Abstract
This paper presents a multi-source data-driven hybrid path planning framework that integrates global A* search with local Deep Q-Network (DQN) optimization to address complex terrestrial routing challenges. By fusing ASTER GDEM terrain data with OpenStreetMap (OSM) road networks, we construct a standardized geospatial database encompassing elevation, traffic, and road attributes. A dynamic-heuristic A* algorithm is proposed, incorporating traffic signals and congestion penalties, and is enhanced by a DQN-based local decision module to improve adaptability to dynamic environments. Experimental results on a realistic urban dataset demonstrate that the proposed method achieves superior performance in risk avoidance, travel time reduction, and dynamic obstacle handling compared to traditional models. This study contributes a unified architecture that enhances planning robustness and lays the foundation for real-time applications in emergency response and smart logistics.