A RIS-Programmable hybrid uplink for grant-Free NOMA networks with distributed signal aggregation

Abstract

Grant-free uplink transmission enables scalable connectivity in dense wireless networks but poses fundamental challenges for reliable signal aggregation under interference- and collision-dominated access. Over-the-air analog aggregation offers low-latency superposition-based processing but is highly vulnerable to collision-induced distortion, whereas digital decoding improves reliability at the expense of reduced parallelism and increased overhead. Reconfigurable intelligent surfaces (RISs) provide a programmable propagation mechanism that can actively reshape uplink signal superposition, rather than merely enhancing individual channel gains. This paper proposes a RIS-programmable hybrid uplink framework for grant-free non-orthogonal multiple access (GF-NOMA) systems, in which the wireless environment is dynamically configured to enable either coherent analog aggregation or successive interference cancellation (SIC)-based decoding depending on instantaneous access conditions. A lightweight adaptive control policy jointly selects the uplink aggregation mode and RIS phase configuration on a per-round basis, allowing explicit regulation of aggregation distortion and access reliability without scheduling or retransmission control. Analytical results characterize the bias-variance behavior of the aggregated gradients under hybrid operation, revealing how RIS-induced coherence and partitioning govern aggregation accuracy. Simulation studies using distributed learning workloads demonstrate significant improvements in convergence efficiency and robustness under dense access, heterogeneous data distributions, and interference-dominated regimes compared with fixed-mode RIS-assisted and grant-free baselines. These results confirm the effectiveness of RIS-enabled adaptive uplink control for reliable signal aggregation in large-scale grant-free wireless networks.