A Multi-Objective Deep Reinforcement Learning Method for Path Planning in Shovel Loading Scenario

The utilization of unmanned mining trucks is gradually expanding. However, planning a path from the starting position to the destination for unmanned mining trucks consumes a significant amount of time due to the irregular environment and the vast search space. On the other hand, deep reinforcement learning holds the potential to expedite the speed of path planning because prior knowledge can be utilized more efficiently. To solve the problem that the network is difficult to converge and easy to get trapped in local optima in shovel loading scenarios and path planning is time-consuming, this paper proposes a method for path planning based on deep reinforcement learning with the residual network and design a reward that considers multiple objectives. This method possesses the advantage that the consumption of time in path planning is reduced and the learning process is easy to converge. Finally, the proposed method is verified in a emulated model using real shovel loading scenario and the result suggests that our method overperforms Hybrid A* in terms of time consumption when the length of paths is similar.