SCI) to independently stand and recover some lost autonomic function. Quantifying the quality of bipedal standing under spinal stimulation is important for spinal rehabilitation therapies and for new strategies that seek to combine spinal stimulation and rehabilitative robots (such as exoskeletons) in real time feedback. To study the potential for automated electromyography (EMG) analysis in SCI, we evaluated the standing quality of paralyzed patients undergoing electrical spinal cord stimulation using both video and multi-channel surface EMG recordings during spinal stimulation therapy sessions. The quality of standing under different stimulation settings was quantified manually by experienced clinicians. By correlating features of the recorded EMG activity with the expert evaluations, we show that multi-channel EMG recording can provide accurate, fast, and robust estimation for the quality of bipedal standing in spinally stimulated SCI patients. Moreover, our analysis shows that the total number of EMG channels needed to effectively predict standing quality can be reduced while maintaining high estimation accuracy, which provides more flexibility for rehabilitation robotic systems to incorporate EMG recordings.

The classical definition and analysis of force closure in robotic grasping focuses only on the fingertip bodies and their contacts with the grasped object. However, the kinematic properties of the hand mechanism can have a non-trivial effect on the actual type of force closure which can be realized at a given grasp. This paper takes a new look at this classical problem, and introduces technical results on force closure, or wrench resistance, which incorporate the hand mechanism's kinematics in this important grasp security measure. Based on a decomposition of the finger contact forces into canonical subspaces, the paper introduces a division of the finger contact forces into active and passive forces, as well as a dual division into resistant and internal contact forces. The paper develops new formal definitions of these concepts, and describes their mutual orthogonality relationships. Based on these definitions, the paper introduces new theorems of wrench resistant grasps which factor the hand mechanism's structure into the analysis of a given grasp. Examples illustrate how the hand mechanism's structure affects its ability to maintain secure wrench resistant grasps.

Barycentric spherical robots (BSRs) rely on a noncollocated center of mass and center of rotation for propulsion. Unique challenges inherent to BSRs include a nontrivial correlation between internal actuation, momentum, and net vehicle motion. A new method is presented for deriving reduced dynamical equations of motion (EOM) for a general class of BSRs which extends and synthesizes prior efforts in geometric mechanics. Our method is an extension of the BKMM approach [1], allowing Lagrangian reduction and reconstruction to be applied to dynamical systems with symmetry-breaking potential energies, such as those encountered by BSRs rolling on a surface. The resulting dynamical equations are of minimal dimension and vehicle motion due to actuation and momenta appear linearly in a simple first-order differential equation. The EOM of a BSR named Moball [2] [3] are derived to illustrate the approach's utility. A simple table summarizes our algorithm's application to popular BSRs in the literature, and the approach is extended to sloped terrains.

Recent proliferation of cyber-physical systems, ranging from autonomous cars to nuclear hazard inspection robots, has exposed several challenging research problems on automated fault detection and recovery. This paper considers how recently developed formal synthesis and model verification techniques may be used to automatically generate information-seeking trajectories for anomaly detection. In particular, we consider the problem of how a robot could select its actions so as to maximally disambiguate between different model hypotheses that govern the environment it operates in or its interaction with other agents whose prime motivation is a priori unknown. The identification problem is posed as selection of the most likely model from a set of candidates, where each candidate is an adversarial Markov decision process (MDP) together with a linear temporal logic (LTL) formula that constrains robot-environment interaction. An adversarial MDP is an MDP in which transitions depend on both a (controlled) robot action and an (uncontrolled) adversary action. States are labeled, thus allowing interpretation of satisfaction of LTL formulae, which have a special form admitting satisfaction decisions in bounded time. An example where a robotic car must discern whether neighboring vehicles are following its trajectory for a surveillance operation is used to demonstrate our approach.

We consider sequential decision problems under uncertainty, where we seek to optimize an unknown function from noisy samples. This requires balancing exploration (learning about the objective) and exploitation (localizing the maximum), a problem well-studied in the multi-armed bandit literature. In many applications, however, we require that the sampled function values exceed some prespecified "safety" threshold, a requirement that existing algorithms fail to meet. Examples include medical applications where patient comfort must be guaranteed, recommender systems aiming to avoid user dissatisfaction, and robotic control, where one seeks to avoid controls causing physical harm to the platform. We tackle this novel, yet rich, set of problems under the assumption that the unknown function satisfies regularity conditions expressed via a Gaussian process prior. We develop an efficient algorithm called SafeOpt, and theoretically guarantee its convergence to a natural notion of optimum reachable under safety constraints. We evaluate SafeOpt on synthetic data, as well as two real applications: movie recommendation, and therapeutic spinal cord stimulation.

This paper studies cages of convex polygonal objects using three point fingers. The fingers are said to cage the object when it is impossible to move the object arbitrarily far from its initial placement without penetrating the fingers. We consider a three-parameter model of the relative position of the fingers, which gives complete generality for three point fingers in the plane. We consider robustness of caging grasps - what variations in the relative position of the fingers is allowed without breaking the cage. Using a simple decomposition of free space around the polygon, we present an algorithm which gives all caging placements of the fingers and a characterization of the robustness of these cages, albeit at the cost of significant computational complexity.

Moball is a wind-driven spherical robot equipped with sensors for in-situ observation of scientifically important and windy environments, e.g., the Earth's polar regions, Mars, and Saturn's moon Titan. More importantly, Moball incorporates an internal triaxial set of linear electromagnetic generators which can be used to harvest wind energy for long-duration self-sustained operation, or to bias its' wind-driven motions as a form of steering. This paper describes our process to optimize the design of a coreless tubular linear generator for Moball so as to improve energy generation and motion control capabilities with the minimal moving generator mass. The performance of three different types of movers was analyzed with the help of finite element analysis. We determined a final optimized structure and its' dimensions involving a single dipole PM and novel slope-shaped back-irons. A prototype of a single-axis linear generator with a length of 0.8 m was fabricated and assembled. Drop and rotating tests were performed to measure the generated power with this machine. The maximum generated power in the rotating test was 1.05 W at 19 rpm when the load resistance was 40 Ω. The experimental results agreed well with our model predictions. The paper concludes with an overview of the current Moball prototype and ongoing work. The design process developed in this paper can serve as a guideline for future design of energy scavenging systems for robots.

The use of the cognitive capabilties of humans to help guide the autonomy of robotics platforms in what is typically called “supervised-autonomy” is becoming more commonplace in robotics research. The work discussed in this paper presents an approach to a human-in-the-loop mode of robot operation that integrates high level human cognition and commanding with the intelligence and processing power of autonomous systems. Our framework for a “Supervised Remote Robot with Guided Autonomy and Teleoperation” (SURROGATE) is demonstrated on a robotic platform consisting of a pan-tilt perception head, two 7-DOF arms connected by a single 7-DOF torso, mounted on a tracked-wheel base. We present an architecture that allows high-level supervisory commands and intents to be specified by a user that are then interpreted by the robotic system to perform whole body manipulation tasks autonomously. We use a concept of “behaviors” to chain together sequences of “actions” for the robot to perform which is then executed real time.

This paper considers how a multi-limbed robot can carry out manipulation tasks involving simultaneous and compatible end-effector velocity and force goals, while also maintaining quasi-static stance stability. The formulation marries a local optimization process with an assumption of a compliant model of the environment. For purposes of illustration, we first develop the formulation for a single fixed based manipulator arm. Some of the basic kinematic variables we previously introduced for multi-limbed robot mechanism analysis in [1] are extended to accomodate this new formulation. Using these extensions, we provide a novel definition for static equilibrium of multi-limbed robot with actuator limits, and provide general conditions that guarantee the ability to apply arbitrary end-effector forces. Using these extended definitions, we present the local optimization problem and its solution for combined manipulation and stance. We also develop, using the theory of strong alternatives, a new definition and a computable test for quasi-static stance feasibility in the presence of manipulation forces. Simulations illustrate the concepts and method.

Multifinger caging offers a rigorous and robust object grasping approach. Focusing on two-finger caging, this paper describes an algorithm for finding all two-finger cage formations of planar polygonal objects based on contact-space formulation. The paper shows that two-finger cages have several useful properties in contact space. First, the critical points of the cage representation in the hand's configuration space appear as critical points of the interfinger distance function in contact space. Second, these critical points can be graphically characterized directly on the object's boundary. Third, contact space admits a natural rectangular decomposition such that all critical points lie on the rectangle boundaries, and the sublevel sets of contact space and free space are topologically equivalent. These properties lead to a caging graph that can be readily constructed in contact space. Starting from a desired immobilizing grasp of a polygonal object, the caging graph is searched for the minimal, intermediate, and maximal caging regions surrounding the immobilizing grasp. An example constructed from real-world data illustrates and validates the method.

Multi-finger caging offers a robust approach to grasping. This paper describes an algorithm to find caging formations of a 3D polyhedron for two point fingers using a lower-dimensional contact-space formulation. The paper shows that contact space has several useful properties. First, the critical points of the cage in the hand's configuration space are identical to the critical points of the interfinger distance in contact space. Second, contact space can be naturally decomposed into 4D regions having useful properties. A geometric analysis of the critical points of the interfinger distance function results in a catalog of grasps in which the cages change topology, leading to a simple test to classify critical points. These properties lead to an easily constructed caging graph whose nodes contain the critical points of contact space. Starting from an immobilizing grasp, this graph can be searched to find local, intermediate, and maximal caging regions around that initial grasp. An implemented algorithm demonstrates the method.

This paper considers the design and optimization of an autonomous electromechanical control and energy scavenging system for the wind-propelled Moball, a spherical mobile sensor platform concept [1, 2]. This mechanism converts mechanical motion to electrical energy, and the same mechanism can function as an actuator to self-generate motion. Simulations of a simplified model on flat ground show that a 2m diameter Moball operating in typical Arctic conditions can generate 1. 8-2.7W of power continuously while being wind-propelled. We also demonstrate a simple motion control algorithm, showing that self-propulsion in windless conditions requires 1-1.5W. Hence, using this mechanism, a Moball can self-generate sufficient energy for long duration missions involving self-propulsion, sensing, and communication in harsh, cold, windy climates (e.g., Polar regions on Earth, or the surface of Titan or Mars) where solar energy may be limited. Simulations with key design parameters are also used to draw general conclusions regarding optimal design for energy recovery. The addition of springs inside the generating mechanism greatly increases the range of wind speeds over which Moball can harvest energy.

