Date: Wednesday, January 31, 2024

Time: 10:00am–12:00noon EST

In-Person Location: Howey N201/N202

Zoom Link: https://gatech.zoom.us/j/99502812564

Kehinde Aina

Robotics PhD Candidate

School of Electrical and Computer Engineering

Georgia Institute of Technology

Committee:

Dr. Daniel I. Goldman (Advisor) - School of Physics, Georgia Institute of Technology

Dr. Sehoon Ha - School of Interactive Computing, Georgia Institute of Technology

Dr. Seth Hutchinson - School of Interactive Computing, Georgia Institute of Technology

Dr. Dana Randall - School of Computer Science, Georgia Institute of Technology

Dr. David Hu - School of Mechanical Engineering, Georgia Institute of Technology

Abstract:

The task of coordinating robot swarms in constrained and crowded environments is often hampered by robots' limited capability to acquire precise state information, crucial for the effective control of the collective to achieve their goals. Such challenges typically stem from inherent uncertainties and unpredictability within these environments. However, recent studies of dense active matter reveal that coordinated behaviors can arise through self-organization principles, where complex global patterns emerge from simple local interactions. This thesis aims to harness the unavoidable features of dense active systems to enable the spontaneous coordination of robot swarms without relying on global control or state information of other robots in the environment. Specifically, we developed an adaptive clog control technique for emergent coordination of robot swarms in narrow passages through unplanned contact or social interactions. Next, we extended the technique to address scenarios where an individual's failure could detrimentally impact group performance. The Active Contact Response (ACR) algorithm was introduced to impart collective fault tolerance within the clog control technique, enabling the repositioning of faulty robots to less obstructive configurations. Additionally, we explored the potential benefits of leveraging stigmergy in addressing the problem of non-stationarity within the multi-agent reinforcement learning framework. Overall, this thesis advances the application of local, social, and physical interactions for the collective coordination of multi-robot systems.