Multi-Service Oriented Joint Channel Estimation and Multi-User Detection Scheme for Grant-Free Massive MTC Networks

To satisfy the highly heterogeneous requirements of Internet of Things applications for the sixth-generation (6G) networks, the machine-type communication (MTC) aims to support multiple types of services having diverse traffic demands, which will cause significant challenges in the grant-free based channel estimation (CE) and multi-user detection (MUD) scheme design for massive MTC (mMTC) networks. To address this challenge, we develop a multi-state Markov chain based transmission model to characterize the diverse time-varying traffic demands for MTC users, where the temporal correlation of user activity and the data length diversity are jointly exploited. Based on the developed transmission model, a multi-service oriented joint CE-MUD scheme is proposed to realize the efficient CE, user activity identification and data detection. Specifically, we first construct the joint block sparse structure for the transmitted pilot and data signals to fully explore the structured sparsity of the pilot and data symbols. Then, we convert the joint CE-MUD into a maximum a posteriori probability (MAP) problem such that the block sparsity of the transmitted signals and the diverse traffic demands provided by the established transmission model can be efficiently exploited. Moreover, we further develop an adjustable prior probability aided Bayesian sparsity adaptive matching pursuit (APP-BSAMP) algorithm to efficiently solve the formulated MAP problem. In the proposed algorithm, we first adjust the prior user activation probabilities through the approximate message passing (AMP) based detector to reduce the impact of the misestimation of user transmission status. Then, we jointly reconstruct the transmitted pilot and data signals under the Bayesian pursuit framework, where the active user set is obtained by maximizing the posterior probabilities. Simulation results show that the proposed scheme can achieve a substantial performance gain over existing methods.