Energy Efficiency-Driven Age of Information Minimization for UAV-Assisted Crowdsensing
Abstract
UAV crowdsensing offers intelligent sensing services with broad coverage, high reliability, and high-quality data for urban environments. However, energy constraints limit UAVs' data collection efficiency and timeliness, which is critical for real-time applications. In this paper, we aim to address these challenges by maximizing the data collection from points of interest (PoIs) while minimizing the average Age of Information (AoI) to ensure data freshness, all under energy constraints. To this end, we propose GDRL-UCS, a Graph Deep Reinforcement Learning-based method that leverages Graph Attention Networks (GAT) to capture spatial dependencies among multiple UAVs and optimize their trajectories and data collection policies. Additionally, GDRL-UCS incorporates Gated Recurrent Units (GRU) to exploit temporal patterns for improved decision-making. Simulation results demonstrate that GDRL-UCS reduces average AoI and enhances energy efficiency compared to state-of-the-art methods.