Mine shuttle cars have extensive applications in fields such as coal tunnel excavation, recovery of leftover coal, "three?down" coal filling and mining, and salt mine mining. With the continuous development of intelligent coal mines and unmanned vehicle technology, mine shuttle cars are gradually moving towards unmanned operation. However, the working environment of mining shuttle cars is complex, especially in the narrow coal mine roadways, where the shuttle cars are confronted with the challenges of trajectory planning and turning difficulties. In response to these problems, taking the unmanned shuttle car as the research object, through the analysis of its steering structure, simulation experiments were conducted to carry out the research on the trajectory planning of the mine shuttle car underground. Firstly, by analyzing the characteristics of the non-integrity constraints of the shuttle car, the kinematic equation was analyzed and derived, and the Hybrid A* algorithm and the TEB algorithm were determined as the basic algorithms for the trajectory planning of the mining shuttle car. Secondly, in view of the problems such as the large number of search nodes and the existence of paths with large curvature in the traditional Hybrid A* algorithm, the strategies of dynamically expanding the step size and dynamically using the RS curve are adopted. The calculation method of obstacle density is proposed and introduced into the steering penalty function. The reversing penalty is added to the cost function for the driving mode of the shuttle car. The generated paths were smoothed using the B-spline method. Global planning simulation experiments were conducted on the algorithms before and after improvement in the double-tunnel tunneling scenario. The results showed that the planning time of the improved D-Hybrid A* algorithm was reduced by 51.02%, the number of node expansions was reduced by 41.3%, and the number of turns was reduced by 36.36%. The efficiency was significantly improved. The generated path is smoother; Finally, aiming at the problems that the traditional TEB algorithm is prone to getting stuck in local optimum and generating trajectory oscillations, the obstacles were reconstructed, and an isolation layer was added outside the expansion layer to improve it. The simulation results of local trajectory planning of the algorithm before and after the improvement in the double-tunnel excavation environment were compared, indicating that the proposed S-TEB algorithm can achieve the functions of safe obstacle avoidance and stopping. Safe driving trajectories can be planned in various scenarios. The average efficiency of the improved S-TEB algorithm has increased by 22.89%, and the average safety performance has increased by 2.3 times compared to the original, which can meet the unmanned driving requirements of mining shuttle vehicles.