Swaminathan Gopalswamy

Multi-agent tracking of multiple moving objects is a well-researched problem in support of many applications. The core sensor fusion algorithms evolve from the Kalman Filter, with different variants on fusion algorithms when multiple sensors are providing similar information. Such fusion is often driven by the uncertainty associated with these sensors; however, such uncertainty is characterized a priori based on the sensor characteristics and potentially a model of the system dynamics. When sensors fail (or even deviate from their presumed characterization), such algorithms also fail or degrade. In this work, we propose to characterize the “health” of individual sensors using a probabilistic health monitor (previously proposed for general cyber-physical systems), as well as the health of the collection of sensors providing observations, and use this characterization to appropriately weight the different sensors when performing sensor fusion. The resultant “Sensor-Health-Aware Resilient Fusion” (SHARF) provides unbiased, consistent, and optimal state estimates. SHARF is applied to the “Infrastructure Enabled Autonomy” (IEA) problem, where sensors mounted on the road-side are used to localize the vehicles on the road. The localization is performed using SHARF on the edge. The proposed SHARF is implemented and evaluated in field experiments with cameras and a RADAR on the infrastructure and with two vehicles on the road driving in real-world scenarios. We demonstrate that SHARF can detect faulty sensory data and incorporate the health status of sensors leading to successful tracking of the target vehicles.

Autonomous navigation in unstructured off-road environments is greatly improved by semantic scene understanding. Conventional image processing algorithms are difficult to implement and lack robustness due to a lack of structure and high variability across off-road environments. The use of neural networks and machine learning can overcome the previous challenges but they require large labeled data sets for training. In our work we propose the use of hyperspectral images for real-time pixel-wise semantic classification and segmentation, without the need of any prior training data. The resulting segmented image is processed to extract, filter, and approximate objects as polygons, using a polygon approximation algorithm. The resulting polygons are then used to generate a semantic map of the environment. Using our framework. we show the capability to add new semantic classes in run-time for classification. The proposed methodology is also shown to operate in real-time and produce outputs at a frequency of 1Hz, using high resolution hyperspectral images.

We present Reshaping Viscoelastic-String Path-Planner a Path Planner that reshapes a desired Global Plan for a Robotic Vehicle based on sensor observations of the Environment. We model the path to be a viscoelastic string with shape preserving tendencies, approximated by a connected series of Springs, Masses, and Dampers. The resultant path is then reshaped according to the forces emanating from the obstacles until an equilibrium is reached. The reshaped path remains close in shape to the original path because of Anchor Points that connect to the discrete masses through springs. The final path is the resultant equilibrium configuration of the Spring-Mass-Damper network. Two key concepts enable RVP (i) Virtual Obstacle Forces that push the Spring-Mass-Damper system away from the original path and (ii) Anchor points in conjunction with the Spring-Mass-Damper network that attempts to retain the path shape. We demonstrate the results in simulation and compare it's performance with an existing Reshaping Local Planner that also takes a Global Plan and reshapes it according to sensor based observations of the environment.

Autonomous vehicles (AVs) are rapidly proliferating. However, their future adoption is dependent on guarantees of their safety and security. Studies have shown that AVs are vulnerable to malicious cyberattacks with potentially lethal consequences. It is, therefore, imperative to deploy defensive techniques to enhance their cybersecurity. The method of dynamic watermarking (DW) has been proposed as an active technique for cyber-physical systems to detect cyberattacks. In this study, DW is implemented on an actual AV. It is shown that the DW technique is effective for timely detection of different types of attacks on the yaw rate measurement under autonomous operation, and thereby, allows recourse to measures such as stopping or pulling off to the side to prevent collisions.

This letter proposes a path planner that reshapes a global path locally in response to sensor-based observations of obstacles in the environment. Two fundamental concepts enable the resultant algorithm (a) a path-following synthetic vehicle whose steering actions are non-myopically optimized to result in a smooth traversible path that meets path curvature constraints, and (b) a path-aware turning moment-field that enables obstacle avoidance while eluding the typical local-minimum-induced stagnation associated with potential field methods. The use of the combination of the two concepts results in a reduced action space over which optimization needs to be performed towards minimizing the path deviation subject to obstacle avoidance, and thus results in an efficient algorithm that can be implemented online. We demonstrate the algorithm in simulations as well as in field experiments, performing real time local path planning and obstacle avoidance on two different vehicle platforms (an ackerman steering vehicle two-axled vehicle, and a differential-steering 4-axled vehicle) in an unstructured off-road terrain.

