Correlation-Aided Joint Activity Detection and Channel Estimation for Multidevice Collaborative Massive Access

This article investigates an uplink grant-free massive access (GF-MA) system, where a large number of Internet of Things devices collaborate to achieve complex applications. For this scenario, device activity identification is a challenging problem due to the interference from massive devices and the limitation in the number of pilot sequences. By utilizing the inherent correlation features in multidevice collaborative scenarios, in this work, we present a detection approach that aims to enhance the accuracy of both activity detection and channel estimation (CE). Specifically, we first propose a task-driven activity (TDA) model to capture the active probability in multidevice collaborative scenarios. Subsequently, considering the TDA model, we propose a message-passing-based algorithm named TDA-JDE for device activity detection and CE. The proposed algorithm jointly processes messages containing channel impulse response (CIR), device activity, and task status information to leverage device activity correlation information. Finally, to obtain the parameters in the TDA model, we propose a parameter estimation algorithm based on the expectation-maximization framework with relaxation and reconstruction strategy (EM-RR). Extensive numerical results show that the proposed algorithm can achieve higher detection accuracy when compared with three benchmark schemes in multidevice collaborative massive access (MA) scenarios.