Efficient Massive Machine Type Communication (mMTC) via AMP
share
We propose efficient and low complexity multiuser detection (MUD) algorithms for Gaussian multiple access channel (G-MAC) for short-packet transmission in massive machine type communications. We formulate the problem of MUD for the G- MAC as a compressive sensing problem with virtual sparsity induced by one-hot vector transmission of devices in code domain. Our proposed MUD algorithms rely on block and non-block approximate message passing (AMP) for soft decoding of the transmitted packets. Taking into account the induced sparsity leads to a known prior distribution on the transmit vector; hence, the optimal minimum mean squared error (MMSE) denoiser can be employed. We consider both separable and non-separable MMSE denoisers for AMP soft decoding. The effectiveness of the proposed MUD algorithms for a large number of devices is supported by simulation results. For packets of 8 information bits, while the state of the art AMP with soft-threshold denoising achieves 5/100 of the converse bound at Eb/N0 = 4 dB, the proposed algorithms reach 4/9 and 1/2 of the converse bound.
READ FULL TEXT
