建設ロボット

[VI-345] On exploring optimal spreading route for automated bulldozer using deep reinforcement learning

*田島 大輔¹、浜本 研一¹、三浦 悟¹、登尾 大地²、タンカラット ブット³、小熊 正¹、黒沼 出¹、中山 裕²、林 和彦²、前原 貴徳³ (1. 鹿島建設（株）、2. （株）小松製作所、3. 理化学研究所)

Keywords:automated construction machines, bulldozer, spreading work, deep reinforcement learning, simulator

Recently, we are promoting research and development of the next generation automated construction system based on the automation technology of construction machinery. In an environment that changes dynamically like a construction site, a function that has an autonomy that is suitable for self-evident automated heavy machinery is required. In this paper, we will report on the development contents of deep reinforcement learning about spreading work of the bulldozer for improving the autonomy of the automated bulldozer.