Active User Detection And Low Complexity Multi User Detection For Unsourced Multiple Access
Abstract
We consider two problems related to uncoordinated multiple access. The first is the design of schemes that can identify the set of active users. The second is the design of low complexity multiuser detection schemes. We consider a multiple access scheme where each user encodes its information by an error correcting code and spreads the coded bits using a spreading sequence that is chosen from a master set of spreading sequences based on part of each user’s message. On the receiver side, the set of active spreading sequences need to identify first and then the bits of the user have to be detected and decoded.
We consider four strategies for evaluating the set of active spreading sequences. The first scheme is a correlation-based energy detector. The second scheme is based on machine learning, which uses the histogram of the output of a matched-filter(MF) as input to a neural network(NN) model. The third scheme is based on using a hypothesis test on the outputs of the matched filter. The fourth scheme is a 2-bit combined energy detector, which is the same way for the original energy detector but combining two bits to consider more about the variable case that can happen in synchronizing spreading sequences.
In the second part of the thesis, assuming the set of active sequences is known, an MMSE estimator is implemented to perform log-likelihood ratios(LLRs) for the active sequences. But inverting the whole active sequences matrix in MMSE has large time complexity. We propose using a clustering method to reduce matrix size. After this active sequences are passed to a list decoder of polar code proceeding iteratively by subtracting the interference due to the successfully decoded sequence from the received signal and repeat the MMSE estimator process on the residual received signal. The cluster size and a decoding schedule are optimized using Monte Carlo simulations.
Subject
Active User DetectionLow Complexity Multi User Detection
Unsourced Multiple Access
IoT
Wireless Communication
Citation
Jeon, Nicholas (2021). Active User Detection And Low Complexity Multi User Detection For Unsourced Multiple Access. Master's thesis, Texas A&M University. Available electronically from https : / /hdl .handle .net /1969 .1 /195837.