Testing and evaluation of field robotic systems requires both experimentation in representative conditions and human supervision to effectively assess components, manage risk, and interpret results. Due to the complexity of robotic sys-tems, we argue this experimentation should be done adaptively by using insights gained from previous trials. Furthermore, we envision an advisory system that could assist experimenters with selecting trial configurations by learning and accounting for human preferences and risk tolerances; however, formal methods for human decision making in the context of field robotic experimentation remains an open question. In this work, we present and analyze a case study for how decisions were made during the testing and evaluation of an off-road, autonomous navigation system. From the perspective of active learning, we find that Bayesian Optimization is a promising mathematical framework for modeling human decision making in adaptive experimental design of field robotics and that a combination of the EI, KG, and PES acquisition functions would likely be useful for realizing an advisory system.

Motion planning in off-road environments re-quires reasoning about both the geometry and semantics of the scene (e.g., a robot may be able to drive through soft bushes but not a fallen log). In many recent works, the world is classified into a finite number of semantic categories that often are not sufficient to capture the ability (i.e., the speed) with which a robot can traverse off-road terrain. Instead, this work proposes a new representation of traversability based exclusively on robot speed that can be learned from data, offers interpretability and intuitive tuning, and can be easily integrated with a variety of planning paradigms in the form of a costmap. Specifically, given a dataset of experienced trajectories, the proposed algorithm learns to predict a distribution of speeds the robot could achieve, conditioned on the environment semantics and commanded speed. The learned speed distribution map is converted into costmaps with a risk-aware cost term based on conditional value at risk (CVaR). Numerical simulations demonstrate that the proposed risk-aware planning algorithm leads to faster average time-to-goals compared to a method that only considers expected behavior, and the planner can be tuned for slightly slower, but less variable behavior. Furthermore, the approach is integrated into a full autonomy stack and demonstrated in a high-fidelity Unity environment and is shown to provide a 30% improvement in the success rate of navigation.

In this work, we introduce Mobile Wireless Infrastructure on Demand: a framework for providing wireless connectivity to multi-robot teams via autonomously reconfiguring ad-hoc networks. In many cases, previous multi-agent systems either assumed the availability of existing communication infrastructure or were required to create a network in addition to completing their objective. Instead our system explicitly assumes the responsibility of creating and sustaining a wireless network capable of satisfying end-to-end communication requirements of a team of agents, called the task team, performing an arbitrary objective. To accomplish this goal, we propose a joint optimization framework that alternates between finding optimal network routes to support data flows between the task agents and improving the performance of the network by repositioning a collection of mobile relay nodes referred to as the network team. We demonstrate our approach with simulations and experiments wherein wireless connectivity is provided to patrolling task agents.

This paper presents an approach and introduces new open-source tools that can be used to evaluate robotic mapping algorithms. Also described is an extensive subterranean mine rescue dataset based upon the DARPA Subterranean (SubT) challenge including professionally surveyed ground truth. Finally, some commonly available approaches are evaluated using this metric.

We develop a hybrid system architecture that enables a team of mobile robots to complete a task in a complex environment by self-organizing into a multihop ad hoc network and solving the concurrent communication and mobility problem. The proposed system consists of a two-layer feedback loop. An outer loop performs infrequent global coordination and a local inner loop determines motion and communication variables. This system provides the lightweight coordination and responsiveness of decentralized systems while avoiding local minima. This allows a team to complete a task in complex environments while maintaining desired end-to-end data rates. The behavior of the system is evaluated in experiments that demonstrate: 1) successful task completion in complex environments; 2) achievement of equal or greater end-to-end data rates as compared to a centralized system; and 3) robustness to unexpected events such as motion restriction.

In this paper, we present a hybrid system architecture that enables a team of robots to self-organize into a multi-hop ad-hoc network allowing for the completion of a given task while providing the desired end-to-end data rates between designated robots. This architecture consists of a two stage feedback loop in which an outer loop provides infrequent global coordination and an inner loop, operating locally, controls the motion and network routing of each robot. The resulting system is able to operate dynamically in complex environments with minimal global coordination as demonstrated through multiple experiments. We conclude with a realistic application of our system, namely patrolling a set of hallways.

In this paper we propose a hybrid architecture that allows a team of mobile robots to self-organize into a multi-hop ad-hoc network and solve the joint mobility and communication problem in complex environments to complete a given task. The system consists of an outer global planning loop and an inner local loop responsible for motion and network routing, arranged in a two-stage feedback system. This system is able to leverage the benefits of previous systems, while avoiding their drawbacks. This results in a lightweight responsive system that is able to operate in complex environments with minimal global coordination while maintaining a minimum end-to-end data rate between robots. Two main benefits of our approach are demonstrated through experimentation superior performance over existing systems and dynamic adjustment to unexpected events. We conclude with a demonstration of the system operating in a realistic scenario, in which the team patrols a set of hallways.

