With the rapid advancement of UAV technology, its military applications have become increasingly widespread. Efficient path planning reduces equipment loss risks and operational combat costs, and as a core technology for autonomous UAV missions which has emerged as a current research hotspot. The limitations of traditional path planning algorithm in dense obstacle environments where fixed iteration counts and step sizes would lead to difficulties in generating feasible paths and excessive computation time. Pertinently this paper proposing a multi-strategy improved based on Deep Deterministic Policy Gradient, termed H-Bi-RRT*-DDPG. This method dynamically adjusts exploration step sizes using reinforcement learning’s experiential advantages, employs Halton sequences in multi-phase sampling to enhance sampling efficiency, and incorporates kinematic constraints with time-cost balancing to improve solution reliability. The simulation experiments comparing the proposed H-Bi-RRT*-DDPG algorithm with standard Bi-RRT* algorithm in typical dense-obstacle environments. The results demonstrate that the H-Bi-RRT*-DDPG algorithm achieves a 79.94% improvement in average solving speed and a 61.43% improvement in stability compared to the standard Bi-RRT* algorithm, confirming its superior computational efficiency and stability. These research findings provide a certain reference value for path planning in future practical applications.