This paper proposes a new method for rigid body pose estimation based on spectrahedral representations of the tautological orbitopes of SE(2) and SE(3). The approach can use dense point cloud data from stereo vision or an RGB-D sensor (such as the Microsoft Kinect), as well as visual appearance data as input. The method is a convex relaxation of the classical pose estimation problem, and is based on explicit linear matrix inequality (LMI) representations for the convex hulls of SE(2) and SE(3). Given these representations, the relaxed pose estimation problem can be framed as a robust least squares problem with the optimization variable constrained to these convex sets. Although this formulation is a relaxation of the original problem, numerical experiments indicates that it is indeed exact - i.e. its solution is a member of SE(2) or SE(3) - in many interesting settings. We additionally show that this method is guaranteed to be exact for a large class of pose estimation problems.

This paper provides kinematic analysis and local motion planning methods for multi-limbed robots. In particular, we consider combined stance and reach tasks for robotic mechanisms whose limbs can be used either as legs or manipulator arms. An example of such a system is the RoboSimian robot participating in the DARPA Robotics Challenge (Figure 1). We develop relationships which model the key quasi-statics and kinematics of these mechanisms: the stance map, the stance Jacobian, and the reach Jacobian, as well as the stance constrained center-of-mass Jacobian. We also introduce characterizations of multi-limbed mechanism configurations in terms of the properties of these maps: local dexterity and limberness. This paper also introduces local planning methods which seek to balance the motion of legs, body, and arms of such mechanisms so as to realize manipulation goals while also maintaining awareness of stance stability issues. Examples with a simple planar model illustrate the methods.

This paper presents a new methodology to craft navigation functions for nonlinear systems with stochastic uncertainty. The method relies on the transformation of the Hamilton-Jacobi-Bellman (HJB) equation into a linear partial differential equation. This approach allows for optimality criteria to be incorporated into the navigation function, and generalizes several existing results in navigation functions. It is shown that the HJB and that existing navigation functions in the literature sit on ends of a spectrum of optimization problems, upon which tradeoffs may be made in problem complexity. In particular, it is shown that under certain criteria the optimal navigation function is related to Laplace's equation, previously used in the literature, through an exponential transform. Further, analytical solutions to the HJB are available in simplified domains, yielding guidance towards optimality for approximation schemes. Examples are used to illustrate the role that noise, and optimality can potentially play in navigation system design.

Many real applications in experimental design need to make decisions online. Each decision leads to a stochastic reward with initially unknown distribution. New decisions are made based on the observations of previous rewards. To maximize the total reward, one needs to solve the tradeoff between exploring different strategies and exploiting currently optimal strategies. This kind of tradeoff problems can be formalized as Multi-armed bandit problem. We recommend strategies in series and generate new recommendations based on noisy rewards of previous strategies. When the reward for a strategy is difficult to quantify, classical bandit algorithms are no longer optimal. This paper, studies the Multi-armed bandit problem with feedback given as a stochastic rank list instead of quantified reward values. We propose an algorithm for this new problem and show its optimality. A real application of this algorithm on clinical treatment is helping paralyzed patient to regain the ability to stand on their own feet.

This paper introduces a tactile or contact method whereby an autonomous robot equipped with suitable sensors can choose the next sensing action involving touch in order to accurately localize an object in its environment. The method uses an information gain metric based on the uncertainty of the object's pose to determine the next best touching action. Intuitively, the optimal action is the one that is the most informative. The action is then carried out and the state of the object's pose is updated using an estimator. The method is further extended to choose the most informative action to simultaneously localize and estimate the object's model parameter or model class. Results are presented both in simulation and in experiment on the DARPA Autonomous Robotic Manipulation Software (ARM-S) robot.

This paper develops an estimation framework for sensor-guided dual-arm manipulation of a rigid object. Using an unscented Kalman Filter (UKF), the approach combines both visual and kinesthetic information to track both the manipulators and object. From visual updates of the object and manipulators, and tactile updates, the method estimates both the robot's internal state and the object's pose. Nonlinear constraints are incorporated into the framework to deal with the an additional arm and ensure the state is consistent. Two frameworks are compared in which the first framework run two single arm filters in parallel and the second consists of the augment dual arm filter with nonlinear constraints. Experiments on a wheel changing task are demonstrated using the DARPA ARM-S system, consisting of dual Barrett- WAM manipulators.

This paper develops a technique for an autonomous robot endowed with a manipulator arm, a multi-fingered gripper, and a variety of sensors to manipulate a known, but poorly observable object. We present a novel grasp planning method which not only incorporates the potential collision between object and manipulator, but takes advantage of this interaction. The natural uncertainty and difficulties in observation in such tasks is modeled as a Partially Observable Markov Decision Process (POMDP). Recent advances in point-based methods as well as a novel state space representation specific to the grasping problem are leveraged to overcome state space growth issues. Simulation results are presented for the combined localization, manipulation, and grasping of a small nut on a table.

Several potentially important science targets have been observed in extreme terrains (steep or vertical slopes, possibly covered in loose soil or granular media) on other planets. Robots which can access these extreme terrains will likely use tethers to provide climbing and stabilizing force. To prevent tether entanglement during descent and subsequent ascent through such terrain, a motion planning procedure is needed. Abad-Manterola, Nesnas, and Burdick [1] previously presented such a motion planner for the case in which the geometry of the terrain is known a priori with high precision. Their algorithm finds ascent/descent paths of fixed homotopy, which minimizes the likelihood of tether entanglement. This paper presents an extension of the algorithm to the case where the terrain is poorly known prior to the start of the descent. In particular, we develop new results for how the discovery of previously unknown obstacles modifies the homotopy classes underlying the motion planning problem. We also present a planning algorithm which takes the modified homotopy into account. An example illustrates the methodology.

This paper considers motion planning and control problems that are motivated by the design of tethered, extreme terrain robots. We abstract the mobility structure of these systems using a tethered differential drive robot with rimless wheels. We analyze several important issues related to this geometry. First it is shown that this vehicle cannot be modeled deterministically unless an additional degree of freedom relative to the standard differential drive vehicle is provided. The simplest kinematically consistent model is one that allows for slight prismatic motion of the axle, approximating the effects of wheel slip. We show that under mild assumptions, such a vehicle's reachable set is dense in SE(2), implying local maneuverability. Next we study some of the constraints which the tether places on the vehicle's motions and derive scaling laws relating wheel and vehicle speeds. Using these results, we provide simple planning and approximate path-following methods that allow tether management. In particular, we consider trajectories produced by solving an optimal control problem to minimize the integral of absolute tether-reeling rate.

Based on a novel contact-space formulation, this paper presents a new algorithm to find two-fingered caging grasps of planar polygonal objects. We show that the caging problem has several useful properties in contact space. First, the critical points of the cage representation in the hand's configuration space appear as critical points of an inter-finger distance function in contact space. Second, the critical points of this distance function can be simply characterized. Third, the contact space admits a rectangular decomposition where the distance function is convex in each rectangle, and all critical points lie on the rectangle boundaries. This property leads to a natural “caging graph,” which can be readily searched to construct the caging sets. An example, constructed from real-world data illustrates and validates the method.

In robotic finger gaiting, fingers continuously manipulate an object until joint limitations or mechanical limitations periodically force a switch of grasp. Current approaches to gait planning and control are slow, lack formal guarantees on correctness, and are generally not reactive to changes in object geometry. To address these issues, we apply advances in formal methods to model a gait subject to external perturbations as a two-player game between a finger controller and its adversarial environment. High-level specifications are expressed in linear temporal logic (LTL) and low-level control primitives are designed for continuous kinematics. Simulations of planar manipulation with our synthesized correct-by-construction gait controller demonstrate the benefits of this approach.

This paper develops an estimation framework for sensor-guided manipulation of a rigid object via a robot arm. Using an unscented Kalman Filter (UKF), the method combines dense range information (from stereo cameras and 3D ranging sensors) as well as visual appearance features and silhouettes of the object and manipulator to track both an object-fixed frame location as well as a manipulator tool or palm frame location. If available, tactile data is also incorporated. By using these different imaging sensors and different imaging properties, we can leverage the advantages of each sensor and each feature type to realize more accurate and robust object and reference frame tracking. The method is demonstrated using the DARPA ARM-S system, consisting of a Barrett™WAM manipulator.

Many manipulation planning problems involve several related sub-problems, such as the selection of grasping points on an object, choice of hand posture, and determination of the arm's configuration and evolving trajectory. Traditionally, these planning sub-problems have been handled separately, potentially leading to sub-optimal, or even infeasible, combinations of the individually determined solutions. This paper formulates the combined problem of grasp contact selection, grasp force optimization, and manipulator arm/hand trajectory planning as a problem in optimal control. That is, the locally optimal trajectory for the manipulator, hand mechanism, and contact locations are determined during the pre-grasping, grasping, and subsequent object transport phase. Additionally, a barrier function approach allows for non-feasible grasps to be optimized, enlarging the region of convergence for the algorithm. A simulation of a simple planar object manipulation task is used to illustrate and validate the approach.

This paper presents a model based approach to autonomous dexterous manipulation, developed as part of the DARPA Autonomous Robotic Manipulation (ARM) program. The developed autonomy system uses robot, object, and environment models to identify and localize objects, and well as plan and execute required manipulation tasks. Deliberate interaction with objects and the environment increases system knowledge about the combined robot and environmental state, enabling high precision tasks such as key insertion to be performed in a consistent framework. This approach has been demonstrated across a wide range of manipulation tasks, and in independent DARPA testing archived the most successfully completed tasks with the fastest average task execution of any evaluated team.

This paper considers the problem of synthesizing correct-by-construction robotic controllers in environments with uncertain but fixed structure. “Environment” has two notions in this work: a map or “world” in which some controlled agent must operate and navigate (i.e., evolve in a configuration space with obstacles); and an adversarial player that selects continuous and discrete variables to try to make the agent fail (as in a game). Both the robot and the environment are subjected to behavioral specifications expressed as an assume-guarantee linear temporal logic (LTL) formula. We then consider how to efficiently modify the synthesized controller when the robot encounters unexpected changes in its environment. The crucial insight is that a portion of this problem takes place in a metric space, which provides a notion of nearness. Thus if a nominal plan fails, we need not resynthesize it entirely, but instead can “patch” it locally. We present an algorithm for doing this, prove soundness (correctness of output), and demonstrate it on an example gridworld.

We investigate the local motion of a planar rigid body with unilateral constraints in the neighborhood of a two-contact frictional equilibrium configuration on a slope. A new sufficient condition of Lyapunov stability is developed in the presence of arbitrary external forces. Additionally, we construct an example, which is stable against perturbations by infinitesimal forces, but does not possess Lyapunov stability against infinitesimal displacements or impulses. The great difference between previous stability criteria and ours leads to further questions about the nature of the exact stability condition.

