Rateless Multiple Access: Asymptotic Throughput Analysis and Improvement With Spatial Coupling

A novel random code-domain non-orthogonal multiple access (NOMA) framework, rateless multiple access (RMA), is studied in this paper. In RMA, instead of granting each user specific radio resources in a fixed and centralized manner, the access point simply assigns to it a random access control function, according to which the user independently chooses a pseudo-random number of symbols each time and transmits their weighted sum until receiving an acknowledgment that indicates the successful recovery of its information. In this way, the transmission process of each individual user as well as that of all the users as a whole, resembles a special linear superposition rateless encoder and, thus, information can be retrieved with the low-complexity belief propagation (BP) algorithm at the access point. In this paper, the asymptotic throughput of RMA with BP decoding is analyzed. Then, in order to bridge the performance gap between BP decoding and the optimal MAP (maximum a posteriori) decoding, we apply the most recently developed coding technique of spatial coupling to RMA and propose an enhanced version, SC-RMA, in which all the user codewords are properly spatially-coupled across different subsets of REs. We prove that SC-RMA asymptotically approaches the channel sum-rate capacity with high adaptability and low-signaling overhead, which makes it a viable candidate for massive access.