Dr. Daniel Zelazo
Faculty of Aerospace Engineering, Israel Institute of Technology, Haifa
Wednesday, July 16, 2014, 16:00 - 17:30
Room 01-016, Georges-Koehler-Allee 101, Freiburg 79110, Germany
Many applications of multi-agent systems assume that all agents in the system are cooperating to achieve a common goal. From an optimization perspective, these systems are attempting to minimize some global objective function in a distributed manner. Other applications, however, may require that each agent in the system behaves selfishly to achieve some local desired objective, but is constrained by certain team goals. In this scenario there must be a compromise between what each individual agent considers optimal and the constraints imposed on the entire team. In this talk, we explore this second scenario and describe a solution method we term the shrinking horizon preference agreement algorithm that allows each agent to distributedly and in real-time negotiate their individual optimal trajectories while satisfying the team constraints. We consider this problem for both fixed and switching communication topologies and in the process reveal connections between distributed optimization algorithms, graph structures, and the classical linear quadratic regulator problem.