Discrete-time decentralized control using the risk-sensitive performance criterion in the large population regime: A mean field approach

摘要: This paper considers a discrete-time decentralized control problem using the risk-sensitive cost function when there is large number of agents. We solve this via mean field theory. first obtain an individual robust controller that local state information and bias term related to term. then construct auxiliary system characterizes best approximation in mean-square sense agents, say N, goes infinity. prove set controllers e-Nash equilibrium, where e can be made arbitrarily close zero N → ∞. Finally, we show view relationship with risk-sensitive, H∞, LQG control, equilibrium features robustness, converges game risk-sensitivity parameter