Intelligent time-delay reduction of nonlinear model predictive control (NMPC) for wheeled mobile robots in the presence of obstacles

Abstract

The speed of solving and processing factors that are beneficial in reaching the desired target is one of the problematic aspects of controlling robots that has been neglected by the majority of researchers. Therefore, it is essential to look into the factors that influence calculation speed and goal achievement, and there should be some solutions to control robots in a lower time without sacrificing accuracy. The speed of processing and operations in wheeled mobile robots (WMRs), as well as the speed of a nonlinear model predictive control (NMPC), are examined in this article. The “Prediction horizon”, which is the most efficient element in increasing the calculations of the NMPC, is determined separately and intelligently at every step based on the magnitude of the error and the significance of the state variable by training a multilayer neural network, to decrease the time-delay in software mode. In addition, the processing speed in the hardware mode has increased due to the investigations conducted and the optimal selection of equipment effective in the speed of performing actuators, such as the use of the U2D2 interface instead of interface boards with their own processing, and the use of the pixy2 as a smart camera. The results have employed that the proposed intelligence method responds 40 to 50% faster compared to the conventional method of NMPC. Also, the path tracking error has been reduced by using the proposed algorithm due to the optimal gain extraction at each step. In addition, there is a comparison of solving speed in hardware mode between the proposed and usual methods. In this regard, about 33% increase in solving speed has been shown.