In pursuit of increasing the operational tempo of a ground robotics platform in unknown domains, we consider the problem of predicting the distribution of structural state-estimation error due to poorly-modeled platform dynamics as well as environmental effects. Such predictions are a critical component of any modern control approach that utilizes uncertainty information to provide robustness in control design. We use an online learning algorithm based on matrix factorization techniques to fit a statistical model of error that provides enough expressive power to enable prediction directly from motion control signals and low-level visual features. Moreover, we empirically demonstrate that this technique compares favorably to predictors that do not incorporate this information.

Autonomous platforms are confronted by a diversity of challenges in unstructured environments, which make monitoring performance a non-trivial task. Some of these environments are so complex that they preclude persistent, nearby operator oversight. This absence of oversight motivates the need for an online monitoring system, specifically for robots operating in difficult environments. We develop a test methodology and set of online monitoring metrics by extending methods for characterizing robotic-systems using offline metrics. We implement this test methodology in an unstructured, outdoor environment and show the resulting performance information gained from our online monitoring solution. This online monitoring approach is generalizable such that it characterizes any robotic system that meets our set of hardware and software criteria.

We begin to bridge the gap between high-level motion planning and execution by adopting models to abstract the complicated skid-steer vehicle dynamics and evaluating their suitability as motion predictors for a feed-forward control framework. We consider three kinematic motion models and a drivetrain model in experiments on two surface types with a small tracked vehicle. We perform statistical analysis of the predictive accuracy of these models when used to create optimal open-loop plans for a set of canonical maneuvers and discuss the applicability of these models for a closed-loop control framework.

Providing reliable end-to-end communication for teams of robots requires the integration of novel routing techniques, motion planning algorithms, and transport level communication protocols. In this paper we look at existing robust routing solutions that provide redundancy at the routing layer and develop the Multi-Confirmation Transmission Protocol (MCTP) to take advantage of that redundancy at the transport level. The resulting system that integrates robust routing and MCTP is evaluated in experiments performed in complex environments. The integrated system is observed to provide a robust architecture that allows for near lossless communication while operating in a complex environment with less traffic than standard confirmation protocols.

We propose an architecture and algorithms for maintaining end-to-end network connectivity for autonomous teams of robots. By adopting stochastic models of point-to-point wireless communication and computing robust solutions to the network routing problem, we ensure reliable connectivity during robot movement in complex environments. We fully integrate the solution to network routing with the choice of node positions through the use of randomized motion planning techniques. Experiments demonstrate that our method succeeds in navigating a complex environment while ensuring that end-to-end communication rates meet or exceed prescribed values within a target failure tolerance.

We consider the combined problem of frontier exploration in a complex indoor environment while seeking a radio source. To do this in an efficient manner, we incorporate radio signal strength (RSS) information into the exploration algorithm by locally sampling the RSS and estimating the 2-D RSS gradient. The algorithm exploits the local motion to collect RSS samples for gradient estimation and seeks to explore in a way that brings the robot to the signal source. This strategy avoids random or exhaustive exploration. An indoor experiment demonstrates the exploration algorithm that uses this information to dynamically prioritize candidate frontiers and traverse to a radio source. Simulations, including radio propagation modeling with a ray-tracing algorithm, enable study of control algorithm tradeoffs and statistical performance.

In this paper, a system architecture to provide end-to-end network connectivity for autonomous teams of robots is discussed. The core of the proposed system is a cyber-physical controller whose goal is to ensure network connectivity as robots move to accomplish their assigned tasks. Due to channel quality uncertainties inherent to wireless propagation, we adopt a stochastic model where achievable rates are modeled as random variables. The cyber component of the controller determines routing variables that maximize the probability of having a connected network for given positions. The physical component determines feasible robot trajectories that are restricted to safe configurations which ensure these probabilities stay above a minimum reliability level. Local trajectory planning algorithms are proposed for simple environments and leveraged to obtain global planning algorithms to handle complex surroundings. The resulting integrated controllers are robust in that end-to-end communication survives with high probability even if individual point-to-point links are likely to fail with significant probability. Experiments demonstrate that the global planning algorithm succeeds in navigating a complex environment while ensuring that end-to-end communication rates meet or exceed prescribed values within a target failure tolerance.

This paper presents the locomotion approach of a novel quadruped robot which is able to carry various effectors for achieving manufacturing tasks in large workspaces. Equipped with lockers on some of the passive joints and clamping devices at the end of its limbs, this quadruped uses eight actuators for achieving manufacturing tasks as well as locomotion tasks. In the following sections, we first present the proposed robot and its two working modes. Then, the locking strategy of the robot is formulated as an optimization problem. Also, a practical method for managing the limbs swinging movement is addressed. At last, the presented approach is applied on two concrete examples. Possessing a low degree of kinematic redundancy, the proposed quadruped shows a reasonable locomotion capacity which allows it to achieve locomotion with respect to some extra constrains in its workspaces.

