Mitchell Colby

Scientific Systems (United States), Oregon State University

Papers

Total Citations

H-Index

About

Mitchell Colby is a researcher specializing in multiagent systems, autonomous robotics, and reinforcement learning, with a particular focus on coordinating teams of robots in complex, real-world environments. His most influential work centers on the challenge of structural credit assignment in tightly coupled multiagent domains — a fundamental problem in getting individual agents to learn behaviors that benefit the collective. His 2016 paper introducing D++, which has garnered 36 citations, represents a significant advance in this area by developing more effective reward-shaping techniques for scenarios where robot actions are deeply interdependent. Colby has also made notable contributions to autonomous space exploration, investigating how multi-robot teams can operate under high uncertainty and communication delays while incorporating high-level human feedback — a practically vital consideration for real mission deployments. His work on local approximations of difference evaluation functions addresses computational barriers that previously limited the scalability of these powerful multiagent reward signals. Across his research, Colby consistently tackles the tension between coordination efficiency and adaptability in dynamic environments, producing work that bridges theoretical reinforcement learning with applied robotics. With over 90 cumulative citations, his contributions have meaningfully shaped the field of cooperative autonomous systems.