Efficient Soft-Output GNSS Signal Parameter Estimation using Signal Compression Techniques

Abstract

The direct implementation of the maximum likelihood estimator for the time-delay estimation problem is practically intractable for navigation signals due to its complexity, especially when due to multipath reception several superimposed replica are taken into account. In particular, the complexity arises from the evaluation of the likelihood, which requires operations upon a data set of many thousand samples for each code period of the received signal. Thus within conventional navigation receivers tracking loops are used for estimating the signal parameters. Recently it has been shown that signal compression techniques can overcome the problem of computational complexity, as the likelihood function can be formulated efficiently upon a reduced data set of much smaller size compared to the original data, where the reduced data set forms a sufficient statistic for the estimated signal parameters. As a consequence, the evaluation of the likelihood becomes feasible with moderate cost, which might enable a paradigm change in terms of information handover within the receivers, as much more information can be handed over from the signal domain to the navigational part of the receiver. Within this paper a novel approach for integrating a conventional tracking loop with a likelihood evaluator is proposed. It is shown that the parameter information can be kept recursively in a data set of small size in a sequential estimation procedure by using a bank of first order IIR filters, such that an approximate posterior can be reconstructed from the filtered data set. The performance of the proposed method is assessed by computer simulations. The results show that the conventional delay lock loop is outperformed with respect to noise performance as well as with respect to the multipath bias whilst providing capabilities with respect to multipath monitoring and mitigation.