Efficient Path planning for Automated Guided Vehicles in Intelligent Logistics System using Double Q-learning

In logistics systems, Multiple Automated Guided Vehicles (AGVs) have been extensively used for transportation. While determining the path by AGVs, Conflict-free route planning is a major research challenge. This study builds a novel efficient path-finding model using a Double QLearning reinforcement algorithm (DQL) to improve the efficiency of multi-robot path planning. Initially, training has been performed using a double Q-learning algorithm to learn the solutions to route-planning problems quickly. This algorithm determines the optimal route for AGV to reach the destination based on rewards without collision and ensures minimal transfer time. AGV selects the action based on Q-table values. Results show that this Double Q-learning path-finding algorithm executes faster than existing algorithms such as the $A^{*}$ algorithm, Conflict Based Search and Autonomous Distributed Route Planning Method (ADRPM) and reduces total picking distance by 5% than other methods. This DQL path finding algorithm has been executed with less computation time than ADRPM by nearly 57%.