A Charging‐Aware Hybrid Deep Q‐Learning Framework for Energy‐Efficient Path Planning of a Single UAV

ABSTRACT This paper proposes a charging‐aware hybrid deep Q‐learning (HDQL) framework for intelligent and energy‐efficient path planning of unmanned aerial vehicles (UAVs) in grid‐based industrial environments. The proposed HDQL model integrates three key components: A*‐guided heuristic reward shaping, battery‐aware decision‐making, and adaptive charging strategy, to ensure reliable navigation under strict energy constraints. A dynamic state‐dependent reward shaping mechanism is designed to consider residual battery level, goal distance, and charger proximity, enabling the UAV to balance path optimality and energy consumption. Extensive experiments were conducted in grid environments of sizes , , and . The experimental results demonstrate that the proposed HDQL framework outperforms baseline algorithms (DQL, HR‐DQL, and HR‐DQL+charging) in terms of success rate, average reward, and path efficiency. The proposed method achieves success rates of 89.44%, 58.70%, and 29.37% for , , and grid environments, respectively. The results also show that HDQL provides faster convergence, shorter navigation paths, and more stable learning performance compared to baseline methods. Additional experiments including sensitivity analysis of the reward shaping parameter, computational time analysis, ablation study, and multi‐charger scenario evaluation further demonstrate the robustness, scalability, and practical applicability of the proposed HDQL framework. Overall, the proposed HDQL framework provides a reliable and energy‐aware solution for UAV path planning in smart industrial environments with energy constraints.