Title:  Active Matter to Programmable Swarms: Statistical Physics Approaches to Collective Aggregation

Committee: 

Dr. Randall, Advisor

Dr. Egerstedt, Co-Advisor       

Dr. Hutchinson, Chair

Dr. Howard

Abstract:

The objective of this research is to develop a computational and experimental framework for collective/modular robotic systems that are sensing and communication constrained, by leveraging principles from statistical physics and active matter. Towards this goal, we develop a system of minimalistic analog robots. This robotic testbed is used as an experimental tool to study and analyze distributed stochastic algorithms for collective behavior, which are in turn inspired by studies on statistical physics algorithms. We extend our studies with grid-based particle and Discrete Element Modeling (DEM) simulations. For the scope of this proposal, we focus on a very basic collective behavior problem: aggregation. This basic collective behavior is a prerequisite for most other collective tasks like object transport and manipulation, information transfer, etc. and constitutes one of the most widely studied collective behavior. While some recent effort has been placed in exploiting simple designs of robots for achieving aggregation, most state of the art swarm programming strategies rely on computationally intensive units and models. Here, we explore an overly simplistic approach in achieving collective aggregation by constructing simple robotic models with bare minimum sensing and computation that can leverage mechanics and physical dynamics to accomplish goals. For future explorations, we will combine ideas from the active matter and distributed computing literature to study the role of robot density and persistence length of trajectories in manipulating aggregation dynamics. The proposed line of research is expected to provide us with novel tools for swarm manipulation and control.