Abstract: Addressing the challenges of collision detection and collision avoidance path planning during the collaborative operation of mining and anchoring equipment in the narrow and restricted spaces of underground coal mines, this paper proposes a method for collision detection and collision avoidance path planning for mining and anchoring equipment in tunneling lanes based on Deep Reinforcement Learning (DRL). LiDAR is utilized for real-time environmental reconstruction of the tunnel, and a path planning training model for mining and drilling equipment is established in a virtual environment. In the constructed three-dimensional virtual scene of the mining face, a hybrid hierarchical bounding box method is applied for virtual collision detection among mining and anchoring equipment, drilling and anchoring equipment, and the tunnel itself. Considering the motion characteristics of the mining and anchoring equipment, this paper introduces a Multi-Agent Experience Sharing mechanism on the basis of the Soft Actor-Critic (SAC) algorithm, proposing the MAES-SAC algorithm. By defining the state space and action space of the agent and designing a corresponding reward and punishment mechanism, the agent is trained. Simulation results indicate that, compared to the PPO algorithm and the SAC algorithm, the MAES-SAC algorithm has improved the average reward value by 8.21% and 7.43% respectively, increased the maximum reward value by 0.25% and 0.14% respectively, reduced the steps to reach the maximum reward value by 3.06% and 6.63% respectively, and decreased the standard deviation by 10.07% and 6.99% respectively. Finally, an experimental platform for collision avoidance path planning and collision perception system for mining and anchoring equipment is constructed. Through virtual-physical motion synchronization testing and collision avoidance trajectory planning experiments, the feasibility and accuracy of the collision avoidance path planning for mining and anchoring equipment are verified. This method provides a new approach for collision perception and collaborative collision avoidance path planning of mining equipment groups in underground coal mines, which is of significant importance for promoting the intelligent construction of mining faces in underground coal mines.