Asymptotically stable reinforcement learning-based neural network controller using adaptive bounding technique

In this paper, a novel asymptotically stable reinforcement learning-based neural network controller using adaptive bounding technique for the tracking problem of a class continuous nonlinear system is proposed. An actor-critic structure is adopted for designing the controller, in which the critic network is tuned by itself and generates the reinforcement learning signal to tune actor network which generates the input signal to the system. The designed controller can achieve asymptotic convergence of the tracking error and performance measurement signal to zero, while ensuring boundedness of parameter estimation errors. No a prior knowledge of bounds of unknown quantities in designing the controller is assumed. Simulation results on a two-link robot manipulator show the satisfactory performance of the proposed control scheme.