Importance sampling simulation of a symbol-by-symbol multiuser detector

Multiuser detection has been shown to outperform the conventional matched-filter receiver on the code-division multiple access (CDMA) channel. Our previous work developed a multiuser detector in which extended Kalman filter (EKF) estimates for user amplitudes and time delays are computed and used in a MAP detector. Importance sampling (IS) has been shown to greatly increase simulation efficiency for a variety of detection problems. In this paper, we describe an IS method which is compatible with our nonlinear time-varying detector. The simulation achieves IS gains on the order of 10⁸, and the implications for computational accuracy are discussed. Simulation results show that the detector greatly outperforms the matched filter detector even under perfect power control, when significant user interference is present. In fact, the detector performance rivals that of BPSK on a purely Gaussian channel, indicating that the interference from other users has been “canceled”