Deep Q-Networkとグラフ探索を組み合わせた複数台移動ロボットの経路計画法
福島 昂之介, 西 竜志, 劉 子昂, 藤原 始史
pp. 207-215
DOI:
10.5687/iscie.37.207抄録
In recent years, automated multiple mobile robots are introduced for transporting luggage and inspecting final products in factories to reduce the burden of human labor shortage. There is a need for mobile robots to develop autonomous systems that make flexible decisions like human operators. We propose a route planning method that combines deep reinforcement learning and graph search methods. In the proposed method, the routing is firstly determined by a graph search algorithm, and deep Q-network (DQN). A deep reinforcement learning method is used to avoid collisions. The proposed method is applied to the multiple drones route planning problem. As a result, a near-optimal routing is obtained that can reach to the destination while avoiding collisions between drones. These results suggest that the route planning problem in a three dimensional environment is successfully solved by using DQN that can process multidimensional states. We generate a learning model for collision avoidance using DQN for both whole observation and partial observation ranges to verify the usefulness of path planning with partial observations through computational experiments.