In this paper, we propose a formulation of the spatial search problem, where a mobile searching agent seeks to locate a stationary target in a given search region or declare that the target is absent. The objective is to minimize the expected time until this search decision of target's presence (and location) or absence is made. Bayesian update expressions for the integration of observations, including false-positive and false-negative detections, are derived to facilitate both theoretical and numerical analyses of various computationally efficient (semi-)adaptive search strategies. Closed-form expressions for the search decision evolution and analytic bounds on the expected time to decision are provided under assumptions on search environment and/or sensor characteristics. Simulation studies validate the probabilistic search formulation and comparatively demonstrate the effectiveness of the proposed search strategies.

This paper presents a strategy for planning robot motions in dynamic, uncertain environments (DUEs). Successful and efficient robot operation in such environments requires reasoning about the future evolution and uncertainties of the states of the moving agents and obstacles. A novel procedure to account for future information gathering (and the quality of that information) in the planning process is presented. To approximately solve the stochastic dynamic programming problem that is associated with DUE planning, we present a partially closed-loop receding horizon control algorithm whose solution integrates prediction, estimation, and planning while also accounting for chance constraints that arise from the uncertain locations of the robot and obstacles. Simulation results in simple static and dynamic scenarios illustrate the benefit of the algorithm over classical approaches. The approach is also applied to more complicated scenarios, including agents with complex, multimodal behaviors, basic robot-agent interaction, and agent information gathering.

This paper considers the motion planning problem that arises when a tethered robot descends and ascends steep obstacle-strewn terrain. This work is motivated by the Axel tethered robotic rover designed to provide access to extreme extra-planetary terrains. Motion planning for this type of rover is very different from traditional planning problems because the tether geometry under high loading must be considered during the planning process. Furthermore, only round-trip paths that avoid tether entanglement are viable solutions to the problem. We present an algorithm for tethered robot motion planning on steep terrain that reduces the likelihood that the tether will become entangled during descent and ascent of steep slopes. The algorithm builds upon the notion of the shortest homotopic tether path and its associated sleeve. We provide a simple example for purposes of illustration.

This paper develops a method to fuse stereo vision, force-torque sensor, and joint angle encoder measurements to estimate and track the location of a grasped object within the hand. We pose the problem as a hybrid systems estimation problem, where the continuous states are the object 6D pose, finger contact location, wrist-to-camera transform and the discrete states are the finger contact modes with the object. This paper develops the key measurement equations that govern the fusion process. Experiments with a Barrett Hand, Bumblebee 2 stereo camera, and an ATI omega force-torque sensor validate and demonstrate the method.

Obstacle avoidance, and by extension collision checking, is a basic requirement for robot autonomy. Most classical approaches to collision-checking ignore the uncertainties associated with the robot and obstacle's geometry and position. It is natural to use a probabilistic description of the uncertainties. However, constraint satisfaction cannot be guaranteed, in this case, and collision constraints must instead be converted to chance constraints. Standard results for linear probabilistic constraint evaluation have been applied to probabilistic collision evaluation; however, this approach ignores the uncertainty associated with the sensed obstacle. An alternative formulation of probabilistic collision checking that accounts for robot and obstacle uncertainty is presented which allows for dependent object distributions (e.g., interactive robot-obstacle models). In order to efficiently enforce the resulting collision chance constraints, an approximation is proposed and the validity of this approximation is evaluated. The results presented here have been applied to robot-motion planning in dynamic, uncertain environments.

This paper presents the Axel robotic rover which has been designed to provide robust and flexible access to extreme extra-planetary terrains. Axel is a lightweight 2-wheeled vehicle that can access steep slopes and negotiate relatively large obstacles due to its actively managed tether and novel wheel design. This paper reviews the Axel system and focuses on its novel paddle wheel characteristics. We show that the paddle design has superior rock climbing ability. We also adapt basic terramechanics principles to estimate the sinkage of paddle wheels on loose sand. Experimental comparisons between the transport efficiency of mountain bike wheels and paddle wheels are summarized. Finally, we present an unfolding wheel prototype which allows Axel to be compacted for efficient transport.

This paper presents an algorithm for human presence detection and tracking using an Ultra-Wideband (UWB) impulse-based mono-static radar. UWB radar can complement other human tracking technologies, as it works well in poor visibility conditions. UWB electromagnetic wave scattering from moving humans forms a complex returned signal structure which can be approximated to a specular multi-path scattering model (SMPM). The key technical challenge is to simultaneously track multiple humans (and non-humans) using the complex scattered waveform observations. We develop a multiple-hypothesis tracking (MHT) framework that solves the complicated data association and tracking problem for an SMPM of moving objects/targets. Human presence detection utilizes SMPM signal features, which are tested in a classical likelihood ratio (LR) detector framework. The process of human detection and tracking is a combination of the MHT method and the LR human detector. We present experimental results in which a mono-static UWB radar tracks human and non-human targets, and detects human presence by discerning human from moving non-human objects.

This paper presents a method whereby an autonomous mobile robot can search for a 3-dimensional (3D) object using an on-board stereo camera sensor mounted on a pan-tilt head. Search efficiency is realized by the combination of a coarse-scale global search coupled with a fine-scale local search. A grid-based probability map is initially generated using the coarse search, which is based on the color histogram of the desired object. Peaks in the probability map are visited in sequence, where a local (refined) search method based on 3D SIFT features is applied to establish or reject the existence of the desired object, and to update the probability map using Bayesian recursion methods. Once found, the 6D object pose is also estimated. Obstacle avoidance during search can be naturally integrated into the method. Experimental results obtained from the use of this method on a mobile robot are presented to illustrate and validate the approach, confirming that the search strategy can be carried out with modest computation.

This paper presents an approach to sensor planning for simultaneous pose estimation and model identification of a moving object using a stereo camera sensor mounted on a mobile base. For a given database of object models, we consider the problem of identifying an object known to belong to the database and where to move next should the object not be easily identifiable from the initial viewpoint. No constraints on the motion of the object nor the robot itself are assumed, which is an improvement on previous methods. Sensor planning is based on the selection of the control action that optimizes a cost metric based on information gain. Experimental results from the implementation of the method on a two-wheeled nonholonomic robot are presented to illustrate and validate the method.

This paper presents a strategy for planning robot motions in dynamic, cluttered, and uncertain environments. Successful and efficient operation in such environments requires reasoning about the future system evolution and the uncertainty associated with obstacles and moving agents in the environment. This paper presents a novel procedure to account for future information gathering (and the quality of that information) in the planning process. After first presenting a formal Dynamic Programming (DP) formulation, we present a Partially Closed-loop Receding Horizon Control algorithm whose approximation to the DP solution integrates prediction, estimation, and planning while also accounting for chance constraints that arise from the uncertain location of the robot and other moving agents. Simulation results in simple static and dynamic scenarios illustrate the benefit of the algorithm over classical approaches.

This paper introduces an algorithm for tracking targets whose locations are inferred from clusters of observations. This method, which we call MHTC, expands the traditional multiple hypothesis tracking (MHT) hypothesis tree to include model hypotheses - possible ways the data can be clustered in each time step - as well as ways the measurements can be associated with existing targets across time steps. We present this new hypothesis framework and its probability expressions and demonstrate MHTC's operation in a robotic solution to tracking neural signal sources.

Consider the task of searching a region for the presence or absence of a target using a team of multiple searchers. This paper formulates this search problem as a sequential probabilistic decision, which enables analysis and design of efficient and robust search control strategies. Imperfect detections of the target's possible locations are made by each search agent and shared with teammates. This information is used to update the evolving decision variable which represents the belief that the target is present in the region. The sequential decision-theoretic formulation presented in this paper provides an analytic framework to evaluate team search systems, as it includes a performance metric (time until decision), a measure of uncertainty (decision confidence thresholds) and imperfect information gathering (detection error). Strategies for cooperative search are evaluated in this context, and comparisons between homogeneous and hybrid search strategies are investigated in numerical studies.

We present a set of potential function components to assist an automated or semi-automated vehicle in navigating a multi-lane, populated highway. The resulting potential field is constructed as a superposition of disparate functions for lane- keeping, road-staying, speed preference, and vehicle avoidance and passing. The construction of the vehicle avoidance potential is of primary importance, incorporating the structure and protocol of laned highway driving. Particularly, the shape and dimensions of the potential field behind each obstacle vehicle can appropriately encourage control vehicle slowing and/or passing, depending on the cars' velocities and surrounding traffic. Hard barriers on roadway edges and soft boundaries between navigable lanes keep the vehicle on the highway, with a preference to travel in a lane center.

This paper presents a miniature robot device and control algorithm that can autonomously position electrodes in cortical tissue for isolation and tracking of extracellular signals of individual neurons. Autonomous electrode positioning can significantly enhance the efficiency and quality of acute electrophysiolgical experiments aimed at basic understanding of the nervous system. Future miniaturized systems of this sort could also overcome some of the inherent difficulties in estabilishing long-lasting neural interfaces that are needed for practical realization of neural prostheses. The paper describes the robot's design and summarizes the overall structure of the control system that governs the electrode positioning process. We present a new sequential clustering algorithm that is key to improving our system's performance, and which may have other applications in robotics. Experimental results in macaque cortex demonstrate the validity of our approach.

This paper considers the stability of an object supported by several frictionless contacts in a potential field such as gravity. The bodies supporting the object induce a partition of the object's configuration space into strata corresponding to different contact arrangements. Stance stability becomes a geometric problem of determining whether the object's configuration is a local minimum of its potential energy function on the stratified configuration space. We use stratified Morse theory to develop a generic stance stability test that has the following characteristics. For a small number of contacts - less than three in 2D and less than six in 3D - stance stability depends both on surface normals and surface curvature at the contacts. Moreover, lower curvature at the contacts leads to better stability. For a larger number of contacts, stance stability depends only on surface normals at the contacts. The stance stability test is applied to quasi-static locomotion planning in two dimensions. The region of stable center-of-mass positions associated with ak-contact stance is characterized. Then, a quasi-static locomotion scheme for a three-legged robot over a piecewise linear terrain is described. Finally, friction is shown to provide robustness and enhanced stability for the frictionless locomotion plan. A full maneuver simulation illustrates the locomotion scheme.

