Fairness-Based Energy-Efficient 3-D Path Planning of a Portable Access Point: A Deep Reinforcement Learning Approach

In this work, we optimize the 3D trajectory of an unmanned aerial vehicle (UAV)-based portable access point (PAP) that provides wireless services to a set ground nodes (GNs). Moreover, as per Peukert effect, consider pragmatic non-linear battery discharge for UAV’s battery. Thus, formulate problem in novel manner represents maximization fairness-based energy efficiency metric and is named fair (FEE). The FEE defines system lays importance on both per-user service fairness PAP’s efficiency. formulated takes form non-convex with non-tractable constraints. To obtain solution represent Markov Decision Process (MDP) continuous state action spaces. Considering complexity space, use twin delayed deep deterministic policy gradient (TD3) actor-critic reinforcement learning (DRL) framework learn maximizes system. We perform two types RL training exhibit effectiveness our approach: first (offline) approach keeps positions GNs same throughout phase; second generalizes learned any arrangement by changing after each episode. Numerical evaluations show neglecting effect overestimates air-time PAP can be addressed optimally selecting flying speed. user fairness, efficiency, hence value improved efficiently moving above GNs. As such, notice massive improvements over baseline scenarios up 88.31%, 272.34%, 318.13% suburban, urban, dense urban environments, respectively