Vision-Based Model Predictive Control of Wheeled Mobile Robot

In this article, a novel control of wheeled mobile robot based on machine vision is considered. One of the common methods for controlling such systems is the use of Model Predictive Control (MPC) algorithms. In these systems, the response speed of the control algorithm and the optimality of these are two basic factors for achieving the optimal performance. Also, the impossibility of achieving precise values of the robot parameters and their variation during the operation of the robot is an important challenge in the implementation of the controller, therefore, this paper focuses on real-time and robust MPC, so that it can ensure the system against uncertainties and environmental disturbances in addition to the optimal and real-time response. Hence, the optimization based on projection recurrent neural network (PRNN) has been used as an optimizer to reduce the calculation time cost. The combination of PRNN optimization with MPC leads to new formulation and constraints that are considered to be the article innovations. Finally, in order to verify the validity of the proposed algorithm, the robot passes through the corridor with the presence of obstacles, which is simulated in the V-REP software. The results show that the optimum control input speed has been increased in comparison with similar methods, and the optimal path selection by the fuzzy system in the presence of obstacles has been well suited.