Collision-Isolated Asynchronous Access for High-Density IoT Networks

Abstract

Low Power Wide Area Networks (LPWANs) such as LoRa commonly adopt unslotted ALOHA for its simplicity, but the resulting throughput ceiling (0.183 frames/frame-time) limits scalability. We propose a heterogeneous multi-channel collision resolution system that partitions $M$ channels into contention channels and dedicated collision resolution channels (CRC). By isolating post-collision retransmissions in CRCs, the system prevents them from being interfered with by new packet transmissions on contention channels, improving reliability and utilization. We develop a renewal-reward–based analytical framework to characterize per-channel throughput and to derive the optimal transmission attempt rate and the optimal ratio between contention and collision resolution channels. Building on this analysis, we design an online Bayesian backoff control algorithm that adapts device backoff rates using real-time backlog estimation with minimal computational and communication overhead. Analysis and simulation results show that the proposed system achieves 0.409 frames/frame-time per channel, exceeding the throughput of unslotted ALOHA by 123%, while maintaining bounded delay and stable operation even under time-varying, temporally correlated traffic.