Link adaptation and multi-objective resource optimization in intelligent wireless networks using power-domain non-orthogonal multiple access

In intelligent wireless networks, achieving reliable communication between vehicles and infrastructure is critical for enhancing user experiences and addressing the demands of next-generation networks. However, maintaining robust connectivity is challenging due to urban environments and network variability in vehicle-to-infrastructure (V2I) and vehicle-to-vehicle (V2V) communication systems. This paper proposes a novel framework for link adaptation and multi-objective resource optimization, leveraging power-domain non-orthogonal multiple access (NOMA) and blind reconfigurable intelligent surfaces (IRS). The proposed method incorporates a multi-agent Deep Reinforcement Learning (DRL) model, where each agent dynamically allocates resources by optimizing power control and scheduling based on real-time network data and traffic patterns. Our approach uses IRS to enhance signal quality and extend coverage even in complex and highly dynamic environments, while the multi-agent DRL framework with graph attention mechanisms enables decentralized and scalable resource management. The agents learn from the environment, adjusting resource allocation across multiple objectives, such as maximizing throughput, improving energy efficiency, and ensuring reliable connectivity. By optimizing power allocation and link adaptation, the framework addresses the challenges of channel variability and improves network performance without requiring precise channel state information (CSI). Simulation results show that the proposed approach achieved significant improvements in both energy efficiency and throughput compared to conventional methods such as NFVMCH and HetVNet. Additionally, the throughput of TRONICS scales effectively, reaching nearly 55 Mbps/Hz with 60 users per cluster, while competing methods only manage up to 26 Mbps/Hz, underscoring its advanced resource optimization capabilities.