Approximate Policy Iteration for Cooperative Control of Multiagent Systems Under Limited Sensing/Communication

Jing Wang, Tianyu Yang, Gennady Staskevich, and Brian Abbe


Multiagent policy iteration, Cooperative control, Multiagent Systems, Neural Network


In this paper, we propose an approximate policy iteration method for cooperative control of multiagent systems under the limited sensing/communication topology. By considering a class of nonlinear multiagent systems, the cooperative control problem is formulated as making all systems achieve consensus while minimizing the individual sensing/communication topology dependent cost functions. To solve the induced multiagent Hamilton-Jacobi-Bellman (HJB) equations, a multiagent policy iteration algorithm is proposed with convergence proof. Neural network parameterization is further employed to approximate value function to deal with unknown system dynamics. Through seeking the least-squares solution based on the measured online sensing/communication data, the approximate multiagent policy iteration algorithm is obtained to solve the posed optimal cooperative control problem for multi agents. Simulation results illustrate the effectiveness of the proposed optimal cooperative control.

Important Links:

Go Back