A digital twin-based decision support approach for AGV scheduling

Abstract

The dynamic and complex environment affects the operational efficiency of automated guided vehicles (AGVs) at automated container terminals, in which artificial intelligent approaches are applied to address optimization problems in dynamic systems. This paper proposes a digital twin-based decision support approach to improve the efficiency of AGV scheduling service. Accordingly, the factors that affect AGV scheduling performance, including conflicts, failures, and battery constraints, are discussed first. Then, the framework and main steps of the proposed approach are described. The physical operation space is mapped into the virtual space, and both spaces keep synchronized to support the verification of solutions. A mathematical programming model and Q-learning algorithm are used to generate the AGV scheduling plan considering battery charging. Numerical experiments are conducted to demonstrate the effectiveness of the proposed approach. Comparisons of the digital twin-based approach, genetic algorithm (GA), and particle swarm optimization (PSO) are also made with different scale experiments. It appears that the proposed digital twin-based approach is superior to GA and PSO when solving small- and large-scale cases. A sensitivity analysis is performed concerning the battery utilization rate, task, and AGV number. Experimental results show that an optimal configuration of AGV and task can improve the battery utilization rate and reduce the completion time.