Validation and Experimentation Toolkit for Spatiotemporal Reasoning (VETS)

To maximize the operational value of robotic systems, warfighters must be able to specify high-level commands to direct their autonomous partner’s activities using multiple interaction methods that effectively convey spatiotemporal goals and constraints. The risks involved with unintended behaviors by autonomous robots due to dissonance in warfighter and spatiotemporal reasoning could be catastrophic to the warfighter’s mission. Therefore, appropriate validation of these command and control (C2) subsystems throughout the robot’s design lifecycle is imperative. Charles River Analytics proposes to design and demonstrate a Validation and Experimentation Toolkit for Spatiotemporal Reasoning (VETS) that enables experimentation and evaluation of spatiotemporal C2 strategies and autonomous reasoning approaches through simulation and monitoring of agents operating in a synthetic environment and receiving commands from a human operator. VETS has three major capabilities: (1) an experiment design capability that simulates autonomous robots operating in synthetic operational environments under varying experimental conditions; (2) a targeted trace capability that provides insight into sources of failure within these experiments, enabling assessment against defined performance metrics of the strengths and weakness of specific human-robot spatiotemporal C2 and reasoning approaches; and (3) integration with existing behavior modeling and human interaction technologies to provide cost-effective and reusable tools to the robotics community.