A central event-triggered nonlinear MPC approach to reduce the computational time of WMR

One of the limitations of applying Nonlinear Model Predictive Control (NMPC) in robotic systems is the high computational burden associated with the continuous solution of the Optimal Control Problem (OCP). In this paper, an intelligent central event-triggered method based on the variation of the gradient between the optimal state error and the actual state is proposed to achieve intermittent solving and reduce the frequency of OCP computations, consequently decreasing the computational time of NMPC. Unlike conventional event-triggered NMPC (ET-NMPC), which can degrade solution accuracy when combined with warm-starting, the proposed approach employs a Multilayer Perceptron Neural Network (MLP-NN) to predict the OCP inputs. This strategy reduces the number of iterations required per solution, enhances convergence, and enables the NMPC to track the trajectory more accurately, mitigating the accuracy loss typically associated with event-triggered methods. Simulation and experimental validation were performed on a wheeled mobile robot (WMR) platform. The results indicate that the proposed intelligent event-triggering mechanism reduces the computational time by 64.7 % compared to traditional NMPC, while improving the event-triggered tracking error by 18 %.