This paper presents the search problem formulated as a decision problem, where the searcher decides whether the target is present in the search region, and if so, where it is located. Such decision-based search tasks are relevant to many research areas, including mobile robot missions, visual search and attention, and event detection in sensor networks. The effect of control strategies in search problems on decision-making quantities, namely time-to-decision, is investigated in this work. We present a Bayesian framework in which the objective is to improve the decision, rather than the sensing, using different control policies. Furthermore, derivations of closed-form expressions governing the evolution of the belief function are also presented. As this framework enables the study and comparison of the role of control for decision-making applications, the derived theoretical results provide greater insight into the sequential processing of decisions. Numerical studies are presented to verify and demonstrate these results

This paper aims at acquiring robust rotation, scale, and translation-invariant feature from a space-variant image by a fovea sensor. A proposed model of eccentricity compensator corrects deformation that occurs in a log-polar image when the fovea sensor is not centered at a target, that is, when eccentricity exists. An image simulator in discrete space remaps a compensated log-polar image using this model. This paper proposes unreliable feature omission (UFO) that reduces local high frequency noise in the space-variant image using discrete wavelet transform. It discards coefficients when they are regarded as unreliable based on digitized errors of the input image. The first simulation mainly tests geometric performance of the compensator, in case without noise. This result shows the compensator performs well and its root mean square error (RMSE) changes only by up to 2.54 [%] in condition of eccentricity within 34.08[deg]. The second simulation applies UFO to the log-polar image remapped by the compensator, taking its space-variant resolution into account. The result draws a conclusion that UFO performs better in case with more white Gaussian noise (WGN), even if the resolution of the compensated log-polar image is not isotropic.

This paper presents a new method to optimally combine motion measurements provided by proprioceptive sensors, with relative-state estimates inferred from feature-based matching. Two key challenges arise in such pose tracking problems: 1) the displacement estimates relate the state of the robot at two different time instants and 2) the same exteroceptive measurements are often used for computing consecutive displacement estimates, a process that renders the errors in these correlated. We present a novel stochastic cloning Kalman filtering (SC-KF) estimation algorithm that successfully addresses these challenges, while still allowing for efficient calculation of the filter gains and covariances. The proposed algorithm is not intended to compete with simultaneous localization and mapping (SLAM) approaches. Instead, it can be merged with any extended-Kalman- filter-based SLAM algorithm to increase its precision. In this respect, the SC-KF provides a robust framework for leveraging additional motion information extracted from dense point features that most SLAM algorithms do not treat as landmarks. Extensive experimental and simulation results are presented to verify the validity of the proposed method and to demonstrate that its performance is superior to that of alternative position-tracking approaches.

This paper introduces a new "assist-as-needed" (AAN) training paradigm for rehabilitation of spinal cord injuries via robotic training devices. In the pilot study reported in this paper, nine female adult Swiss-Webster mice were divided into three groups, each experiencing a different robotic training control strategy: a fixed training trajectory (fixed group, A), an AAN training method without interlimb coordination (Band Group, B), and an AAN training method with bilateral hind-limb coordination (Window Group, C). Fourteen days after complete transection at the mid-thoracic level, the mice were robotically trained to step in the presence of an acutely administered serotonin agonist, quipazine, for a period of six weeks. The mice that received AAN training (Groups B and C) show higher levels of recovery than Group A mice, as measured by the number, consistency, and periodicity of steps realized during testing sessions. Group C displays a higher incidence of alternating stepping than Group B. These results indicate that this training approach may be more effective than fixed trajectory paradigms in promoting robust post-injury stepping behavior. Furthermore, the constraint of interlimb coordination appears to be an important contribution to successful training

This paper presents a decentralized motion planning algorithm for the distributed sensing of a noisy dynamical process by multiple cooperating mobile sensor agents. This problem is motivated by localization and tracking tasks of dynamic targets. Our gradient-descent method is based on a cost function that measures the overall quality of sensing. We also investigate the role of imperfect communication between sensor agents in this framework, and examine the trade-offs in performance between sensing and communication. Simulations illustrate the basic characteristics of the algorithms

This paper presents a multi-scale point and line based representation of two-dimensional range scan data. The techniques are based on a multi-scale Hough transform and a tree representation of the environment's features. The multi-scale representation can lead to improved robustness and computational efficiencies in basic operations, such as matching and correspondence, that commonly arise in many localization and mapping procedures. For multi-scale matching and correspondence we introduce a chi2criterion that is calculated from the estimated variance in position of each detected line segment or point. This improved correspondence method can be used as the basis for simple scan-matching displacement estimation, as a part of a SLAM implementation, or as the basis for solutions to the kidnapped robot problem. Experimental results (using a Sick LMS-200 range scanner) show the effectiveness of our methods

This paper defines wide angle foveated (WAF) imaging. A proposed model combines Cartesian coordinate system, a log-polar coordinate system, and a unique camera model composed of planar projection and spherical projection for all-purpose use of a single imaging device. The central field-of-view (FOV) and intermediate FOV are given translation-invariance and, rotation and scale-invariance for pattern recognition, respectively. Further, the peripheral FOV is more useful for camera's view direction control, because its image height is linear to an incident angle to the camera model's optical center point. Thus, this imaging model improves its usability especially when a camera is dynamically moved, that is, overt-attention. Moreover, simulation results of image extraction show advantages of the proposed model, in view of its magnification factor of the central FOV, accuracy of scale-invariance and flexibility to describe other WAF vision sensors

This paper revisits the multi-robot boundary coverage problem in which a group of k robots must inspect every point on the boundary of a 2-dimensional environment. We focus on the case in which revision of the original inspection plan may be necessary due to changes in the robot team size or the environment. Building upon prior work, which presented a graph-based approach to path planning for this problem, we present a graph representation of the task that is greatly reduced in complexity and a path revision algorithm appropriate for addressing such changes

This paper describes a fixture planning method that minimizes object deflection under external loads. The method takes into account the natural compliance of the contacting bodies and applies to two-dimensional and three-dimensional quasirigid bodies. The fixturing method is based on a quality measure that characterizes the deflection of a fixtured object in response to unit magnitude wrenches. The object deflection measure is defined in terms of frame-invariant rigid body velocity and wrench norms and is therefore frame invariant. The object deflection measure is applied to the planning of optimal fixture arrangements of polygonal objects. We describe minimum-deflection fixturing algorithms for these objects, and make qualitative observations on the optimal arrangements generated by the algorithms. Concrete examples illustrate the minimum deflection fixturing method. Note to Practitioners-During fixturing, a workpiece needs to not only be stable against external perturbations, but must also stay within a specified tolerance in response to machining or assembly forces. This paper describes a fixture planning approach that minimizes object deflection under applied work loads. The paper describes how to take local material deformation effects into account, using a generic quasirigid contact model. Practical algorithms that compute the optimal fixturing arrangements of polygonal workpieces are described and examples are then presented.

This paper develops a formal connection between the power dissipation method (PDM) and Lagrangian mechanics, with specific application to robotic systems. Such a connection is necessary for understanding how some of the successes in motion planning and stabilization for smooth kinematic robotic systems can be extended to systems with frictional interactions and overconstrained systems. We establish this connection using the idea of a multiple-model system, and then show that multiple-model systems arise naturally in a number of instances, including those arising in cases traditionally addressed using the PDM. We then give necessary and sufficient conditions for a dynamic multiple-model system to be reducible to a kinematic multiple-model system. We use this result to show that solutions to the PDM are actually kinematic reductions of solutions to the Euler-Lagrange equations. We are particularly motivated by mechanical systems undergoing multiple intermittent frictional contacts, such as distributed manipulators, overconstrained wheeled vehicles, and objects that are manipulated by grasping or pushing. Examples illustrate how these results can provide insight into the analysis and control of physical systems

This paper considers multiple-contact arrangements where several bodies grasp, fixture, or support an object via frictional point contacts. Within a strictly rigid-body modeling paradigm, when an external wrench (i.e., force and torque) acts on the object, the reaction forces at the contacts are typically indeterminate and span an unbounded linear space. This paper analyzes the contact reaction forces within a generalized quasi-rigid-body framework that keeps the desirable geometric properties of rigid-body modeling, while also including more realistic physical effects. We describe two basic principles that govern the contact mechanics of quasi-rigid bodies. The main result is that for any given external wrench acting on a quasi-rigid object, the statically feasible contact reaction forces lie in a bounded Polyhedral set that depends on the external wrench, the grasp's geometry, and the preload forces. Moreover, the bound does not depend upon any detailed knowledge of the contact mechanics parameters. When some knowledge of the parameters is available, the bound can be sharpened. The polyhedral bound is useful for "robust" grasp and fixture synthesis. Given a set of external wrenches that may act upon an object, the grasp's geometry and preload forces can be chosen such that all of these external wrenches would be automatically supported by the contacts

Intense activity in neural prosthetic research has recently demonstrated the possibility of robotic interfaces that respond directly to the nervous system. The question remains of how the flow of information between the patient and the prosthetic device should be designed to provide a safe, effective system that maximizes the patient’s access to the outside world. Much recent work by other investigators has focused on using decoded neural signals as low-level commands to directly control the trajectory of screen cursors or robotic end-effectors. Here we review results that show that high-level, or cognitive, signals can be decoded from planned arm movements. These results, coupled with fundamental limitations in signal recording technology, motivate an approach in which cognitive neural signals play a larger role in the neural interface. This proposed paradigm predicates that neural signals should be used to instruct external devices, rather than control their detailed movement. This approach will reduce the effort required of the patient and will take advantage of established and on-going robotics research in intelligent systems and human-robot interfaces.

This paper describes a novel miniature robot that can autonomously position recording electrodes inside cortical tissue to isolate and maintain optimal extracellular action potential recordings. The system consists of a novel motorized miniature recording microdrive and a control algorithm. The microdrive was designed for semi-chronic operation and can independently position four electrodes with micron precision over a 5mm range using small (3mm diameter) piezoelectric linear actuators. The autonomous positioning algorithm is designed to detect, align and cluster action potentials, and then command the microdrive to optimize and maintain the neural signal. This system is shown to be capable of autonomous operation in monkey cortex.

This paper introduces the multi-robot boundary coverage problem, wherein a group of k robots must inspect every point on the boundary of a 2-dimensional test environment. Using a simplified sensor model, this inspection problem is converted to an equivalent graph representation. In this representation, the coverage problem can be posed as the k-Rural Postman Problem (kRPP). We present a constructive heuristic which finds a solution to the kRPP, then use that solution to plan the robots’ inspection routes. These routes provide complete coverage of the boundary and also balance the inspection load across the k robots. Simulations illustrate the algorithm’s performance and characteristics.

This paper introduces a machine vision system, which is suitable for cooperative works between the human and computer. This system provides images inputted from a stereo camera head not only to the processor but also to the user’s sight as binocular wide-angle foveated (WAF) information, thus it is applicable for Virtual Reality (VR) systems such as tele-existence or training experts. The stereo camera head plays a role to get required input images foveated by special wide-angle optics under camera view direction control and 3D head mount display (HMD) displays fused 3D images to the user. Moreover, an analog video signal processing device much inspired from a structure of the human visual system realizes a unique way to provide WAF information to plural processors and the user. Therefore, this developed vision system is also much expected to be applicable for the human brain and vision research, because the design concept is to mimic the human visual system. Further, an algorithm to generate features using Discrete Fourier Transform (DFT) for binocular fixation in order to provide well-fused 3D images to 3D HMD is proposed. This paper examines influences of applying this algorithm to space variant images such as WAF images, based on experimental results.

