Nonlinear Model Predictive Motion Control of Differential Wheeled Robots

Real-time motion control of a nonholonomic mobile robot in a dynamic environment, especially in the case of multiobjective control, is a challenging problem. Model predictive control (MPC) as an optimization based control algorithm has the ability to deal with complex systems, like multiple-input and multiple-output (MIMO) system, in a dynamic environment. However, due to the complexity of optimization algorithms, the implementation of MPC in real-time applications, especially for the systems with fast transient behaviors is very challenging. With the advent of processors with the ability of parallel computing like FPGAs and GPUs, the application of MPC has become reachable. In this study, the algorithm of optimization problem as the core part of the MPC for motion control of a two-wheel differential robot was developed. Considering the final objective of coding the optimization algorithm on FPGA, the sequential quadratic programming (SQP) method was selected as the optimization algorithm. The specific algorithm equations and matrices were derived based on a simplified nonlinear model. The algorithms were then be coded in MATLAB and used to control a two-wheel robot in the simulation. This paper present the MPC design process and simulation results for the cases of path tracking and point tracking.