Information Freshness-Aware Task Offloading in Air-Ground Integrated Edge Computing Systems

This paper investigates an air-ground integrated multi-access edge computing system, which is deployed by infrastructure provider (InP). Under a business agreement with the InP, third-party service provides services to subscribed mobile users (MUs). MUs compete for shared spectrum and resources over time achieve their distinctive goals. From perspective of MU, we deliberately define age update capture staleness information from refreshing computation outcomes. Given system dynamics, model interactions among as stochastic game. In Nash equilibrium without cooperation, each MU behaves in accordance local states conjectures. We can hence transform game into single-agent Markov decision process. As another major contribution, develop online deep reinforcement learning (RL) scheme that adopts two separate double Q-networks approximate Q-factor post-decision Q-factor, respectively. The RL allows optimize behaviours unknown dynamic statistics. Numerical experiments show our proposed outperforms baselines terms average utility under various conditions.