Improved multicanonical algorithm for outage probability estimation in MIMO channels

Abstract

Multicanonical Monte Carlo (MMC) is an adaptive importance sampling technique which employs a blind adaptation algorithm to converge to the optimal biasing distribution. In this paper, we propose an improved MMC algorithm for fast estimation of outage probabilities in Multiple Input Multiple Output (MIMO) channels. The algorithm uses an improved estimator which can provide smooth estimates with high reliability at very low error probabilities. The proposed estimator uses moving average filtering to smooth the visits histograms at each iteration thereby reducing the stochastic fluctuations between iterations. We compare the proposed estimator with the well known Berg's update and the simulation results show that the new estimator can accurately estimate lower error probabilities with the same number of samples.