This paper proposes a new synthetic jet actuation concept for small, low speed, highly maneuverable AUVs. Synthetic jet thrusters, which produce jets of vortex rings, are inspired by the pulsatile jet propulsion of salps, jellyfish, and squid. To assess the potential utility of this scheme, we developed synthetic jet actuator prototypes, and verified their function via both force measurement and flow visualization experiments. We used a genetic-algorithm based technique for optimizing the actuation profile of the thrusters. Also presented is an initial discussion of vehicle design. Our conclusion is that synthetic jet thrusters are a viable propulsion method for small underwater vehicles.

Summary form only given. This talk summarized our efforts to develop new technologies whose aim is to improve the quality and quantity of the information derived from extracellular recordings. This work is motivated by ongoing activities at Caltech to develop neural prostheses based on the brain's parietal reach region (PRR). The talk first reviewed our progress towards developing a functioning neural prosthesis in order to motivate the need to develop long-lasting chronic interfaces between electrodes and neurons. The second half of the talk focuses on our efforts to develop a new class of "movable" electrodes that autonomously isolate a neural cell so as to optimize the recorded signal quality, and then maintain optimal signal quality using feedback. Such devices are likely to improve the reliability and robustness of future chronic neural prosthetic systems. We also summarized current research in neural decoding algorithms, whose aim is to extract the maximum information content from the recorded signals.

In this paper we characterize the impact of imperfect communication on the performance of a decentralized mobile sensor network. We first examine and demonstrate the trade-offs between communication and sensing objectives, by determining the optimal sensor configurations when introducing imperfect communication. We further illustrate the performance degradation caused by non-ideal communication links in a decentralized mobile sensor network. To address this, we propose a decentralized motion-planning algorithm that considers communication effects. The algorithm is a cross-layer design based on the proper interface of physical and application layers. Simulation results will show the performance improvement attained by utilizing this algorithm.

We examine the problem of distributed estimation when only one sensor can take a measurement per time step. We solve for the optimal recursive estimation algorithm when the sensor switching schedule is given. We then consider the effect of noise in communication channels. We also investigate the problem of determining an optimal sensor switching strategy. We see that this problem involves searching a tree in general and propose two strategies for pruning the tree to minimize the computation. The first is a sliding window strategy motivated by the Viterbi algorithm, and the second one uses thresholding. The performance of the algorithms is illustrated using numerical examples.

The mechanics of friction and compliance in multi-contact arrangements is key to understanding and predicting grasp stability and dynamic response to external loads. This paper introduces a comprehensive model for the nonlinear force-displacement relationship at a frictional contact. The model is given in an analytic lumped parameter form suitable for on-line grasping applications, and is entirely determined by material and geometric properties of the contacting bodies. The force-displacement law predicts a nonlinear tangential stiffening as the normal load increases. As a result, the composite stiffness matrix of a frictional grasp is asymmetric, indicating that such grasps are not governed by any potential energy. The consequences for grasp stability are investigated. We formulate a rule for preloading frictional grasps which guarantees stable response at the individual contacts. Then we obtain a criterion for selecting contact points which guarantees overall grasp stability. The synthesis rule and its effect on grasp stability is illustrated with a simple 2D example.

This paper computes and analyzes the natural compliance of fixturing and grasping arrangements. Traditionally, linear-spring contact models have been used to determine the natural compliance of multiple contact arrangements. However, these models are not supported by experiments or elasticity theory. We derive a closed-form formula for the stiffness matrix of multiple contact arrangements that admits a variety of nonlinear contact models, including the well-justified Hertz model. The stiffness matrix formula depends on the geometrical and material properties of the contacting bodies and on the initial loading at the contacts. We use the formula to analyze the relative influence of first- and second-order geometrical effects on the stability of multiple contact arrangements. Second-order effects, i.e., curvature effects, are often practically beneficial and sometimes lead to significant grasp stabilization. However, in some contact arrangements, curvature has a dominant destabilizing influence. Such contact arrangements are deemed stable under an all-rigid body model but, in fact, are unstable when the natural compliance of the contacting bodies is taken into account. We also consider the combined influence of curvature and contact preloading on stability. Contrary to conventional wisdom, under certain curvature conditions, higher preloading can increase rather than decrease grasp stability. Finally, we use the stiffness matrix formula to investigate the impact of different choices of contact model on the assessment of the stability of multiple contact arrangements. While the linear-spring model and the more realistic Hertz model usually lead to the same stability conclusions, in some cases, the two models lead to different stability results.

This paper describes an experimental fixturing system wherein fixel reaction forces, workpiece loading, and workpiece displacements are measured during simulated fixturing operations. The system's configuration, its measurement principles, and tests to characterize its performance are summarized. This system is used to experimentally determine the relationship between workpiece displacement and variations in fixel preload force or workpiece loading. We compare the results against standard theories, and conclude that commonly used linear spring models do not accurately predict workpiece displacements, while a non-linear compliance model provides better predictive behavior.

This paper introduces a smooth control algorithm for controlling fully actuated distributed manipulation systems that operate by frictional contact. The control law scales linearly with the number of actuators and is simple to implement. Moreover, we prove that control law has desirable robustness properties in the presence of the nonsmooth mechanics inherent in distributed manipulation systems that rely upon frictional contact. This algorithm has been implemented on an experimental distributed manipulation test-bed, whose structure is briefly reviewed. The experimental results confirm the validity and performance of the algorithm.

This paper describes an experimental modular distributed manipulation system upon which one can implement a variety of control schemes. We have shown elsewhere that when one includes the nonsmooth effects of friction into a model of distributed manipulation, nonsmooth feedback laws must generally be used to control distributed manipulators. We summarize results obtained with this experimental system that confirm the validity of control schemes proposed by the authors in recent papers. We describe the control algorithms in some detail and include specifics of the experimental set-up and experimental results.

This paper presents an algorithm to find the line-based map that best fits sets of two-dimensional range scan data. To construct the map, we first provide an accurate means to fit a line segment to a set of uncertain points via maximum likelihood formalism. This scheme weights each point's influence on the fit according to its uncertainty, which is derived from sensor noise models. We also provide closed-form formulas for the covariance of the line fit, along with methods to transform line coordinates and covariances across robot poses. A Chi-squared based criterion for "knitting" together sufficiently similar lines can be used to merge lines directly (as we demonstrate) or as part of the framework for a line-based SLAM implementation. Experiments using a Sick LMS-200 laser scanner and a Nomad 200 mobile robot illustrate the effectiveness of the algorithm.

This paper considers 2D contact arrangements where several bodies grasp, fixture, or support an object via frictional point contacts. Within a strictly rigid body modelling paradigm, when an external wrench (i.e. force and torque) acts on the object, the reaction forces at the contacts are indeterminate and span an unbounded linear space. This paper analyzes the contact forces within a quasi-rigid body framework that keeps the desirable geometric properties of rigid body modelling, while also includes more realistic physical effects. Using two principles governing the mechanics of quasi-rigid contacts, we show that for any given external wrench acting on the object, the contact forces lie in a bounded polyhedral set. The polyhedral bound depends on the external wrench, the grasp's geometry, and the preload forces. But it does not depend on any detailed knowledge of the contact mechanics parameters. The bound is useful for "robust" grasp and fixture synthesis. Given a collection of external wrenches that may act on an object, the grasp's geometry and preload forces can be chosen such that all of these external wrenches would be automatically supported by the contacts.

Based on a recently developed "generalized averaging theory", we present a generic approach for the design of stabilizing feedback controller for biomimetic locomotive systems. The control laws exponentially stabilize in the average, and they apply to a very wide class of systems. Two examples are given: a "kinematic biped" that demonstrates how our theory handles discontinuities, and the snakeboard, which is an underactuated mechanical system with drift.

Presents a trajectory planning algorithm for mobile robots which may be subject to kinodynamic constraints. Using computational methods from noncommutative harmonic analysis, the algorithm efficiently constructs an approximation to the robot's reachable-state density function. Based on a multiscale approach, the density function is then used to plan a path. One variation of the algorithm exhibits time complexity that is logarithmic in the number of steps. Simulations illustrate the method.

Considers the task of trajectory stabilization for a fish-like robot by means of feedback. We use oscillatory control inputs and apply correction signals at the endpoints of each periodic input signal. Such a strategy can be proven to cause the system to converge to a desired trajectory. We present a specific model of a planar carangiform fish, and verify the stabilization results with simulations and with experiment on a planar robotic fish system that is propelled using carangiform-like movements.

Considers the global exponential stability of planar distributed manipulation control schemes. The programmable vector field approach is a commonly proposed method for distributed manipulation control. The authors (2001) showed that when one takes into account the discreteness of actuator arrays and the mechanics of actuator/object contact, the controls designed by the programmable vector field approach can be unstable at the desired equilibrium configuration. We show here how a discontinuous feedback law that locally stabilizes the manipulated object at the equilibrium can be combined with the programmable vector field approach to control the object's motions. We prove that the combined system is globally exponentially stabilizable even in the presence of changes in contact state. Simulations illustrate the results.

Introduces a "weighted" matching algorithm to estimate a robot's planar displacement by matching two-dimensional range scans. The influence of each scan point on the overall matching error is weighted according to its uncertainty. We develop uncertainty models that account for effects such as measurement noise, sensor incidence angle, and correspondence error. Based on models of expected sensor uncertainty, our algorithm computes the appropriate weighting for each measurement so as to optimally estimate the displacement between two consecutive poses. By explicitly modeling the various noise sources, we can also calculate the actual covariance of the displacement estimates instead of a statistical approximation of it. A realistic covariance estimate is necessary for further combining the pose displacement estimates with additional odometric and/or inertial measurements within a localization framework. Experiments using a Nomad 200 mobile robot and a Sick LMS-200 laser range finder illustrate that the method is more accurate than prior techniques.

Introduces a generalized framework, termed "stochastic cloning," for processing relative state measurements within a Kalman filter estimator. The main motivation and application for this methodology is the problem of fusing displacement measurements with position estimates for mobile robot localization. Previous approaches have ignored the developed interdependencies (cross-correlation terms) between state estimates of the same quantities at different time instants. By directly expressing relative state measurements in terms of previous and current state estimates, the effect of these crosscorrelation terms on the estimation process is analyzed and considered during updates. Simulation and experimental results validate this approach.

We present a general motion planning algorithm for robotic systems with a "stratified" configuration space. Such systems include quasi-static legged robots and kinematic models of object manipulation by finger repositioning. Our method is an extension of a nonlinear motion planning algorithm for smooth systems to the stratified case, where the relevant dynamics are not smooth. The method does not depend upon the number of legs or fingers; furthermore, it is not based on foot placement or finger placement concepts. Examples demonstrate the method.

