How can intelligent machines assess their competency to complete a task? This question has come into focus for autonomous systems that algorithmically make decisions under uncertainty. We argue that machine self-confidence—a form of meta-reasoning based on self-assessments of system knowledge about the state of the world, itself, and ability to reason about and execute tasks—leads to many computable and useful competency indicators for such agents. This article presents our body of work, so far, on this concept in the form of the Factorized Machine Self-Confidence (FaMSeC) framework, which holistically considers several major factors driving competency in algorithmic decision-making: outcome assessment, solver quality, model quality, alignment quality, and past experience. In FaMSeC, self-confidence indicators are derived via “problem-solving statistics” embedded in Markov Decision Process solvers and related approaches. These statistics come from evaluating probabilistic exceedance margins in relation to certain outcomes and associated competency standards specified by an evaluator. Once designed, and evaluated, the statistics can be easily incorporated into autonomous agents and serve as indicators of competency. We include detailed descriptions and examples for Markov Decision Process agents and show how outcome assessment and solver quality factors can be found for a range of tasking contexts through novel use of meta-utility functions, behavior simulations, and surrogate prediction models. Numerical evaluations are performed to demonstrate that FaMSeC indicators perform as desired (references to human subject studies beyond the scope of this article are provided).

Human supervision of multiple fielded robots is a challenging task which requires a thoughtful design and implementation of both the underlying infrastructure and the human interface. It also requires a skilled human able to manage the workload and understand when to trust the autonomy, or manually intervene. We present an end-to-end system for human-robot interaction with a heterogeneous team of robots in complex, communication-limited environments. The system includes the communication infrastructure, autonomy interaction, and human interface elements. Results of the DARPA Subterranean Challenge Final Systems Competition are presented as a case study of the design and analyze the shortcomings of the system.

We present our framework Behaviors, Objectives and Binary states for Coordinated Autonomous Tasks (BOBCAT), a multi-agent decision making and task management system for autonomous robots. BOBCAT builds on behavior-based systems and the Belief-Desire-Intention model for decision-making, by using mission objectives as the basis for selecting tasks, or behaviors. We describe how BOBCAT is formulated and present the results of a real-world implementation in the context of exploring austere underground environments in the DARPA Subterranean Challenge.

We present a data-driven framework for strategy synthesis for partially-known switched stochastic systems. The properties of the system are specified using linear temporal logic (LTL) over finite traces (LTLf), which is as expressive as LTL and enables interpretations over finite behaviors. The framework first learns the unknown dynamics via Gaussian process regression. Then, it builds a formal abstraction of the switched system in terms of an uncertain Markov model, namely an Interval Markov Decision Process (IMDP), by accounting for both the stochastic behavior of the system and the uncertainty in the learning step. Then, we synthesize a strategy on the resulting IMDP that maximizes the satisfaction probability of the LTLf specification and is robust against all the uncertainties in the abstraction. This strategy is then refined into a switching strategy for the original stochastic system. We show that this strategy is near-optimal and provide a bound on its distance (error) to the optimal strategy. We experimentally validate our framework on various case studies, including both linear and non-linear switched stochastic systems.

The paper proposes a reliable and robust planning solution to the long range robotic navigation problem in extremely cluttered environments. A two-layer planning architecture is proposed that leverages both the environment map and the direct depth sensor information to ensure maximal information gain out of the onboard sensors. A frontier-based pose sampling technique is used with a fast marching cost-to-go calculation to select a goal pose and plan a path to maximize robot exploration rate. An artificial potential function approach, relying on direct depth measurements, enables the robot to follow the path while simultaneously avoiding small scene obstacles that are not captured in the map due to mapping and localization uncertainties. We demonstrate the feasibility and robustness of the proposed approach through field deployments in a structurally complex warehouse using a micro-aerial vehicle (MAV) with all the sensing and computations performed onboard.

