Efficient Multi-Stage Active Device Identification for Massive Random Access

Efficiently identifying active devices with minimal latency is crucial in massive machine-type communication networks characterized by sparse and sporadic device activity. This paper addresses the above challenge by introducing a novel active device identification strategy which employs a multi-stage framework that iteratively refines partial estimates of active devices through feedback and hypothesis testing, leading to an exact recovery. In our proposed method, active devices transmit binary preambles independently in each stage, utilizing feedback signals from the BS. Meanwhile, the BS utilizes non-coherent binary energy detection. In addition to theoretical bounds, practical implementations of our multi-stage active device identification schemes using Belief Propagation (BP) techniques are presented. Our simulation results demonstrate that the multi-stage strategy is superior to the single-stage one introduced in our earlier work and performs close to the theoretical bound, even when considering overhead costs related to feedback.