Development of artificial intelligence-assisted collision avoidance algorithms for automated guided vehicles

Automated guided vehicles are transportation systems widely used in industrial environments such as factories, warehouses, and distribution centers, providing reduced labor costs and increased efficiency in production processes. Ensuring the control and coordination of vehicles is crucial for safe and efficient transportation in multi-vehicle systems. In this thesis, modeling and control methods are proposed to prevent collisions in automated guided vehicle systems operating in environments with shared work zones and overlapping routes. The modeling and control methods developed for the coordination of multi-vehicle systems were tested through simulation applications. In the proposed methods, finite state machines were used to model the movements of AGVs in their working environments, while Q-learning, one of the most common reinforcement learning algorithms, was employed to prevent collisions. During the vehicle modeling phase, the states of other vehicles in the collision zones were considered, and simplicity and clarity in the created models were aimed for. In line with these objectives, decentralized modeling approaches were used to reduce the complexity in systems involving multiple vehicles. The finite state automata models, specifically designed for each vehicle, were defined as the environment model in the Q-learning algorithms, which would control the vehicles. Q-tables, containing the actions each vehicle should take in their respective states, were obtained through the Q-learning algorithms designed according to these models. In various systems involving a large number of vehicles and numerous collision zones, the Q-tables for controlling the vehicles were used in simulation applications, and the proposed methods were tested and validated. The results of the simulation applications demonstrate that the proposed methods can potentially prevent collisions and significantly enhance overall efficiency.