This paper presents a distributed optimization framework and its local utility design for communication-aware information gathering by mobile robotic sensor networks. The main idea of the optimization is that each robot decides based on its local utility that considers the decisions of other neighbor robots higher in a given hierarchy. The local utility is designed as conditional mutual information that captures sensing and communication properties. Sampling procedures using a specific measurement set and particle methods are applied to compute the designed utility, which allows nonlinear, non-Gaussian properties of targets, sensing, and communication. Simulation results describe the presented distributed optimization shows more improved estimates and entropy reduction than another approach that does not consider communication properties. Simulation results also verify the presented distributed optimization using the described approach for information computation has better results than using other approaches that simplify the communication-aware information.

Autonomous navigation in unknown environments with the intent of exploring all traversable areas is a significant challenge for robotic platforms. In this paper, a simple yet reliable method for exploring unknown environments is presented based on bio-inspired reactive control and metric-topological planning. The reactive control algorithm is modeled after the spatial decomposition of wide and small-field patterns of optic flow in the insect visuomotor system. Centering behaviour and small obstacle detection and avoidance are achieved through wide-field integration and Fourier residual analysis of instantaneous measured nearness respectively. A topological graph is estimated using image processing techniques on a continuous occupancy grid. Node paths are rapidly generated to navigate to the nearest unexplored edge in the graph. It is shown through rigorous field-testing that the proposed control and planning method is robust, reliable, and computationally efficient.

This paper presents a method for target localization and tracking in clutter using Bayesian fusion of vision and Radio Frequency (RF) sensors used aboard a small Unmanned Aircraft System (sUAS). Sensor fusion is used to ensure tracking robustness and reliability in case of camera occlusion or RF signal interference. Camera data is processed using an off-the-shelf algorithm that detects possible objects of interest in a given image frame, and the true RF emitting target needs to be identified from among these if it is present. These data sources, as well as the unknown motion of the target, lead to a heavily non-linear non-Gaussian target state uncertainties, which are not amenable to typical data association methods for tracking. A probabilistic model is thus first rigorously developed to relate conditional dependencies between target movements, RF data and visual object detections. A modified particle filter is then developed to simultaneously reason over target states and RF emitter association hypothesis labels for visual object detections. Truth model simulations are presented to compare and validate the effectiveness of the RF + visual data fusion filter.

Drones have carried weapons, drugs, explosives and illegal packages in the recent past, raising strong concerns from public authorities. While existing drone monitoring systems only focus on detecting drone presence, localizing or fingerprinting the drone, there is a lack of a solution for estimating the additional load carried by a drone. In this paper, we present a novel passive RF system, namely DroneScale, to monitor the wireless signals transmitted by commercial drones and then confirm their models and loads. Our key technical contribution is a proposed technique to passively capture vibration at high resolution (i.e., 1Hz vibration) from afar, which was not possible before. We prototype DroneScale using COTS RF components and illustrate that it can monitor the body vibration of a drone at the targeted resolution. In addition, we develop learning algorithms to extract the physical vibration of the drone from the transmitted signal to infer the model of a drone and the load carried by it. We evaluate the DroneScale system using 5 different drone models, which carry external loads of up to 400g. The experimental results show that the system is able to estimate the external load of a drone with an average accuracy of 96.27%. We also analyze the sensitivity of the system with different load placements with respect to the drone's body, flight modes, and distances up to 200 meters.

We consider the problem of real-time path-planning in a spatiotemporally varying wind-field with moving obstacles. We are provided with changing wind and obstacle predictions along a (D + 1)-dimensional space-time lattice. We present an Any-Time algorithm that quickly finds an αβ-suboptimal solution (a path that is not longer than αβ times the optimal time-length), and then improves α and β while planning time remains or until new wind/obstacle predictions trigger a restart. The factor α comes from an α-overestimate of the A*-like cost heuristic. β is proportional to motion modeling error. Any-Time performance is achieved by: (1) improving the connectivity model of the environment from a discrete graph to a continuous cost-field (decreasing β); (2) using the established method of incrementally deflating α. Our method was deployed as the global planner on a fixed-wing unmanned aircraft system that uses Doppler radar and atmospheric models for online real-time wind sensing and prediction. We compare its performance vs. other state-of-the-art methods in simulated environments.

