A Partial Gaussian Tree Approximation (PGTA) Detector for Random Multiple Access Oriented SCMA Uplink With Codebook Collisions

Abstract

By means of random multiple access (RMA), there is a prevailing tendency for massive machine-type terminals to have access into networks without numerous signaling interactions and frequent resource scheduling. Being one of the promising techniques to fulfill this requirement, sparse code multiple access (SCMA) is anticipated to notably broaden connections for the communication system. However, when applying RMA in the SCMA uplink, as massive terminals would randomly transmit their packets in an uncoordinated way, it is quite possible for some terminals to share an identical codebook so as to introduce the codebook collision problem. In this case, an indefinite number of terminals with possible codebook collisions would competitively share a same block of channel resources, resulting in an overloading status and deteriorated performance. To deal with this problem, we illustrate first the potential utilization of codebook collisions for RMA and propose then a partial Gaussian tree approximation (PGTA) detector to relieve heavy computational burden due to overloading conditions. With approximate decomposition of likelihood functions that has rarely been discussed, the proposed PGTA detector with proper message passing rules is able to approach the performance of a traditional message passing detector with considerably lower complexity.