MATCH-SLAM: A Multi Agent Tasking and Control metHod for Simultaneous Localization and Mapping

Physical Sciences Inc. (PSI) proposes MATCH-SLAM, an innovative"Multiple Agent Tasking and Control metHod for Simultaneous Localization and Mapping"for multi-agent cooperative and autonomous exploration of tunnels, caves, indoor, or other GPS denied environments. Current tunnel mapping efforts using autonomous ground vehicles are limited by platform mobility and power requirements, while aerial approaches currently lack robustness, effective coordination methods, and autonomy. PSI"s design gives specific attention to the complex interactions between heterogeneous agents and how to fuse their disparate sensor types, the constraints of limited on-board computational resources, limited platform mobility, and bandwidth-limited communications. This approach promises to enable efficiently coordinated exploration and mapping by many classes of autonomous systems, including PSI"s own highly capable, highly mobile, low-power, low-cost micro-air vehicle. This is achieved using PSI"s novel planning algorithms which make use of a multi-dimensional cost/score matrix for heterogeneous agent cooperative tasking, specific agent actions behavior patterns tailored to reduce crucial uncertainties, and a state estimator designed to maintain map consistency across agents in the presence of constrained or intermittent communication. PSI proposes to test and demo this technology during the program both in software simulations and on its own micro-air vehicle platforms in representative environments.