Inverse Reinforcement Learning with DQN for Interference-Aware UAV Path Planning

This paper investigates the multi-objective optimization of path planning and power allocation in Unmanned Arieal Vehicle (UAV)-assisted networks, aiming to minimize the bidirectional interference between UAV and ground user equipment (UE) under energy constraints. A Dueling and Double Deep Q-Network (D3QN)-based inverse reinforcement learning (IRL) multi-objective optimization model is developed in this paper. The proposed model is capable of adaptively adjusting the reward function in response to dynamic environmental changes. The experimental results show good performance in both immediate feedback from single experiments and long-term average rewards across multiple experiments. This indicates that the proposed model has strong adaptability in different scenarios.