Optimizing UAV deployment for maximizing coverage and data rate efficiency using multi-agent deep deterministic policy gradient and Bayesian optimization

Abstract

The rise of connected vehicles has highlighted the crucial need to cater diverse Quality of Service (QoS) demands in intricate vehicular networks. To address this, the burgeoning utilization of Unmanned Aerial Vehicles (UAVs) across various applications has garnered significant attention. UAVs, acting as aerial Base Stations (BSs), improve network coverage and performance in critical communication scenarios. However, challenges such as limited coverage range and unpredictable QoS hinder UAVs from continuously covering urban or rural areas. To tackle these challenges, we introduce a novel multi-agent deep deterministic policy gradient (MADDPG) approach incorporating Bayesian Optimization to optimize UAV trajectories in both rural and urban environments. Our principal aim is to maximize vehicle coverage while ensuring efficient QoS. Comparative evaluations against benchmark algorithms including Genetic Algorithm (GA), Particle Swarm Optimization (PSO), Sine Cosine Algorithm (SCA), and Greedy methods. In rural environments, proposed framework achieves a mean coverage rate of 85.75%, surpassing MADDPG by 4.49%, GA by 8.72%, PSO by 10.02%, SCA by 8.90%, and Greedy by 14.06%. In urban settings, proposed framework maintains superior performance, with a mean coverage rate of 83.78%, outperforming MADDPG by 6.31%, GA by 9.81%, PSO by 9.38%, SCA by 12.78%, and Greedy by 19.53%. Additionally, the system achieves a 95.6% convergence rate, optimizing MADDPG hyperparameters efficiently. The implications of the energy penalty in the proposed algorithm have given the outlook on the tradeoff, that the overall energy consumption can reduce up to 8.3%, it may also result in a decrease in coverage efficiency by around 5.5%.