This paper describes a test-bed for planar micro and mesoscale manipulation tasks and a framework for planning based on quasi-static models of mechanical systems with intermittent frictional contacts. We show how planar peg-in-the-hole assembly tasks can be designed using randomized motion planning techniques with Mason's models for quasi-static manipulation. Simulation and experimental results are presented in support of our methodology. We develop this further into a systematic approach to incorporating uncertainty into planning manipulation tasks with frictional contacts. We again consider the canonical problem of assembling a peg into a hole at the mesoscale using probes with minimal actuation but with visual feedback from an optical microscope. We consider three sources of uncertainty. First, because of errors in sensing position and orientation of the parts to be assembled, we must consider uncertainty in the sensed configuration of the system. Second, there is uncertainty because of errors in actuation. Third, there are geometric and physical parameters characterizing the environment that are unknown. We discuss the synthesis of robust planning primitives using a single degree-of-freedom probe and the automated generation of plans for mesoscale manipulation. We show simulation and experimental results of our work.

In this paper we explore methods for the online mapping of received radio signal strength with mobile robots and localizing the source of the radio signal. By utilizing Gaussian processes, we are able to build an online model of the signal-strength map that can, in turn, be used to provide the current maximum likelihood estimate of the source location. Furthermore, using the estimate of the source location, the Gaussian process model allows for prediction of received signal strength with confidence bounds in regions of the environment that have not been explored. Finally, we develop a control law for collecting samples of the signal strength with mobile robots that allows for online estimation of the radio signal source.

Different robotic tasks can be solved by controlling a robot to circulate along curves. These include, for example, border inspection and surveillance, multirobot manipulation, and pattern generation. In a previous, work we have proposed a vector field approach for robot convergence and circulation along time-varying curves embedded in N-dimensional spaces. In the present work we instantiate this approach for three-dimensional spaces and, for the first time, show the efficacy of this method to control actual robots. Besides new theoretical analysis when constant speed control is applied, we present experimental results with aerial (quadrotors) and ground (differential-driven) robot.

We study radio signal propagation in indoor environments using low-power devices leveraging the Zigbee and Bluetooth specifications. We present results from experiments where two robots equipped with radio signal devices and enabled to control and localize autonomously in an indoor hallway and laboratory environment densely sample RSSI at various times over several days. We show that simulated RSSI measurements using existing radio signal models and experimentally gathered RSSI measurements match closely, suggesting that for robotics applications requiring predicted RSSI, low-power radio signal devices are a well-posed sensing modality.

Manipulation and assembly tasks are typically characterized by many nominally rigid bodies coming into frictional contacts, possibly involving impacts. Manipulation tasks are difficult to model because uncertainties associated with friction and assembly tasks are particularly hard to analyze because of the interplay between process tolerance and geometric uncertainties due to manufacturing errors. Manipulation at the meso (hundred microns to millimeters) and micro (several microns to tens of microns) scale is even harder for several reasons. It is difficult to measure forces at the micro-netwon level reliably using off-the-shelf force sensors and good force-feedback control schemes have not proved successful. It is hard to manufacture general-purpose end effectors at this scale and it is even more difficult to grasp and manipulate parts at the micro and meso level than it is at the macro level. Finally, the lack of good models of the mechanics of contact interactions at this scale means that model-based approaches to planning and control are difficult.

We present a decentralized approach to multi- robot manipulation where the team of robots surround and trap an object and transport it, by dragging or pushing, to the goal configuration in an environment with obstacles. The proposed feedback controllers are obtained by sequentially composing vector fields or behaviors and are decentralized in the sense that robots do not exchange each other's state information. Rather, cooperative manipulation is achieved by relying solely on each robot's local information and a global knowledge of the task. We present computer simulations and experimental results obtained using our multi-robot testbed.

This paper discusses the design and current capabilities of a new software tool, dVC, capable of simulating planar systems of bodies experiencing unilateral contacts with friction. Since different problems require different levels of accuracy, dVC provides user-selectable body types (rigid or locally-compliant), motion models (first-order, quasi-static, dynamic), and several state-of-the-art time-stepping methods. One can also choose to include friction between each body and the plane of motion. To support optimal and robust part design, dVC also allows on-the-fly changes to parameters of the geometric and physical models. The results obtained for three representative planar problems are presented: the design of a passive part-orienting device, the planning of a mesoscale assembly operation, and the design of a grasp strategy.

We consider the task of controlling a large team of nonholonomic ground robots with an unmanned aerial vehicle in a decentralized manner that is invariant to the number of ground robots. The central idea is the development of an abstraction for the team of ground robots that allows the aerial platform to control the team without any knowledge of the specificity of individual vehicles. This happens in much the same way as a human operator can control a single robot vehicle by simply commanding the forward and turning velocities without a detailed knowledge of the specifics of the robot. The abstraction includes a gross model of the shape of the formation of the team and information about the position and orientation of the team in the plane. We derive controllers that allow the team of robots to move in formation while avoiding collisions and respecting the abstraction commanded by the aerial platform. We provide simulation and experimental results using a team of indoor mobile robots and a three-dimensional, cable-controlled, parallel robot which serves as our indoor unmanned aerial platform.

This paper describes a test-bed for planar micro manipulation tasks and a framework for planning based on quasi-static models of mechanical systems with frictional contacts. We show how planar peg-in-the-hole assembly tasks can be designed using randomized motion planning techniques with Mason's models for quasi-static manipulation. Finally, we present simulation and experimental results in support of our methodology