Adaptive MPC-Based Multi-Terrain Trajectory Tracking Framework for Mobile Spherical Robots

Abstract

Multi-terrain trajectory tracking control is a prerequisite for robots to execute tasks in the wild and unknown environments. However, due to the uncertainties in both kinematics and dynamics, the current trajectory tracking framework for mobile robots, such as spherical robots cannot function effectively on multiple terrains, especially uneven and unknown ones. Thus, an efficient online adaptive algorithm that can rapidly obtain uncertainties online and assist the robot in adapting to multiple terrains is required. In this article, we propose an adaptive model predictive control-based instruction planner (VAN-MPC) and a multi-terrain trajectory tracking framework (VANMHH). The VAN-MPC is composed of three components as follows. 1) An adaptive dual-RBF neural network that online estimates all uncertainties based on the adaptive law and composite error. 2) A model predictive controller with a control Lyapunov function constraint. 3) The variable step-size algorithm for the adaptive controller to accelerate convergence and enhance stability. Without modifying the bottom velocity and direction controllers, we finally develop the multi-terrain trajectory tracking framework VANMHH just by employing the new instruction planner VAN-MPC. The practical experiments on multiple terrains demonstrate its effectiveness and robustness.