An integrated framework for Multi-AMR based CL-CBS and MPC-APF in warehousing scenario

Abstract

Aiming at the problem of Multi-Autonomous mobile robots (Multi-AMR) performing autonomous handling tasks in warehouse scenario, this paper proposes a framework that combines the Car like-Conflict based search algorithm (CL-CBS) and the Model predictive control-Artificial potential field algorithm (MPC-APF) is proposed for local trajectory replanning and tracking control. First, the CL-CBS is employed at the global trajectory planning layer; the algorithm uses a binary tree-based conflict search algorithm at the top-level and a spatiotemporal Hybrid-A* algorithm at the lower-level, which allows Multi-AMR to plan collision-free trajectories in compliance with the Ackermann kinematic characteristics. Second, at the trajectory replanning layer, the quintic polynomial equation is employed to fit segments to the discrete points with temporal information to enhance the smoothness and feasibility of the trajectory. Then, an function is proposed which incorporates the features of the APF in the form of an obstacle avoidance function into the optimization solution of the MPC. Finally, at the trajectory tracking control layer, a leapfrog speed planning is proposed, and a dynamics model is used to perform tracking control on the trajectories input from the replanning layer. Moreover, a structured warehousing map is built on virtual environments to validate the framework, and the results verify its safety and feasibility.