Look-ahead control of a car-like mobile robot via reinforcement learning

In this paper, the performance improvement problem of a reference trajectory tracking for a car-like mobile robot with nonholonomic constraints in the presence of external disturbances, nonlinearities and uncertain parameters is investigated. For this purpose, at first the dynamic and kinematic equations of the car-like mobile robot are expressed and then the look-ahead control method in two dimensional is used for the tracking of the car-like mobile robot. The purposed controller will be designed by using the dynamic surface control method with the reinforcement learning based on the actor-critic neural network and an adaptive robust controller is proposed to compensate the effects of external disturbances. Moreover, the prescribed performance control method will be utilized to improve the transient state and steady state. An actor neural network is used to estimate unknown nonlinearities and uncertainties and a critic neural network is employed to evaluate system performance. In the sequel, the direct Lyapunov method will be used to prove the closed-loop control system stability and uniform ultimate boundedness of tracking errors. Finally, the effectiveness and efficiency of the proposed control scheme are confirmed by using MATLAB software.