Distributed predictive motion planning of automated guided vehicles: serial versus parallel schemes

Abstract

The article presents two distributed motion planning methods for dynamically coordinating multiple automated guide vehicles (AGVs) in an industrial setting, aiming to enhance their flexibility, robustness and scalability. A predictive motion model is utilized to describe the transport process mathematically as a dynamical system. Subsequently, alternating direction method of multipliers (ADMM)-based decomposition techniques, both serial and parallel, are developed to coordinate the AGVs and mitigate the computational burden. The efficacy of the distributed methods is verified through testing in industrial scenarios, demonstrating that they can significantly improve multi-AGV system transport productivity with minimal computational effort. Notably, the parallel scheme exhibits superior performance in coordinating the AGVs compared to the serial scheme.