MechE PhD Prospectus Defense: Brennan Brodt

TITLE: Robots in Limbo: Competition and Collaboration in Uncertain Environments

ABSTRACT: The successful deployment of autonomous systems often rely on robotic agents having ac-cess to complete information regarding their environments and objectives. However, in real world scenarios, these comprehensive models are not readily available and must be built over the course of operation. Motivated by applications in wildlife monitoring and disas-ter recovery, we explore how robots can intelligently use limited information to generate behaviors that accomplish their individual and team objectives. We also leverage concepts from multi-objective optimization theory to study how robots can safely and effectively plan over the many mission specifications that may be necessary in real world scenarios. This prospectus focuses these investigations on multi-agent applications with three key sources of environmental uncertainty: unknown objectives, hazards, and obstacles. We first present how agents can strategically infer team objectives by observing other well-behaved agents and, consequently, how to obfuscate these objectives from more malicious actors. Then, we present how Pareto optimality can be used to complete objectives in dangerous environ-ments using only local information gathered by our agents. We also present how a team of agents can exploit differences in sensing and capabilities to explore unknown environments in search of objectives and deliver any necessary services to these locations. Finally, we discuss how our methods can enable agents to accomplish complex objectives in highly uncertain, dangerous, and competitive environments.

COMMITTEE: ADVISOR/CHAIR Professor Alyssa Pierson, ME/SE; Professor Sean Andersson, ME/SE; Professor Andrew Sabelhaus, ME/SE; Professor Roberto Tron, ME/SE