Path tracking via speed adjustment for car-like robots based on a multilayer MPC framework

Abstract

Car-like robots are front wheel steering robots with a structure similar to that of unmanned vehicles and are widely used in manufacturing, warehousing, and other industries because of their advantages, such as simple structure and high load-bearing capacity. The path tracking control of car-like robots, characterized by a small range of system constraints and a low degree of component standardization, is garnering increasing widespread attention. Several studies have been conducted on this topic; however, these studies have not considered the correlation between active speed adjustment and path tracking. Existing control systems set the longitudinal speed to a low value to ensure high path-tracking control accuracy. Meanwhile, among active speed control strategies for other mobile equipment, multilayer model predictive control (MMPC), based on model predictive control (MPC), has the advantage of high compatibility with error sources. However, existing MMPC systems adopt linear MPC (LMPC) as the bottom path tracking control algorithm. Typical LMPC design methods are unable to take into account the reference path information in front of car-like robots and are not accurate enough at high longitudinal speeds; therefore, the longitudinal speeds of existing MMPC systems are still low. To address these problems, an MMPC-based framework for active speed adjustment and path-tracking control is introduced. The framework combines a feedforward MPC (FMPC) bottom path-tracking control algorithm with high accuracy and real-time performance and a nonlinear MPC (NMPC) top-speed adjustment algorithm with high long-term predictive accuracy. This new system aims to enhance car-like robots’ path tracking capabilities while actively adjusting their speed. The effectiveness of the proposed MMPC system is validated through joint simulations using MATLAB and CarSim. Results show that the proposed system achieves high-accuracy path tracking at high average traveling speeds, recording a maximum displacement error of 0.1838 m and a heading error of 0.1350 rad at an average traveling speed of 4.2859 m·s−1. Compared with constant-speed FMPC, NMPC systems, and the existing MMPC system, the proposed system demonstrated higher accuracy at higher longitudinal speeds. Under the same working conditions, the error of the FMPC system is dispersed, and the proposed MMPC system can reduce the maximum displacement error by 46.29% and 62.22% compared with the existing NMPC and MMPC systems. When higher accuracy of path tracking can be guaranteed, the average traveling speed of the proposed MMPC system is higher, and this system can increase the average traveling speed by 43.06% and 317.48% compared with existing FMPC and MMPC systems with smaller errors, approaching the performance index of the NMPC system.

Keywords

• car-like robots
• path tracking
• active speed adjustment
• predictive control
• multilayer control