Soft Actor–Critic optimization for efficient NOMA uplink in intelligent vehicular networks

Abstract

This paper addresses the need for efficient data transmission in congested intelligent vehicular networks using Non-Orthogonal Multiple Access (NOMA). We present a novel uplink communication model that maximizes NOMA’s spectral efficiency, enabling multiple vehicles to share frequency channels effectively. Our dynamic optimization framework adjusts channel assignments, power levels, and bandwidth allocations, significantly enhancing spectral and energy efficiencies critical for battery-limited intelligent vehicles. To address complex optimization challenges, we develop a Heterogeneous-action Multi-agent Soft Actor–Critic (HMSAC) algorithm with an enhanced attention mechanism, improving learning efficiency and convergence speed. Rigorous simulations show that our approach notably boosts energy efficiency, reliability, and scalability compared to baseline models. Finally, the challenges with NOMA implementation in dense networks and propose future research directions are also discussed.