MADDPG Based Path Planning for Multi-UAV Information Collection in IoT Networks

Multi-UAV-assisted information collection is an effective solution for large-scale Internet of Things (IoT) networks, but it is challenged by limited onboard energy, charging requirements, and the need for coordinated path planning among multiple UAVs. This paper investigates a multi-UAV information collection problem in which UAVs cooperatively collect data from distributed IoT devices under energy constraints, charging operations, and collision avoidance constraints, with the objective of minimizing the total mission completion time. The considered problem is formulated as a multi-agent Markov decision process. To address the coupled and high-dimensional decision-making nature of the system, a multi-agent deep deterministic policy gradient (MADDPG) framework with centralized training and decentralized execution is developed for cooperative UAV path planning. An energy-aware reward function is designed to jointly account for information collection efficiency, energy consumption, and charging behavior. Simulation results demonstrate that the proposed approach achieves shorter mission completion time and better scalability compared with DDPG-based and MAPPO-based methods.