AoI and Energy Co-Optimization for UAV-IoT Systems: A Temporal-Critical DRL Solution

Real-time urban sensing in unmanned aerial vehicles (UAVs)-assisted mobile edge computing (MEC) requires timely data collection and energy efficiency. However, conventional Age of Information (AoI) metrics suffer from cumulative bias in multi-sortie missions, and existing hover-point (HP) deployment ignores variations in ground device traffic. This paper introduces a Sortie Relative AoI (SR-AoI) to eliminate inter-sortie bias and a traffic-weighted K-means strategy for optimized clustering. A Temporal and Criticality-Aware Double Deep Q-Network (TAC-DDQN) based on deep reinforcement learning (DRL) is proposed for joint HP-based path planning and task scheduling, featuring a temporally adaptive discount factor and criticality-driven prioritized experience replay to balance information freshness and energy consumption. Simulation results demonstrate that the proposed approach achieves superior performance in both AoI and energy efficiency compared to baseline algorithms.