Dynamic Event‐Based Optimal Control for Multiplayer Nonzero‐Sum Games With Asymmetric Input and State Constraints

ABSTRACT This work investigates the optimal control of multiplayer nonzero‐sum games (NZSGs) subject to asymmetric input and state constraints by using a dynamic event‐based adaptive dynamic programming (ADP) method. Considering the multiplayer NZSGs captures a wider range of practical scenarios. A control barrier function is used to deal with constrained states to facilitate the use of optimal control techniques. By defining a nonquadratic cost function, the Nash equilibrium solutions are obtained by searching a group of optimal control policies. To simplify the training process and reduce the computing burden, a single neural network is constructed to implement the ADP method and the experience replay technique is introduced to ensure no need for persistence of excitation. In conclusion, our proposed ADP method generalizes well to various types of NZSGs. A two‐player and a three‐player NZSG are simulated to show the effectiveness of this method.