Optimizing UAV Path Planning and Energy Management Using Reinforcement-Learning and Digital Twins Simulation
Résumé
In many industrial applications, optimizing the path planning and energy management of unmanned aerial vehicles (UAVs) is crucial for efficient and sustainable operations. This work presents a framework that integrates Reinforcement Learning (RL) and Digital Twin (DT) technologies to optimize UAV trajectory planning while managing energy resources. The framework aims to minimize energy consumption and maximize operational efficiency by dynamically adjusting the UAV’s path and recharging schedule. The UAVs operate in an industrial environment where they must visit multiple checkpoints and avoid obstacles while adhering to battery constraints. A deep reinforcement learning approach, specifically Q-learning, is used to optimize the UAVs’ decision-making processes in real-time. The proposed solution significantly improves energy efficiency and adaptability when compared to traditional path-planning algorithms such as A* and Dijkstra. Experimental results demonstrate the effectiveness of the proposed framework in achieving 100% checkpoint coverage with minimal battery usage, highlighting its potential for large-scale, real-time industrial applications.