Considers the design of motion control algorithms for robot fish. We present modeling, control design, and experimental trajectory tracking results for an experimental planar robotic fish system that is propelled using carangiform-like locomotion. Our model for the fish's propulsion is based on quasi-steady fluid flow. Using this model, we propose gaits for forward and turning trajectories and analyze system response under such control strategies. Our models and predictions are verified by experiment.

Analyzes the stability of distributed manipulation control schemes. A commonly proposed method for designing a distributed actuator array control scheme assumes that the system's control action can be approximated by a continuous vector force field. The continuous control vector field idealization must then be adapted to the physical actuator array. However, we show that when one takes into account the discreteness of actuator arrays and realistic models of the actuator/object contact mechanics, the controls designed by the continuous approximation approach can be unstable. For this analysis we introduce and use a "power dissipation" method that captures the contact mechanics in a general but tractable way. We show that the quasi-static contact equations have the form of a switched hybrid system. We introduce a discontinuous feedback law that can produce stability which is robust with respect to variations in contact state.

Conventional nonholonomic motion planning and control theories do not directly apply to "overconstrained vehicles", such as the Sojourner vehicle of the Mars Pathfinder mission. This paper discusses some basic issues of motion planning and control for this potentially important class of mobile robots. A power dissipation approach is used to model the governing equations of overconstrained vehicles that move quasi-statically. These equations are shown to be switched hybrid systems. Notions from standard geometric control, such as the Lie bracket, are extended to these switched systems. We then develop a controllability test for such systems. We explore motion planning primitives in the context of simplified examples.

Analyzes the global stability of distributed manipulation control schemes. The "programmable vector field" approach, which assumes that the system's control actions can be approximated by a continuous vector force field, is a commonly proposed scheme for distributed manipulation control. In practical implementations, the continuous control force field idealization must then be adapted to the specifics of the discrete physical actuator array. However, in Murphey and Burdick (2001) it was shown that when one takes into account the discreteness of actuator arrays and realistic models of the actuator/object contact mechanics, the controls designed by the continuous approximation approach can be unstable at the desired equilibrium configuration. We introduced a discontinuous feedback law that locally stabilizes the manipulated object at the equilibrium. However, the stability of this feedback law only holds in a neighborhood of the equilibrium. In this paper we show how to combine the programmable vector field approach and our local feedback stabilization law to achieve a globally stable distributed manipulation control system. Simulations illustrate the method.

The classical notion of force closure is formulated for multifingered hands, where the fingers actively apply any desired force consistent with friction constraints at the contacts. This paper considers a simpler notion of passive force closure, where each finger obeys some force-displacement law that depends on the finger's joint parameters. The fingers apply initial preload grasping forces, and the grasped object is stabilized against external disturbances by the automatic response of the grasping fingers. After motivating the usefulness of passive force closure, we characterize the conditions for its existence. Then we introduce the passive stability set, defined as the collection of external wrenches that can be passively resisted by a given grasp. We introduce a class of grasp arrangements where the grasping mechanism is compliant while the grasped object is rigid. Such compliant-rigid systems are common, and for these systems the passive closure set can be computed in closed form. Simulation results demonstrate the computation of the passive closure set for two and three-finger planar grasps.

This paper describes a minimalist hopping robot that can perform basic exploration tasks on Mars or other moderate gravity bodies. We show that a single actuator can control the vehicle's jumping and steering operations, as well as the panning of an on-board camera. Our novel thrusting linkage also leads to good system efficiency. The inherent minimalism of our hopping paradigm offers interesting advantages over wheeled and legged mobility concepts for some types of planetary exploration. The paper summarizes the evolutionary development of the system, issues relevant to the design of such jumping systems, and experimental results obtained with system prototypes.

This paper discusses biomechanical issues that are related to the locomotion of a robotic endoscope in the human small intestine. The robot propels itself by pushing against the intestinal walls, much like a pipe crawler. However, the small intestine is not a rigid pipe; and locomotion in it is further complicated by the fact that the bowel is susceptible to damage. With the goal of engineering a safe and reliable machine, the biomechanical properties of the small bowel are studied and related to the mechanics of robotic endoscope locomotion through the small intestine.

This paper presents both formal as well as practical well-definedness conditions for kinematic metric functions. To formulate these conditions, we introduce an intrinsic definition of a rigid body's configuration space. Based on this definition, the principle of objectivity is introduced to derive a formal condition for well-definedness of kinematic metric functions, as well as to gain physical insight into left, right and bi-invariances on the Lie group SE(3). We then relate the abstract notion of objectivity to the more intuitive notion of frame-invariance, and show that frame-invariance can be used as a practical condition for determining objective functions. Examples demonstrate the utility of objectivity and frame-invariance.

This paper studies a form of robotic fish movement that is analogous to the carangiform style of swimming seen in nature. We propose a simple quasi-steady fluid flow model for predicting the thrust generated by the flapping tail. We then describe an experimental system, consisting of a three-link robot, that has been constructed in order to study carangiform-like swimming. Experimental results obtained with this system suggest that the simplified propulsion model is reasonably accurate. The input parameters that realize optimum thrust are experimentally determined. Finally, we consider some issues in maneuvering.

We show how motion planning and control ideas for smooth nonholonomic systems can be extended to legged quasi-static locomotion via the notion of "stratified" configuration spaces and "stratified" control theory. We particularly consider "minimalist" legged systems, which are not well handled by conventional theories based on foot placement. We briefly discuss controllability issues, and then present a motion planning algorithm for stratified systems. The method does not depend upon the number of legs, nor is it based on foot placement concepts.

This paper presents a systematic approach to quantifying the effectiveness of compliant grasps and fixtures of an object. The approach is physically motivated and applies to the grasping of two- and three-dimensional objects by any number of fingers. The approach is based on a characterization of the frame-invariant features of a grasp or fixture stiffness matrix. In particular, we define a set of frame-invariant characteristic stiffness parameters, and provide physical and geometric interpretation for these parameters. Using a physically meaningful scheme to make the rotational and translational stiffness parameters comparable, we define a frame-invariant quality measure, which we call the stiffness quality measure. An example of a frictional grasp illustrates the effectiveness of the quality measure. We then consider the optimal grasping of frictionless polygonal objects by three and four fingers. Such frictionless grasps are useful in high-load fixturing applications, and their relative simplicity allows an efficient computation of the globally optimal finger arrangement. We compute the optimal finger arrangement in several examples, and use these examples to discuss properties that characterize the stiffness quality measure.

With the success of Mars Pathfinder's Sojourner rover, a new era of planetary exploration has opened, with demand for highly capable mobile robots. These robots must be able to traverse long distances over rough, unknown terrain autonomously, under severe resource constraints. This paper reviews issues which are critical for successful autonomous navigation of planetary rovers. We report on the "Wedgebug" algorithm for planetary rover navigation. This algorithm is complete, correct, requires minimal memory for storage of its world model, and uses only on-board sensors, which are guided by the algorithm to efficiently sense only the data needed for motion planning. The implementation of a version of Wedgebug on the Rocky7 Mars rover prototype at the Jet Propulsion Laboratory is described, and experimental results from operation in simulated martian terrain are presented.

Presents an approach to planning minimum-deflection fixtures for tasks whose characteristics are well understood. Based on an accurately defined notion of deflection, we define a quality measure that characterizes the workpiece's deflection with respect to a set of external wrenches determined by the tasks. A scheme is proposed to model task wrenches, which can be used for practical manufacturing operations. This task modelling scheme is then used to obtain a convenient formulation of the task-dependent quality measure, which allows the quality measure to be efficiently computed. An example is presented to show that our approach can be effectively employed for planning compliant fixtures that are best suited to specified tasks.

Motivated by considerations of shape changing propulsion of underwater robotic vehicles, this paper analyses the mechanics of deformable bodies operating in an ideal fluid. The application of methods from geometric mechanics results in a compact and insightful formulation of the problem. We develop an explicit formula for the fluid mechanical connection, in terms of the fluid potential function, for this class of systems. The connection can be used to analyze many issues in motion planning and control. The theory is illustrated by application to an amoeba-like device.

We present a general method for determining controllability of a class of kinematic legged robots. The method is general in that it is independent of the robot's morphology; in particular, it does not depend upon the number of legs. Our method is based on an extension of a nonlinear controllability test for smooth systems to the legged case, where the relevant mechanics are not smooth. Our extension is based on the realization that legged robot configuration spaces are stratified. The result is illustrated with a simple example.

Much prior work in mobile robot path planning has been based on assumptions that are unrealistic for exploration of planetary terrains. Based on the first author's experience with the Mars Pathfinder mission, this paper reviews issues that are critical for successful autonomous navigation of planetary rovers. No currently proposed methodology accurately addresses all of these issues. We report on an extension of the recently proposed "TangentBug" algorithm. The implementation of this extended algorithm on the Rocky 7 Mars Rover prototype at the Jet Propulsion Laboratory is described and experimental results are presented. In addition, limitations encountered by the Sojourner rover in actual Martian terrain suggest that terrain traversability is a key issue for future interplanetary rover autonomous planning algorithms.

This paper presents an approach to planning compliant grasps and fixtures in which the object exhibits minimal deflection under external disturbances. The approach, which applies to general two- and three-dimensional grasps and fixtures represented by any quasi-rigid compliance model, employs a quality measure that characterises the grasped or fixtured object's worst-case deflection caused by disturbing wrenches lying in the unit wrench ball. To ensure well-defined notions of deflection and wrench balls, frame-invariant rigid body velocity and wrench norms are used. As illustrated by its application to fixtures of polygonal objects, our minimum-deflection approach can be effectively applied to planning grasps and fixtures where deflection significantly influences performance.

Presents a control approach for multiple nonholonomic wheeled mobile robots of the Hilare-type to form group formations. To control the formation, each robot has its own coordinate system and it controls its relative positions to its neighboring robots. Particularly, it has a vector called "a formation vector, and the formation is controllable by the vectors. Since the robots have nonholonomic constraints, it is not possible for them to directly move in omni-directions, which means that such nonholonomic vehicles cannot be asymptotically stabilized by smooth static-state feedback control laws. We introduce a smooth time-varying feedback control law whose asymptotic stability is guaranteed in a mathematical framework, averaging theory. The validity of this law is verified by computer simulations.

We present a framework for designing stable control schemes for systems whose dynamics change. The idea is to develop a controller for each of the regions defined by different dynamic characteristics and design a switching scheme that guarantees the stability of the overall system. We derive sufficient conditions for the stability of the switching scheme for systems evolving on a sequence of embedded manifolds. An important feature of the proposed framework is that if the conditions are satisfied by pairs of controllers adjacent in the hierarchy, the overall system will be stable. This makes the application of our results particularly straight forward. The methodology is applied to stabilization of a shimmying wheel, where changes in the dynamic behaviour are due to switches between sliding and rolling.

