论文部分内容阅读
针对一类未知的连续非线性系统,提出一个基于单网络近似动态规划(ADP)的近似最优控制方案.该方案通过设计一个新型的递归神经网络(RNN)辨识器放松了系统模型需已知或部分已知的要求,并利用一个神经网络(NN)近似系统的性能指标函数消除了常规ADP方法中的控制网络.通过Lyapunov理论分析严格证明了闭环系统内所有信号一致最终有界,并且所获得的性能指标函数和控制输入分别收敛到最优性能指标函数和最优控制输入的小邻域内.仿真结果验证了所提出控制方案的有效性.
Aiming at a class of unknown continuous nonlinear systems, an approximate optimal control scheme based on single network approximation dynamic programming (ADP) is proposed. The scheme relaxes the need to know the system model by designing a new RNN recognizer Or partially known requirements and eliminates the control network in the conventional ADP method by using the performance index function of a neural network (NN) approximation system. It is strictly proved by Lyapunov theory that all the signals in the closed-loop system are uniformly and ultimately bounded, The obtained performance index function and control input converge to the optimal performance index function and the optimal control input in a small neighborhood, respectively. The simulation results verify the effectiveness of the proposed control scheme.