This paper examines the problem of distributed path planning for a mobile sensor network comprised of communication-aware robots performing general information gathering missions. Mutual information is derived for distributed sensing over packet erasure channels that model multi-hop communication. We model distributed path planning as a non-cooperative game and derive utility functions that are optimized locally by each robot. Each robot computes the control input in a distributed manner that results in a combined action that can be bounded by the optimal centralized result by utilizing sub-modularity in certain cases. It is shown that when the communication model includes multi-hop communication to expand the coverage of the sensor network, the property of sub-modularity is lost. We further show that the additional global knowledge required for the local computation of utility functions can be learned by simple consensus approaches. Finally, we discuss a sampling approach to approximate the proposed utility functions in order to reduce the associated computational requirements.

A data ferry enables communication in sparse sensor networks by combining physical movement with wireless relaying of data. Optimizing an unmanned aircraft's motion and communication link scheduling, however, requires knowledge of the communication environments through which it moves. Aspects of the radio frequency environment can be opportunistically learned through the process of communicating with sensor nodes while ferrying, allowing models of the radio environment to be improved. This work analyzes the integration of ferry optimization with using a Gaussian process to learn the radio environment. The unmanned aircraft's trajectory is initially optimized with an a priori model. After flying one period of the trajectory, RF variations observed by the ferry are used to train a Gaussian process and improve the estimate of the environment. Through this iterative process, ferry performance improves rapidly, achieving 80% of optimal within 4 iterations, and 93% after 9 iterations, as the Gaussian process is able to converge rapidly to the true radio frequency environment.

This work presents a decentralized mobility control algorithm for an optimal end-to-end communication chain using a team of unmanned aircraft acting solely as communication relays. The chaining controller drives the location of a virtual control point, using estimates of the communication objective function gradient calculated using stochastic approximation techniques, to locations of improved relaying for unmanned aircraft. The gradient estimate is derived from observation data of the communication objective function taken along a path generated by the aircraft orbiting about the control point. Flight experiments show that an unmanned aircraft can measure the signal-to-noise-and-interference ratio fields from IEEE 802.11b/g (WiFi) communication links; generate estimates of the field gradients using the least-squares gradient estimation method; and use the gradient estimates to drive a control point to a location that improves communication capacity.

This paper presents an aerial robot system designed to perform in situ sampling of supercell thunderstorms, including those that produce tornadoes. A description of the system architecture and concept of operations is given, along with a brief discussion of the evolution of critical system components. Results are given from two field campaigns, an airmass boundary intercept in Northern Colorado during the spring of 2009, and two supercell transects during the Spring 2010 campaign of the VORTEX2 project. The 2010 results represent the first aloft sampling of the rear-flank downdraft in a tornadic supercelluar storm.

This paper presents a vision-based navigation solution for unmanned aircraft operations on airfield surfaces in GPS-denied environments. The unmanned aircraft system ground operations management system (UGOMS) described here combines measurements from a computer vision system and inertial sensors with an airport layout database to provide real-time position determination on the airfield surface. UGOMS provides both absolute position of the aircraft as well as relative position to airport surface elements such as runway hold lines and taxiway edges. The key technical components of UGOMS are computer vision algorithms that classify image regions, Markov localization using particle filters, and a navigation architecture which incorporates the localization information. An overview of the overall UGOMS architecture is presented as well as preliminary test results using an uncertain airfield map to highlight the performance capabilities of the system.

This paper presents a system for in situ atmospheric sensing using an aerial robot system in the presence of a strong wind field. The geostatistical concept of the variogram is used to characterize regions of the environment with high variability, which is assumed to correlate with scientific interest. After regions of interest are identified, ordered upwind methods are used to generate feasible trajectories in the face of strong background wind. The feasible trajectories are combined with the variogram characterization to select feasible paths that travel through the regions of highest interest. The system is tested in simulation using data from a simulated severe storm.

