UPNC-MAC: A Collision-Aware and PD-NOMA-Enabled MAC Protocol for Distributed Underwater Acoustic Networks

Distributed Underwater Acoustic Networks (UANs) facilitate flexible, node-initiated transmissions, which are vital for real-time data exchange in marine applications. However, their decentralized architecture faces multiple challenges, including multiuser interference, resource contention, and inefficient channel utilization. Although various decentralized Medium Access Control (MAC) protocols have been proposed, most struggle to effectively balance collision avoidance with transmission efficiency. Power Domain Non-Orthogonal Multiple Access (PD-NOMA) offers the potential to enhance spectral efficiency by enabling simultaneous transmissions. However, conventional PD-NOMA schemes typically depend on centralized power allocation, limiting adaptability in dynamic underwater environments. To overcome both the challenges of decentralized MAC design and the limitations of conventional PD-NOMA, we propose Underwater Power-Domain NOMA with Collision-aware MAC (UPNC-MAC), the first MAC protocol to combine dynamic PD-NOMA with real-time collision prediction for distributed UANs. Unlike traditional PD-NOMA approaches, UPNC-MAC supports user-initiated, on-demand transmissions with adaptive power allocation, achieving greater flexibility. It incorporates real-time channel state analysis and a collision prediction mechanism to proactively detect and mitigate potential transmission conflicts. A flexible user-pairing strategy further selects suitable sender-receiver pairs for concurrent transmissions. At the receiver, Successive Interference Cancellation (SIC) is employed to suppress interference and recover valid signals. This integrated design not only reduces collisions but also maximizes the number of concurrent transmissions, significantly enhancing time-frequency resource efficiency. Simulation results demonstrate that, compared to the other three distributed MAC protocols, UPNC-MAC improves goodput efficiency by at least 3.56× and reduces end-to-end delay by up to 1.65× across most scenarios, while maintaining a comparable protocol overhead.