Abstract

Modern autonomous systems increasingly require teams of highly heterogeneous agents to work together to achieve a group goal. For example, aerial drones may maintain a scouting formation while guiding ground robots toward a waypoint for a search-and-rescue mission. Controlling such systems requires a framework that accommodates heterogeneity in (i) agent state spaces and dynamics, (ii) local and group objectives and constraints, and (iii) interaction topology. This talk presents the first steps towards such a unified framework based on the theory of cellular sheaves, which frames multi-agent coordination as a local-to-global problem. This framework uniformly models diverse coordination problems over heterogeneous agents and topologies, and sheaf-diffusion-based algorithms are developed to solve these problems via purely local computation and communication. Numerous applications are presented ranging from robot teams to satellite rendezvous to structural modeling.

About the Speaker

Tyler Hanks recently joined Florida Tech as an Assistant Professor of Computer Science after receiving his Ph.D. from the University of Florida and his B.S. from the University of South Florida. Hanks's doctoral research was funded by an NSF Graduate Research Fellowship, and he was honored as a finalist for the UF AI Research Dissertation Award. Hanks's research focuses on the application of programming language theory, category theory, and sheaf theory to improve the design, analysis, and safe control of multi-agent systems, with a particular emphasis on robotic and LLM agent systems. Hanks has presented his research to diverse conferences such as the IEEE Conference on Decision and Control (CDC), SIAM Conference on Mathematics of Data Science, and AMS Joint Mathematics Meetings, and he serves as a reviewer for CDC and IEEE Control Systems Letters.