Adaptive optimal kinematic tracking control of nonholonomic mobile robot

An adaptive feedback velocity control laws is proposed for the kinematic system of nonholonomic mobile robot to guarantee that the posture error of tracking a reference trajectory asymptotically stable in optimally way. We transformed the tracking problem into a regulation problem by redefining new system states and inputs to make the pre-defined cost function finite, then proposed an online policy iteration (PI) algorithm by using a single neural networks (NNs) to solve of Hamilton-Jacobi-Bellman (HJB) equation approximately. This method learns online in real-time to approximate the cost function and then the adaptive optimal controller can be computed directly according to the cost function. Simulation results are provided to demonstrate the effectiveness of the proposed approach.