The problems of area patrol, perimeter surveillance, and target tracking are combined into a single reconnaissance framework. This integration is based on application of ordered upwind methods to track the propagation of the target position boundary. The ordered upwind methods provide a computationally efficient method for computing the target track boundary over a discretized nonuniform mesh of the environment. Planning methods based on the target track boundary are discussed. Simulation results demonstrate the use of ordered upwind methods for reconnaissance tasks.

This paper presents a hierarchical control architecture that enables cooperative surveillance by a heterogeneous aerial robot network comprised of mothership unmanned aircraft and daughtership micro air vehicles. Combining the endurance, range, and processing capabilities of the motherships with the stealth, flexibility, and maneuverability of swarms of daughterships enables robust control of aerial robot networks conducting collaborative operations. The hierarchical control structure decomposes the system into components that take advantage of the abilities of the different types of vehicles. The motherships act as distributed databases, fusion centers, negotiation agents, and task supervisors while daughtership control is achieved using cooperative vector field tracking. This paper describes the overall architecture and then focuses on the assignment and tracking algorithms used once sub- teams of daughtership vehicles have been deployed. A summary of the communication, command, and control structure of a heterogeneous unmanned aircraft system is also given in this paper along with hardware-in-the-loop and software simulation results verifying several components of the distributed control architecture.

This paper presents a new task assignment algorithm that integrates area search and target tracking. A new tracking metric is derived that encodes the ability of each robot to reach an unsensed target before the uncertainty in its position passes a given upper bound. Target activation and tracking assignment using this method enable broader participation by individual robots in a cooperative surveillance architecture. Area search is added to the cooperative surveillance architecture through the use of a coordinated coverage map that represents the probability of new targets being detected in regions of the environment. A key feature of the coverage map is that it is updated by all robots, not just robots performing area search. Thus, tracking robots aid the search process. Simulation results validate the new integrated assignment process.

This paper presents an approach to approximate information content for active sensing tasks. The unscented transform is used to represent probability distributions by a set of representative sample points that capture the first and second moments of the distribution. Using these sample points, the effects of nonlinear operators on a probability distribution of active sensing costs can be approximated. Simulation results validate the approximation for bearings-only geolocalization of a stationary target and tracking of an uncertain moving target.

This paper explores the role of meshed airborne communication networks in the operational performance of small unmanned aircraft systems. Small unmanned aircraft systems have the potential to create new applications and markets in civil domains, enable many disruptive technologies, and put considerable stress on air traffic control systems. We argue that of the existing networked communication architectures, only meshed ad hoc networking can meet the communication demands for the large number of small aircraft expected to be deployed in future. Experimental results using the heterogeneous unmanned aircraft system are presented to show that meshed airborne communication is feasible, that it extends the operational envelope of small unmanned aircraft at the expense of increased communication variability, and that net-centric operation of multiple cooperating aircraft is possible. Additionally, the ability of airborne networks of small unmanned aircraft to exploit controlled mobility to improve performance is discussed.

This paper presents a coordination scheme for standoff tracking of an uncertain moving target by cooperating mobile robots with range and bearing sensors. An algorithm is described that transforms a guidance vector field developed for loitering about circular patterns into a vector field that tracks ellipses. Specific elliptical tracking patterns are developed based on the covariance matrix of a stochastic estimation process. Simulation results demonstrate the ability of two aerial robots to coordinate their orbit around an uncertain target moving with constant velocity.

The performance of monocular vision based target tracking is a strong function of camera motion. Without motion, the target estimation problem is unsolvable. By designing the camera path, the best possible estimator performance can be achieved. This paper describes a trajectory design method based on the predicted target state error covariance. This method uses a pyramid, breadth-first search algorithm to generate real-time paths that achieve a minimum uncertainty bound in fixed time or a desired uncertainty bound in minimum time.