Markov Chain, Monte Carlo Global Search and Integration for Bayesian, GPS, Parameter Estimation

Bayesian estimation techniques are applied to the problem of time and frequency offset estimation for Global Positioning System receivers. The estimation technique employs Markov Chain Monte Carlo (MCMC) to estimate unknown system parameters, utilizing a novel, multi-dimensional, Bayesian, global optimization strategy for initializing a Metropolis-Hastings proposal distribution. The technique enables the design of a high performance multi-user GPS receiver, capable of overcoming the near-far problem when the relative signal power is on the order of 5 dB (single antenna element) and 20 dB (4 antenna element array) and providing dramatically improved performance over conventional matched filter techniques against interference and jamming when the relative jammer and satellite signal power is on the order of 20 dB (4 antenna element array). INTRODUCTION A new class of GPS receivers is proposed herein that offers improved performance in heavy interference and heavy multipath environments [1-4, 10]. Instead of using a simple sliding correlator to measure time of arrivals (TOAs) and Doppler frequencies, we employ a multi-emitter, statistical signal model, similar to that employed in multi-user detection (MUD) receivers [12-14]. We also propose jointly estimating Doppler frequencies and TOA using nearly optimal Bayesian estimation techniques [1-4]. The resulting receiver enjoys improved performance, is able to receive more GPS satellite signals, has multipath immunity and elegantly resolves the near-far problem when the relative signal power between the weakest