Multiagent systems arise in several domains of engineering. Examples include multi-stage supply chains, arrays of sensor networks for aggregate imagery, autonomous highways, and formations of unmanned aerial vehicles. In these contexts, the individual subsystems are governed by dynamics and typically constraints, and the control objective is often achieved by cooperation. To be viable, the control approach for these types of systems should be distributed, rather than centralized, particularly when the scale of the overall system is large. Receding Horizon Control (RHC) is well suited to a wide range of multiagent problems, in that RHC can admit very general objectives, dynamic models and constraints. In this talk, a distributing implementation of RHC is presented, using the venue of multiple autonomous vehicles. The resulting control law is provably asymptotically stabilizing and scalable (i.e., agents compute locally for themselves, communicating only with those agents to whom they are coupled). While the algorithm is presented for multiple vehicles, a venue in which the agent dynamics are decoupled, the same algorithm is proven stabilizing when the agent dynamics are coupled, as in the case of supply chain systems. Simulations comparing the distributed implementation to a centralized implementation will be presented, for an example multiple autonomous vehicle mission, and in a multi-stage supply chain management problem.
Back to Control Seminars Page