Real-time optimal trajectory planning of mobile robot in presence of obstacle through Generalized Regression Neural Network

In this paper, a new approach for optimal real-time path planning of wheeled mobile robots based on Generalized Regression Neural Network (GRNN) and optimal control is presented. Optimal control is used to find an accurate mathematical solution for mobile robot with considering kinodynamic constraints. However, the drawback of this method is to be more time-consuming than standard real time systems. In this study, through the optimal control procedure, the best path is obtained, in terms of distance, input torque, and velocity. Then, the network is trained by these data, and optimal control is replaced by GRNN. It generates the path with acceptable reduction in cost function (at time of 0.3 s), which is suitable for real-time application of mobile robot. The simulation results demonstrate the ability of trained neural network in generating paths with less computational time cost in comparison with other methods which are merely generated by the optimal control procedure.