Abstract:
Multi-robot systems have been widely studied for extending its capability of accomplishing complex tasks through cooperative behaviors. In large-scale multi-robot behavior mixing, the heterogeneous robotic team executes simultaneously multiple behaviors or sequences of behaviors with various task-prescribed controllers in real time to increase efficiency in parallel tasks. Key to the success of behavior mixing lies in the ability to retain safety (e.g. no inter-robot collisions) and robustness (e.g. dynamic network connectivity, failure of robots) to enable smooth execution and coordination among robots. In this talk, I will present our most recent works on a unified optimization-based control framework synthesizing the original behavior-prescribed multi-robot controllers to generate provably collision-free and robustly connected multi-robot behaviors. The proposed control synthesis framework can be applied to any multi-robot controllers to ensure safety and connectivity in a minimally invasive manner, yielding significantly increased flexibility for large-scale behavior mixing. We will provide simulation results on at least 40 robots executing multiple behaviors to demonstrate the effectiveness and efficiency of the proposed methods.

Committee:
Katia Sycara (advisor)
Maxim Likhachev
George Kantor
Jayanth Krishna Mogali