Bayesian Synthetic Likelihood for Calibration of Stochastic Radio Channel Model

Abstract

This paper presents a novel Bayesian Synthetic Likelihood (BSL) method for calibration of stochastic radio channels without multi path parameter estimation. To calibrate a stochastic channel model, we apply a Markov Chain Monte Carlo (MCMC) algorithm with a Metropolis accept/reject criterion and synthetic likelihood obtained from data generated using the model. The proposed method is applied to calibrate the Turin model and the polarized propagation graph model. Simulation examples show that the BSL method yield similar calibration accuracy to the state-of-the-art method based on Approximate Bayesian Computation (ABC).