Improving scheduling in multi-AGV systems by task prediction

Automated guided vehicles (AGVs) are driverless robotic vehicles that pick up and deliver materials. Finding ways to improve efficiency while preventing deadlocks is a core issue in designing AGV systems. In this paper, we propose an approach to improve the efficiency of traditional deadlock-free scheduling algorithms. Typically, AGVs have to travel to designated starting locations from their parking locations to execute tasks, the time required for which is referred to as preparation time. The proposed approach aims at reducing the preparation time by predicting the starting locations for future tasks and then making decisions on whether to send an AGV to the predicted starting location of the upcoming task, thus reducing the time spent waiting for an AGV to arrive at the starting location after the upcoming task is created. Cases in which wrong predictions have been made are also addressed in the proposed method. Simulation results show that the proposed method significantly improves efficiency, up to 20-30% as compared with traditional methods.