Channel-aware hierarchical graph network for optimizing sliding window RLNC in challenging wireless environments

Sliding window random linear network coding has emerged as a powerful technique for enhancing the performance of communication networks, particularly in multicast and wireless networks. However, its performance is significantly affected by varying channel conditions such as multipath fading, interference, and noise. To overcome this challenge, novel adaptive channel coefficient engineering is required to dynamically optimize RLNC functionality across varying channel conditions. This paper presents a novel channel-aware hierarchical graph network (CAHGN) that captures the complex interdependencies between various channel features (such as signal-to-noise ratio, path loss, fading factor, interference level, channel magnitude, and phase) and generates an optimized channel coefficient metric to enhance coding performance. The effectiveness of the CAHGN model depends on two key elements: stochastic feature aggregation and dynamic redundancy factor. The stochastic feature aggregation process provides adaptive and weighted channel features by selecting the most relevant features, which are incorporated using the posterior distribution of feature weights. The dynamic redundancy factor controls redundancy levels according to changing feature weights. Simulation results demonstrate that the generated channel coefficient metric improves throughput (by 47.4%), reliability (by 18.2%), and spectral efficiency (by 50%) compared to the traditional RLNC technique under varying wireless channel conditions. Thus, the CAHGN model provides consistent near-optimal performance across different channel environments, reducing recovery time by approximately 60.9%.