Using a configuration-space approach, the paper develops a 2nd-order mobility theory for rigid bodies in contact. A major component of this theory is a coordinate invariant 2nd-order mobility index for a body, B, in frictionless contact with finger bodies A/sub 1/,...A/sub k/. The index is an integer that captures the inherent mobility of B in an equilibrium grasp due to second order, or surface curvature, effects. It differentiates between grasps which are deemed equivalent by classical 1st-order theories, but are physically different. We further show that 2nd-order effects can be used to lower the effective mobility of a grasped object, and discuss implications of this result for achieving new lower bounds on the number of contacting finger bodies needed to immobilize an object. Physical interpretation and stability analysis of 2nd-order effects are taken up in the companion paper.

For part I, see ibid., p.696-708. The paper considers how forces are produced by compliance and surface curvature effects in systems where an object a is kinematically immobilized to second-order by finger bodies A/sub l/,...,A/sub k/. A class of configuration-space based elastic deformation models is introduced. Using these elastic deformation models, it is shown that any object which is kinematically immobilized to first or second-order is also dynamically locally asymptotically stable with respect to perturbations. Moreover, it is shown that for preloaded grasps kinematic immobility implies that the stiffness matrix of the grasp is positive definite. The stability result provides physical justification for using second-order effects for purposes of immobilization in practical applications. Simulations illustrate the concepts.

This work considers sensor based motion planning for rod-shaped robots in unknown environments. The motion planning scheme is based on the rod hierarchical generalized Voronoi graph (rod-HGVG). The rod-HGVG is a roadmap for rod-like robots, and is an extension of a prior roadmap for point-like robots. We give an incremental method to construct the rod-HGVG thereby enabling exploration of unknown environments. An important practical feature of the algorithm is its sole reliance upon the use of work space distance measurements to objects that are within line of sight. Such measurements can be readily provided by conventional range sensors. Moreover, motion planning in a configuration space is achieved without explicitly constructing each configuration space obstacle. A key result derived in this paper is the distance gradient between two convex sets.

We present a general trajectory generation scheme for a class of "kinematic" legged robots. The method does not depend upon the number of legs, nor is it based on foot placement concepts. Instead, our method is based on an extension of a nonlinear trajectory generation algorithm for smooth systems to the legged case, where the relevant mechanics are not smooth. Our extension is based on the realization that legged robot configuration spaces are stratified. The algorithm is illustrated with a simple example.

This paper presents a systematic approach for quantifying the quality of compliant grasps. Appropriate tangent and cotangent subspaces to the object's configuration space are studied, from which frame-invariant characteristic compliance parameters are defined. Physical and geometric interpretations are given to these parameters, and a practically meaningful method is proposed to make the parameters comparable. A frame-invariant quality measure is then defined, and grasp optimization using this quality measure is discussed with examples.

This paper presents a method to compute stiffness matrices for compliant grasps and fixtures. While the linear spring contact model has been widely used by robotics researchers, it is in general not accurate for practical applications. More realistic models, including the well-verified Hertz model, are incorporated by use of overlap functions. We derive a stiffness matrix formula that considers surface and material properties of the contacting bodies and applies to both planar and solid grasps. The effects of contact geometry are analyzed and illustrated with examples.

This paper considers the gravitational stability of a frictionless 3-dimensional object in contact with immovable objects. Arbitrarily curved objects are considered. This paper also shows how to determine the region over which the object's center of mass can move while the object maintains a given set of contacts and remains in stable equilibrium. We present symbolic solutions for up to three contacts and discuss numerical solutions for larger numbers of contacts. This analysis has application in planning the motions of quasi-statically walking robots over uneven terrain and the manipulation of heavy objects.

Sensor based planning for rod-shaped robots is necessary for the realistic deployment of noncircular symmetric robots into unknown environments. To this end, the rod hierarchical generalized Voronoi graph (rod-HGVG), introduced in this paper, is a roadmap for rod-like robots. A key feature of this roadmap is that it can be incrementally constructed using distance (range) information. This planning paradigm is an extension of previous work on sensor based planning for point robots.

This paper considers the problem of serpentine, or snake-like, locomotion from the perspective of geometric mechanics. A particular model based on Hirose's active cord mechanism is analyzed. Using the kinematic constraints, we develop a connection, which describes the net motion of the machine as a function of variations in the mechanism's shape variables. We present simulation results demonstrating three types of locomotive gaits, one of which bears an obvious resemblance to the serpentine motion of a snake. We also discuss how these algorithms can be used to optimize certain inputs given the particular choice of physical parameters for a snake robot.

This paper considers the relationship between force and form closure. Based on a previously developed mobility theory, the authors give precise definitions for 1/sup st/ and 2/sup nd/ order form closure for frictionless grasps. The authors also introduce the new concept of 2/sup nd/ order force closure. The authors show for the case of frictionless contacts, a grasp is 1/sup st/ order force closure if and only if it is 1/sup st/ order form closure. The authors further show that a grasp is 2/sup nd/ order force closure if and only if it is also 2/sup nd/ order form closure.

A "modular" robotic system consists of standardized joint and link units that can be assembled into a number of different kinematic configurations. Given a predetermined set of modules, this paper considers the problem of finding an "optimal" module assembly configuration for a specific task. The authors formulate the solution as a discrete optimization procedure. The formulation is based on an assembly incidence matrix representation of a modular robot and a general task-oriented objective function that can incorporate many realistic task criteria. Genetic algorithms (GA) are employed to solve this optimization problem, and a canonical method to represent a modular assembly in terms of genetic strings is introduced. An example involving a 3-DOF manipulator configuration is presented to demonstrate the feasibility of this approach.

This paper prescribes an incremental procedure to construct the generalized Voronoi graph (GVG) and the hierarchical generalized Voronoi graph (HGVG) detailed in the companion paper. The procedure requires only local distance sensor measurements, and therefore the method can be used as a basis for sensor based planning algorithms.

This paper introduces a 1-dimensional network of curves termed the generalized Voronoi graph (GVG) and its extension, the hierarchical generalized Voronoi graph (HGVG), which can be used as a basis for a roadmap or retract-like structure. The GVG and HGVG provide a basis for sensor based path planning in an unknown static environment. In this paper, the GVG and HGVG are defined and some of their properties are exploited to show their utility for motion planning. A companion paper describes how to use the GVG and HGVG for the purposes of sensor based planning.

In both robot grasping and robot locomotion, we wish to hold objects stably in the presence of gravity. We present a derivation of second-order stability conditions for a supported heavy object, employing the tool of Stratified Morse theory. We then apply these general results to the case of objects in the plane.

This paper studies the mechanics of undulatory locomotion. This type of locomotion is generated by a coupling of internal shape changes to external non-holonomic constraints. Employing methods from geometric mechanics, the authors use the dynamic symmetries and kinematic constraints to develop a specialized form of the dynamic equations which govern undulatory systems. These equations are written in terms of physically meaningful and intuitively appealing variables that show the role of internal shape changes in driving locomotion.

This paper develops new lower bounds on the number of frictionless fingers or fixtures which are required to immobilize planar objects. We study in detail the case of objects with smooth boundaries and polygonal objects. Analogous results for the case of piecewise smooth objects follow directly from the analysis presented herein. These results have obvious applications to fixture planning and grasp planning, as we show that it is possible to immobilize objects with fewer fingers than was previously thought possible.

This paper describes the development of a prototype robotic endoscope for gastrointestinal diagnosis and therapy. The goal of this device is to access, in a minimally invasive fashion, the portions of the small intestine that cannot be accessed by conventional endoscopes. This paper describes the macroscopic design and function of the device, and the results of preliminary experiments that validate the concept.

This paper considers the kinematics of hyper-redundant (or "serpentine") robot locomotion over uneven solid terrain, and presents algorithms to implement a variety of "gaits". The analysis and algorithms are based on a continuous backbone curve model which captures the robot's macroscopic geometry. Two classes of gaits, based on stationary waves and traveling waves of mechanism deformation, are introduced for hyper-redundant robots of both constant and variable length. We also illustrate how the locomotion algorithms can be used to plan the manipulation of objects which are grasped in a tentacle-like manner. Several of these gaits and the manipulation algorithm have been implemented on a 30 degree-of-freedom hyper-redundant robot. Experimental results are presented to demonstrate and validate these concepts and our modeling assumptions.

"Hyper-redundant" robots have a very large or infinite degree of kinematic redundancy. This paper develops new methods for determining "optimal" hyper-redundant manipulator configurations based on a continuum formulation of kinematics. This formulation uses a backbone curve model to capture the robot's essential macroscopic geometric features. The calculus of variations is used to develop differential equations, whose solution is the optimal backbone curve shape. We show that this approach is computationally efficient on a single processor, and generates solutions in O(1) time for an N degree-of-freedom manipulator when implemented in parallel on O(N) processors. For this reason, it is better suited to hyper-redundant robots than other redundancy resolution methods. Furthermore, this approach is useful for many hyper-redundant mechanical morphologies which are not handled by known methods.

This paper describes some initial steps towards sensor based path planning in an unknown static environment. The method is a based on a sensor-based incremental construction of a one-dimensional retract of the free space. In this paper we introduce a retract termed the generalized Voronoi graph, and also analyze the roadmap of Canny and Lin's opportunistic path planner (1990, 1993). The bulk of this paper is devoted to the application of nonsmooth analysis to the Euclidean distance function. We show that the distance function is in fact nonsmooth at the points which are required to construct the plan. This analysis leads directly to the incorporation of simple and realistic sensor models into the planning scheme.>

Analysis and simulations are performed for a simplified model of a commercially available variant of the skateboard, known as the Snakeboard. Although the model exhibits basic gait patterns seen in a large number of locomotion problems, the analysis tools currently available do not apply to this problem. The difficulty lies primarily in the way in which the nonholonomic constraints enter into the system. As a first step towards understanding systems represented by their model the authors present the equations of motion and perform some controllability analysis for the snakeboard. The authors also perform numerical simulations of possible gait patterns which are characteristic of snakeboard locomotion.>

Using a configuration-space approach, this paper develops a coordinate invariant 2nd order mobility index for a body, B, in frictionless contact with finger bodies A_1, ..., A_k. The index captures the inherent mobility of B in an equilibrium grasp due to 2nd order, or surface curvature, effects. It differentiates between grasps which are deemed equivalent by the classical 1st order theories, but are physically different. In a companion paper we discuss applications and provide physical justification for using 2nd order immobility effects.

