Hierarchical Control of Omnidirectional Mobile Robots Incorporating NMPC With Deep Neural Network

ABSTRACT

Omnidirectional mobile robots (OMRs) equipped with mecanum wheels possess superior mobility, enabling them to move in any direction and pivot in place within confined spaces. This unique advantage endows them with significant application potential across various fields. The development and application of these robots require safe motion control, which necessitates maintaining a safe distance from obstacles and achieving high‐precision trajectory tracking during motion. To achieve safe obstacle avoidance and high‐precision trajectory tracking control for OMRs while ensuring real‐time control capabilities, this article proposes a hierarchical control strategy. The high‐level module integrates nonlinear model predictive control (NMPC) and artificial potential fields (APF) to plan safe avoidance trajectories. The low‐level module uses a deep neural network (DNN) to approximate the NMPC trajectory tracking control method and employs quadratic programming (QP) to correct the outputs solved by the DNN, thus improving solving speed while maintaining control performance similar to that of NMPC. Simulation results demonstrate that the proposed scheme effectively achieves safe obstacle avoidance and trajectory tracking control.