Distributed Learning and Cooperative Control of Multi-Agent Systems

This seminar presents a novel class of resource-constrained multi-agent systems for cooperatively predicting an unknown field of interest to achieve a global goal. A measurable unknown field represents the collection of scalar quantities of interest (such as chemical concentration or biomass of algal blooms) transported via physical processes. The conventional inverse problem approach based on physical transport models is too computationally costly for resource-constrained multi-agent systems. For agents to efficiently predict the field of interest, statistical models using kernel regression and Gaussian processes have been developed. Different navigation strategies for different goals such as prediction and tracing of a field of interest are proposed by exploiting the predictive posterior statistics of spatial prediction. We provide a class of algorithms for distributed learning and cooperative control of a multi-agent system so that a global goal of the overall system is achieved from locally acting agents. Convergence properties of the proposed multi-agent systems using kernel regression were analyzed by the (ODE) approach. Several simulation results demonstrate the effectiveness of the proposed algorithms based on different environmental models. Our scheme provides agents with robust intelligence based on the prediction of an unknown field; and, hence, it allows agents to be versatile for various scenarios in uncertain environments.
Jongeun Choi received his Ph.D. and M.S. degrees in Mechanical Engineering from the University of California at Berkeley in 2006 and 2002 respectively. He also received a B.S. degree in Mechanical Design and Production Engineering from Yonsei University at Seoul, Republic of Korea in 1998. He is currently an Assistant Professor with the Department of Mechanical Engineering at the Michigan State University. Dr. Choi’s research interests include adaptive, learning, distributed and robust control, with applications to unsupervised competitive learning algorithms, self-organizing systems, distributed learning and coordination algorithms for autonomous vehicles, multiple robust controllers and biomedical problems.