We investigate the contact forces generated by 2/sup nd/ order effects for a body B, in frictionless contact with finger bodies A/sub 1/,..., A/sub k/. A simple paradox shows that rigid body models are inadequate to explain how contact forces are generated by 2/sup nd/ order effects. A class of configuration-space based elastic deformation models are introduced, and are shown to explain the restraining forces produced by surface curvature. Using these elastic deformation models, we prove that any object which is kinematically immobilized to 1/sup st/ or 2/sup nd/ order is also dynamically locally asymptotically stable with respect to perturbations.>

This paper presents novel and efficient kinematic modeling techniques for "hyper-redundant" robots. This approach is based on a "backbone curve" that captures the robot's macroscopic geometric features. The inverse kinematic, or "hyper-redundancy resolution," problem reduces to determining the time varying backbone curve behavior. To efficiently solve the inverse kinematics problem, the authors introduce a "modal" approach, in which a set of intrinsic backbone curve shape functions are restricted to a modal form. The singularities of the modal approach, modal non-degeneracy conditions, and modal switching are considered. For discretely segmented morphologies, the authors introduce "fitting" algorithms that determine the actuator displacements that cause the discrete manipulator to adhere to the backbone curve. These techniques are demonstrated with planar and spatial mechanism examples. They have also been implemented on a 30 degree-of-freedom robot prototype.>

The authors consider how to enumerate the nonisomorphic assembly configurations of a modular robotic system. They introduce an assembly incidence matrix (AIM) to represent a modular robot assembly configuration. Then they use symmetries of the module geometry and graph isomorphisms to define an equivalence relation on the AIMs. Equivalent AIMs represent isomorphic robot assembly configurations. Based on this equivalence relation, the authors propose an algorithm for generating nonisomorphic assembly configurations of an n-link tree-like robot with different joint and link module types. Examples demonstrate that this method offers significant improvement over a brute force enumeration process.

This paper presents a dual-resolution local sensor based planning method for hyper-redundant robot mechanisms. Two classes of sensor feedback control methods, working at different sampling rates and different spatial resolutions, are considered: full shape modification (FSM), and partial shape modification (PSM). FSM and PSM cooperate to utilize a mechanism's hyper-redundancy to enable both local obstacle avoidance and end-effector placement in real-time. These methods have been implemented on a thirty degree of freedom hyper-redundant manipulator which has 11 ultrasonic distance measurement sensors and 20 infrared proximity sensors. The implementation of these algorithms in a dual CPU real-time control computer, an innovative sensor bus architecture, and a novel graphical control interface are described. Experimental results obtained using this test bed show the efficacy of the proposed method.

Two-finger antipodal point grasping of arbitrarily shaped smooth 2-D and 3-D objects is considered. An object function is introduced that maps a finger contact space to the object surface. Conditions are developed to identify the feasible grasping region, F, in the finger contact space. A "grasping energy function", E, is introduced which is proportional to the distance between two grasping points. The antipodal points correspond to critical points of E in F. Optimization and/or continuation techniques are used to find these critical points. In particular, global optimization techniques are applied to find the "maximal" or "minimal" grasp. Further, modeling techniques are introduced for representing 2-D and 3-D objects using B-spline curves and spherical product surfaces.>

A geometric application of screw theory is used to develop a recursive algorithm for computing all singular configurations of revolute-jointed manipulators with arbitrary geometry and an arbitrary number of joints. The depth of the recursion is linear in the number of joints, n, while the computational burden is proportional to 2/sup n-2/. This method does not require explicit construction of the Jacobian matrix elements or a determinant operation. Further, the screw axis of the singular motion is determined at no additional cost. The bifurcations of this algorithm are also explored.>

The authors consider two-finger antipodal point grasping of arbitrarily shaped 2D and 3D objects. An object function which maps a finger contact space to the object surface is introduced. Conditions are developed to identify the feasible grasping region F in the finger contact space. A grasping energy function E is introduced which is proportional to the distance between two grasping points. The antipodal points correspond to critical points of E at F. Optimization and/or continuation techniques are used to find these critical points. In particular, global optimization techniques are applied to find the maximal grasp. Modeling techniques for representing 2D and 3D objects using B-spline curves and spherical product surfaces are described.>

Hyper-redundant robots have a very large or infinite degree of kinematic redundancy. The authors develop methods for determining the optimal configurations which satisfy task constraints while minimizing a weighted measure of mechanism bending and extension. These methods are based on a continuous backbone curve which captures the robot's essential macroscopic geometric features. The calculus of variations is used to develop differential equations whose solution is the optimal backbone curve shape. The optimal distribution of frames along the backbone curve is also considered.>

3R manipulator singularities and geometries based on genericity are categorized. A recursive application of screw theory is used to generate singular configurations and provide a geometric interpretation of nongenericity. A generic manipulator classification scheme based on homotopy class is introduced. Nongeneric geometries are interpreted as bifurcations of generic geometries with respect to kinematic parameter values. Some conjectures on the classes of manipulators which can change pose without passing through a singularity are also given.>

A method for the global configuration resolution for kinematically redundant manipulators is presented. This method is based on an efficient global optimization algorithm which uses a sub-energy tunneling function and terminal repellers. This optimization algorithm is reviewed, and its specialization to redundancy resolution is developed. Applications of this method and comparisons to null-space projection are presented.>

A method is presented for generating inverse kinematic solutions for hyper-redundant manipulators of fixed or variable length. This method uses a continuous backbone curve to capture the macroscopic geometric features of the manipulator. The inverse kinematics of the backbone curve can be used directly to specify the geometry of a wide variety of hyper-redundant manipulator morphologies. The hyper-redundant manipulators are broken nonredundant segments which have closed form inverse kinematic solutions. The kinematic constraints for each segment are specified independently by the backbone curve, and the kinematics of the total manipulator can therefore be solved in parallel. The method is demonstrated with planar and spatial variable geometry truss manipulators.>

Simple schemes for analyzing the kinematics of hyper-redundant robot locomotion over solid terrain are considered. These schemes are based on the concepts of amplitude varying and traveling wave gaits, which are idealized models of inchworm and caterpillar locomotion. The kinematics of these gaits is formulated for hyper-redundant robots of both constant and variable length for locomotion over both flat and irregular terrain. Hyper-redundant locomotion concepts are applied to a novel grasping and fine manipulation scheme based on a grasping wave.>

Discrete dynamical systems theory is applied to the analysis of simplified hopping robot models which are analogous to M.H. Raibert's (1986) experimental machines. A two-dimensional model is presented which includes both forward and vertical hopping dynamics and a foot placement algorithm. These systems are analyzed using a Poincare return map, and hopping behavior is investigated by constructing the return map bifurcation diagrams with respect to system parameters. The bifurcation diagrams exhibit period doubling which results in unexpected hopping behavior.>

Hyper-redundant robots have a large or infinite number of degrees of freedom. Such robots are analogous to snakes or tentacles and are useful for operation in highly constrained environments and novel forms of locomotion. The paper reviews newly developed methods for the kinematic analysis of hyper-redundant manipulators. These methods can be applied to a wide variety of hyper-redundant morphologies and lead to very efficient inverse kinematic, path planning, obstacle avoidance, locomotion, and grasping schemes. It also reviews the design and implementation of a planar 30 degree of freedom variable geometry truss hyper-redundant robot.>

Novel kinematic algorithms for implementing planar hyperredundant manipulator obstacle avoidance is presented. Unlike artificial potential field methods, the method outlined is strictly geometric. Tunnels are defined in a workspace in which obstacles are presented. Methods of differential geometry are then used to formulate equations which guarantee that sections of the manipulator are confined to the tunnels and therefore avoid the obstacles. A general formulation is given with examples to illustrate this approach.>

Discrete dynamical systems theory is applied to the dynamic stability analysis of a simplified hopping robot. A Poincare return map is developed to capture the system dynamics behavior, and two basic nondimensional parameters which influence the systems dynamics are identified. The hopping behavior of the system is investigated by constructing the bifurcation diagrams of the Poincare return map with respect to these parameters. The bifurcation diagrams show a period-doubling cascade leading to a regime of chaotic behavior, where a strange attractor is developed. One feature of the dynamics is that the strange attractor can be controlled and eliminated by tuning an appropriate parameter corresponding to the duration of applied hopping thrust. Physically, the collapse of the strange attractor leads to globally stable uniform hopping motion.>

The author takes a global rather than instantaneous look at the inverse kinematics of redundant manipulators. This approach is based on a manifold mapping reformulation of manipulator kinematics. While the kinematic problem has an infinite number of solutions for redundant manipulators, the infinity of solutions can be grouped into a finite and bounded set of disjoint continuous manifolds. Each of these manifolds, termed self-motion manifolds, physically corresponds to a distinct self-motion of the manipulator, and the number, geometry, and characterizations of the self-motion manifolds are investigated.>

To provide COSMOS, a dynamic model based manipulator control system, with an improved dynamic model, a PUMA 560 arm was disassembled; the inertial properties of the individual links were measured; and an explicit model incorporating all of the non-zero measured parameters was derived. The explicit model of the PUMA arm has been obtained with a derivation procedure comprised of several heuristic rules for simplification. A simplified model, abbreviated from the full explicit model with a 1% significance criterion, can be evaluated with 805 calculations, one fifth the number required by the recursive Newton-Euler method. The procedure used to derive the model is laid out; the measured inertial parameters are presented, and the model is included in an appendix.

This paper presents a method for the generation of efficient manipulator dynamic equations in symbolic form. The efficiency is obtained by the use of simplification rules during the process of equation generation. These simplifications are based on the structure of manipulator dynamics, on simplifications that arise from common manipulator geometries, and from other heuristic simplification rules. This algorithm has been implemented in a lisp-based program, EMDEG (Efficient Manipulator Dynamic Equation Generator). The development of the algorithm, the derivation of simplification rules, and some implementation aspects are discussed; and an example is presented.

The robotics group of the Stanford Artificial Intelligence Laboratory is currently developing a new computational system for robotics applications. Stanford's NYMPH system uses multiple NSC 32016 processors and one MC68010 based processor, sharing a common Intel Multibus. The 32K processors provide the raw computational power needed for advanced robotics applications, and the 68K provides a pleasant interface with the rest of the world. Software has been developed to provide useful communications and synchronization primitives, without consuming excessive processor resources or bus bandwidth. NYMPH provides both large amounts of computing power and a good programming environment, making it an effective research tool.

In this paper we present a unified approach for the control of manipulator motions and active forces based on the operational space formulation. The end-effector dynamic model is used in the development of a control system in which the generalized operational space end-effector forces are selected as the command vector. This formulation provides a framework for natural and efficient integration of both end-effector force and motion control. A "generalized position and force specification matrix" is used for the specification of tasks that involve simultaneous motion and force operations. Flexibility in the force sensor, end-effector, and environment, and problems related to impact are discussed. The real-time operational space control system, COSMOS, has been recently implemented in the NYMPH multiprocessor system. Results of experiments involving contact and force step input response are presented.