This paper considers a multiagent, connected, robotic fleet where the primary functionality of the agents is sensing. A distributed multi-sensor control strategy maximizes the value of the collective sensing capability of the fleet, using an information-driven approach. Each agent individually performs sensor processing (Kalman Filtering and Joint Probabilistic Data Association) to identify trajectories (and associated distributions). Using communications with its neighbors the agents enhance the prediction of the trajectories using a Consensus of Information approach that iteratively calculates the Kullback-Leibler average of trajectory distributions, enabling the calculation of the value of the collective information for the fleet. The dynamics of the agents, the evolution of the identified trajectories for each agent, and the dynamics of individual observed objects are captured as a Partially Observable Markov Decision Process (POMDP). Using this POMDP and applying rollout with receding horizon control, an optimized non-myopic control policy that maximizes the collective fleet information value is synthesized. Simulations are performed for a scenario with three heterogeneous UAVs performing coordinated target tracking that illustrate the proposed methodology and compare the centralized approach with a contemporary sequential multiagent distributed decision technique.

Typical numerical solutions to planning problems under time-dependent constraints (like traffic signals) involve searching in time plus state space. We consider a sampled discrete spatial formulation of the vehicle dynamics. This allows us to propose an optimal planning algorithm with much reduced search-space and time complexity, for vehicles moving across signalized intersections with full knowledge of the traffic Signal, Phasing and Timing (SPaT) information. Then we extend these results to partial knowledge by casting the problem as a Markov Decision Process (MDP). The proposed algorithms are demonstrated through numerical simulations that show a five-fold improvement in runtime compared with a standard time-state formulation, while providing comparable improvements in fuel economy with no vehicle dynamic constraints or traffic rules violated.

When harvesting grain crops on large farms, a combine collects the grain while a grain cart transports the grain by commuting between the combine and a semi-trailer parked by the roadside. There are several issues associated with human-operated grain carts: labor shortage and increasing labor cost, operational imprecision and inefficiency as well as safety hazards, all of which can potentially be addressed if grain carts were autonomous. This paper presents a motion planning algorithm and the associated navigation solution for autonomous grain carts. The algorithm features a novel integration of Artificial Potential Field (APF) with Fuzzy Logic Control (FLC). A set of simulation tests were carried out, comparing the proposed APF+FLC planner with a simple APF planner. The test results verified the effectiveness, robustness, and efficiency of the proposed planning algorithm in performing the logistical tasks in harvest operations where unharvested crops were the only obstacles as well as when random static or dynamic obstacles existed. In addition, a set of mobile robot tests implementing the proposed navigation solution were conducted, in which the robot representing the grain cart autonomously accomplished the logistical tasks in the harvest operations, verifying the effectiveness and practicality of the navigation solution.

Cyber physical systems (CPSs) are characterized by behaviors emergent from the interactions of software systems with physical systems. Many complex CPSs utilize “independent monitoring” of the system behavior for any undesirable effects so that appropriate mitigation actions can potentially be taken. Typical independent monitoring systems for CPSs focus only on the violation of specified constraints that are detected “after-the-fact,” and any preemptive attempts to prevent the violation are handled by the core software systems. In this paper, we propose an independent health monitoring mechanism as a human-recognizable measure of trustworthiness of the system. Specifically, this mechanism not only detects actual violations of constraints but also calculates the possibility to violate the constraints. The proposed health monitor has two components: 1) health measurement functions that are specific and tailored to each component of a CPS and 2) a probabilistic health-state determination that is generic and directly derived from Bayesian estimation methods. The implementation and performance of the proposed monitoring mechanism were demonstrated with an autonomous vehicle self-navigating to a free parking space in MATLAB/Simulink simulations, which showed that the mechanism was able to promptly and accurately estimate the health state of the monitored system and help prevent accidents/catastrophes.