A novel hydraulic track motion control method of anti‐impact drilling robot based on MI‐DDPG

With the complex environment and poor road conditions in underground coal mines, accurate autonomous traveling of hydraulic tracks is a prerequisite for the Anti‐Impact Drilling Robot (AIDR) to achieve the goal of unmanned operation. In this paper, an AIDR path‐tracking control method that integrates reinforcement learning and manual instruction is proposed to reduce the tracking deviation during traveling and enhance the robustness of the control system. Based on AIDR kinematic modeling, a heading angle reward factor, a lateral deviation reward factor, a steering radius deviation reward factor, and a line speed reward factor are added to the deep deterministic policy gradient (DDPG) reward function. The path‐tracking problem is transformed into a multi‐objective optimization problem. For the problem of low efficiency of DDPG, manual instruction deep deterministic policy gradient (MI‐DDPG) is proposed by optimizing the empirical pool part using a manual instruction method. The results show that compared with the traditional path tracking control method, the method proposed in this paper has a maximum lateral deviation of 0.079 m during straight‐line travel and a maximum steering radius deviation of 0.062 m during circular steering of the AIDR under the cinder soft pavement (a typical harsh road condition in underground coal mines). This method can effectively improve the accuracy of AIDR path tracking, especially adapting to the complex and soft road environment of underground coal mines.