Initial Excitation-Based Optimal Control for Continuous-Time Linear Nonzero-Sum Games
针对连续时间线性非零和博弈，提出基于初始激励的策略迭代和值迭代算法，在可在线验证的初始激励条件下求解纳什均衡，无需持续激励或数据存储。
In this article, the initial excitation-based optimal control methods are presented for continuous-time linear nonzero-sum games. The traditional reinforcement learning-based optimal control methods for continuous-time linear nonzero-sum games require the persistent excitation condition or data storage to guarantee the convergence of the algorithms. To relax the above conditions, the initial excitation-based policy iteration and value iteration algorithms are presented to obtain the Nash equilibrium solution under an online-verifiable initial excitation condition. The properties of the initial excitation-based policy iteration and value iteration algorithms are analyzed. Simulation examples are provided to show the efficiency of the presented methods.