Energy-Efficient Multi-Agent UAV Path Planning for Green IoT Systems

Efficient UAV navigation in large-scale six-generation (6G)-enabled green internet of things (GIoT) systems presents significant challenges due to stringent energy constraints, dynamic network conditions, and the need for robust connectivity and high coverage. Existing methods often neglect critical factors such as real-time adaptability to network variations, scalable multi-agent coordination, and integrated communication-energy optimization. To address these gaps, this paper proposes a novel multi-agent UAV path-planning framework based on the proximal policy optimization (PPO) algorithm, explicitly designed for connectivity-aware navigation. The approach integrates simultaneous wireless information and power transfer (SWIPT)-based energy harvesting, dynamic obstacle avoidance, and communication-driven reward shaping to enable sustainable, efficient UAV operations. Extensive simulations in a custom Gymnasium environment, modeled on precision agriculture, validate the framework. Experimental results with three UAV agents demonstrate 100% mission success, up to 0.0096 J/bit energy efficiency, average communication latency below 11 ms, and improved coverage with zero collisions for two agents. Compared to classical planners (A*, Dijkstra) and baseline PPO methods, the proposed model achieves a 15% higher cumulative reward and superior energy-latency trade-offs. Fully decentralized decision-making further enables scalable, practical deployment. This framework enables real-time UAV coordination by integrating 6G communication, energy, and dynamic environment adaptation for Green IoT.