Transforming wireless network efficiency with UAV path planning using deep reinforcement learning

Abstract

Unmanned aerial vehicles (UAVs) are emerging as a key enabler for enhancing wireless network performance in dynamic and challenging environments. This paper systematically studied a deep reinforcement learning (DRL) framework for optimizing the trajectories of UAVs operating as mobile base station (MoBS). The problem is formulated as a Markov decision process (MDP) and solved using two DRL algorithms: Deep Q-network (DQN) and Proximal Policy Optimization (PPO). Seven reward functions are designed to maximize average channel capacity and fairness among ground terminals (GTs). The work further presents an in-depth analysis of reward shaping, diverse network configurations, and hyperparameter tuning. Simulation results show that the proposed approach outperforms static and non-adaptive baselines, achieving near-optimal performance under uncertain GT movement patterns. Scalability evaluations demonstrate strong generalization to larger network areas, providing valuable insights for DRL-based UAV path planning in next-generation wireless networks.