Deep learning based and age of information aware resource management in HAPS-aided V2X systems

In this thesis, a novel resource management framework based on the multi-agent deep deterministic policy gradient (MADDPG) algorithm is proposed for high altitude platform station (HAPS)-enabled vehicle-to-everything (V2X) communication systems. The developed framework aims to optimize key performance indicators such as information freshness, energy efficiency, and communication capacity in highly dynamic vehicular environments. The proposed system integrates HAPS into the communication infrastructure as a complementary component to terrestrial base stations and roadside units, providing wider coverage and enhanced line-of-sight connectivity. Resource allocation problem is formulated as an optimization problem involving multiple agents, each representing a vehicle platoon. To handle the complexity of the environment, agents utilize the MADDPG algorithm to learn optimal communication strategies over time. The reward function is designed as a multi-objective structure that minimizes age-of-information (AoI), satisfies channel capacity constraints, and respects transmission power limits. This design enables agents to jointly learn policies that balance information freshness and resource efficiency. Simulation results show that the proposed HAPS-assisted MADDPG framework significantly outperforms both classical and fully decentralized approaches in terms of convergence speed, stability, and overall reward performance. In particular, HAPS integration allows for more stable and up-to-date data transmission across varying inter-vehicle distances, leading to significant improvements in AoI performance. Comparative evaluations also demonstrate that the proposed method offers strong adaptability to environmental dynamics and ensures superior communication quality under constrained network conditions.