An online single-network adaptive algorithm for continuous-time nonlinear optimal control

In this paper, we propose an online adaptive neural-algorithm to solve the CT nonlinear optimal control problems. Compared to the existing methods, which adopt the architecture with two neural networks (NNs) for actor-critic implementations, only one NN for critic is used to implement the algorithm, simplifying the structure of the computation model. Moreover, we also provide a generalized learning rule for updating the NN weights, which covers the existing critic update rules as special cases. The theoretical and numerical results are given under the required persistent excitation condition to verify and analyze stability and performance of the proposed method.