Fast-convergent model predictive control of omni-directional mobile robots

摘要

Omni-directional wheeled mobile robots (OWMRs) present high manoeuvrability and can move in all directions of the plane. However, for active-caster-type OWMRs (ACT-OWMRs), satisfying physical constraints of multiple motors is challenging. To guarantee motor constraints and improve the tracking control performance, a new model predictive control (MPC)-based controller termed as 'MPC-FCR' is designed, which embeds a real-time fast-convergent trajectory planner to generate high efficient gentle reference trajectory. Such a framework has two main advantages. First, the gentle trajectory generation module is formulated as a constrained optimisation problem, which can achieve fast convergence performance while satisfying the robot constraints. Second, both motor and robot constraints are considered in MPC-FCR control law. The proposed MPC-FCR is implemented to achieve real-time performance within a control period of 20 milliseconds. Experimental results show the superior performance of MPC-FCR in terms of convergence speed and the capability to handle physical constraints, compared with the classical controller.