Bayesian PIC-DSC detector

This detection method is proposed in A Bayesian Receiver With Improved Complexity-Reliability Trade-Off in Massive MIMO Systems by Alva Kosasih. It has three modules: BSO does the parallel interference cancellation, BSE does the Bayesian symbol estimation, and DSC does the update.

Kosasih, A., Miloslavskaya, V., Hardjawana, W., She, C., Wen, C. K., & Vucetic, B. (2021). A Bayesian receiver with improved complexity-reliability trade-off in massive MIMO systems. IEEE Transactions on Communications, 69(9), 6251-6266.

How to install

Currently, we offer three options to install this tool.

• Install through Matlab Add-Ons
  • Install through Matlab Get Add-Ons: search whatshow_phy_detect_bpic and install it.
  • Install through .mltbx: Go to Releases to download the file in the latest release to install.
• Install through pip

  pip install whatshow-phy-detect-bpic

  • import this module

    from whatshow_phy_detect_bpic import BPIC
    

• Install through git under another local repositiory

  git submodule add [email protected]:whatshow/Phy_Detect_BPIC.git Modules/Detect_BPIC

  • import this module
    • Matlab

      addpath("Modules/Detect_BPIC");

    • Python

      if '.' not in __name__ :
          from Modules.Detect_BPIC.BPIC import BPIC
      else:
          from .Modules.Detect_BPIC.BPIC import BPIC

How to use

All Bayesian PIC-DSC detector codes are uniform in matlab and python as a class of BPIC. This class is the whole process of the detection. This section will illustrate the methods of this class following the detection process.

• BPIC
  @constellation: the constellation, a vector.
  @bso_mean_init: 1st iteration method in BSO to calculate the mean. Default: BPIC.BSO_INIT_MMSE, others: BPIC.BSO_INIT_MRC, BPIC.BSO_INIT_ZF (BPIC.BSO_INIT_NO should not be used but you can try)
  @bso_mean_cal: other iteration method in BSO to calculate the mean. Default: BPIC.BSO_MEAN_CAL_MRC (BPIC.BSO_MEAN_CAL_ZF should not be used but you can try)
  @bso_var: use approximate or accurate variance in BSO. Default: BPIC.BSO_VAR_APPRO, others: BPIC.BSO_VAR_ACCUR
@bso_var_cal: the method in BSO to calculate the variance. Default: BPIC.BSO_VAR_CAL_MRC, others: BPIC.BSO_VAR_CAL_MRC (BSO_VAR_CAL_ZF should not be used but you can try)
  @dsc_ise: how to calculate the instantaneous square error. Default: BPIC.DSC_ISE_MRC, others: BPIC.DSC_ISE_NO, BPIC.DSC_ISE_ZF, BPIC.DSC_ISE_MMSE
@dsc_mean_prev_sour: the source of previous mean in DSC. Default: BPIC.DSC_MEAN_PREV_SOUR_BSE, others: BPIC.DSC_MEAN_PREV_SOUR_DSC
@dsc_var_prev_sour: the source of previous variance in DSC. Default: BPIC.DSC_VAR_PREV_SOUR_BSE, others: BPIC.DSC_VAR_PREV_SOUR_DSC
@min_var: the minimal variance.
  @iter_num: the maximal iteration.
  @iter_diff_min: the minimal difference in DSC to early stop.
  @detect_sour: the source of detection result. Default: BPIC.DETECT_SOUR_DSC, others: BPIC.DETECT_SOUR_BSE.
  

  // paper version 1: for BSO, MMSE in 1st iteration but MRC in others
  bpic = BPIC(sympool);
  // paper version 2: MRC in all iterations
  bpic = BPIC(sympool, "bso_mean_init", BSO_MEAN_INIT_MRC); % matlab
  bpic = BPIC(sympool, bso_mean_init=BSO_MEAN_INIT_MRC); # python
  // other configurations
  % matlab
  bpic = BPIC(sympool, "bso_mean_init", BPIC.BSO_MEAN_INIT_MMSE, "bso_var", BPIC.BSO_VAR_APPRO, "bso_var_cal", BPIC.BSO_VAR_CAL_MMSE, "dsc_ise", BPIC.DSC_ISE_MMSE, "detect_sour", BPIC.DETECT_SOUR_BSE);
  # python
  bpic = BPIC(sympool, bso_mean_init=BPIC.BSO_MEAN_INIT_MMSE, bso_var=BPIC.BSO_VAR_APPRO, bso_var_cal=BPIC.BSO_VAR_CAL_MMSE, dsc_ise=BPIC.DSC_ISE_MMSE, detect_sour=BPIC.DETECT_SOUR_BSE);

• detect: the estimated symbols from Tx
  @y: the received signal, a vector
  @H: the channel matrix, a matrix
  @No: the noise power, a scalar
  @sym_map: whether use hard mapping
  

  // symbol estimation - soft
  x_est = bpic.detect(y, H, No);
  // symbol estimation - hard
  x_est = bpic.detect(y, H, No, "sym_map", true); % matlab
  x_est = bpic.detect(y, H, No, sym_map=true); # python

Samples

Before running any sample code, please make sure you are at the root path of this repository. Also, Matlab codes require running init in the command window first to load directories.

• Test
  • Test/test_mimo: test the performance of Bayesian PIC-DSC detector in MIMO case.
  • Test/test_case_01: compare the output from this module and Alva's original code.