Abstract:
Information-based search and coverage are important in planetary exploration and disaster response applications. Efficient information acquisition can help with increasing geological understanding or situational awareness. Heterogeneous robots, each with different sensing and motion modalities, can be coordinated to optimize search and coverage in a target region. Information maps, which estimate the importance of visiting various areas within the search region, can guide robots to explore a region more efficiently. However, these information maps can be low-resolution, inaccurate, or incomplete. This work presents methods of using the spectral decomposition of an information map to distribute robotic agents with diverse skill sets to different sub-parts of the overall search and coverage problem, such that each robot is leveraging its unique capabilities to best contribute to the team’s goal. Additionally, this thesis proposes methods of refining the information map over a region as more information is gained and consequently predictively adapting agent-task allocations. Finally, this thesis proposes a method of optimizing team selection alongside agent-task allocation, optimizing both, in turn, to determine the optimal team to search a given region and the best allocation of tasks for the selected agents. The effectiveness of agent-task allocations is assessed using search and coverage metrics, including target discovery time and ergodicity. The planner employed is ergodic trajectory optimization, which directs robots to spend time in areas based on the expected amount of information there, thus balancing exploration and exploitation. Experiments conducted using real-world data sources, including probability distributions of scientifically significant minerals at Cuprite, NV, roadmaps, and flood maps generated by Hurricane Ian, demonstrate the efficacy of the presented agent-task allocation methods.

Thesis Committee Members:
David Wettergreen, Co-chair
Howie Choset, Co-chair
Andrea Bajcsy
Robin Murphy, Texas A&M University
Guillaume Sartoretti, National University of Singapore

More Information