Joint Channel Estimation and Active-User Detection for Massive Access in Internet of Things—A Deep Learning Approach
Zhendong Mao, Xiqing Liu, Mugen Peng, Zijie Chen, Guiming Wei
Abstract
For conventional signaling, the length of the orthogonal pilot is required at least equal to the total number of user antennas. However, it is not recommended in the Internet of Things (IoT) due to the expensive cost paid in massive connectivities. Thanks to the sporadic nature of the massive connected users where a considerable fraction of users are inactive within a coherence time, the nonorthogonal pilot can be utilized with the joint channel estimation and active-user detection being modeled as a compressive sensing problem. According to the different antenna configuration methods employed by the base station, the constructed problems in this work are formulated into the single measurement vector and the multiple measurement vectors recovery problems. Also, we develop a model-driven deep learning algorithm to solve the problems based on the traditional alternative direction method of multipliers (ADMM) algorithm, where the iteration operation is unfolded into the network layer. The network parameters are learned with the help of the stochastic gradient descent algorithm. Simulation results show that the proposed approach can achieve better performance than an ADMM algorithm under the same computational complexity.