A DECENTRALISED COLLISION AVOIDANCE METHOD BASED ON Q-LEARNING FOR MULTI-AGV SYSTEMS

Abstract

Automated guided vehicles are widely used transport systems in factories, warehouses, and distribution centres. The control and coordination of vehicles is of great importance for safe and efficient transport in multi-vehicle systems. In this study, a collision avoidance strategy is proposed for automated guided vehicle systems operating in environments with shared work zones and conflicting routes. In the proposed method, finite state machines are used to model the motion of vehicles in the environment. Q-learning, one of the most common algorithms of reinforcement learning, is used for collision avoidance. In the presented strategy, a decentralised control approach is utilized to reduce the computational complexity. The proposed method is validated through simulations involving multi-vehicle system applications with several collision zones. Simulation results demonstrate that the proposed method can avoid potential collisions and significantly improve overall efficiency.