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 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