Deep Reinforcement Learning for Semantic-Aware AoI Minimization in UAV-Assisted Communication Networks

Abstract

Unmanned Aerial Vehicles (UAVs) are envisioned as key enablers for time-sensitive data collection in 6G cognitive networks. Semantic communication offers a promising solution to overcome bandwidth scarcity by transmitting only essential information. However, the heavy computational burden of semantic extraction is often ignored, which can lead to significant processing delays and stale information. In this letter, we investigate a semantic-aware Age of Information (AoI) minimization problem in UAV-assisted networks, explicitly accounting for the trade-off between transmission latency and computation latency. We formulate a joint optimization problem for UAV trajectory planning and semantic compression level selection. To solve this problem with a hybrid action space (continuous trajectory and discrete compression levels), we propose a Hybrid-Action Proximal Policy Optimization (HA-PPO) deep reinforcement learning algorithm. Simulation results demonstrate that the proposed scheme significantly reduces the Semantic-aware AoI compared to traditional bit-based and greedy semantic baselines. Crucially, our results reveal that intelligent switching between semantic extraction and raw transmission based on the UAV’s computing capacity is essential for maintaining information freshness.