Quasi-optimization of all path planning in scooping all gravel for wheel loader by reinforcement learning

Mining operations at the mines need to be automated in order to remove the danger of operater. In this paper, we propose a sub-optimal path plan for a wheel loader to automate the scooping operation of all target sediments. By using deep reinforcement learning, the scooping point of the wheel loader is determined interactively according to the shape of the target sediments, then the total moving distance of the wheel loader becomes a quasi-minimum. By comparing the four route determination methods, we show that our automation method by deep reinforcement learning is useful.