No abstract

No abstract

This paper presents a novel time-variant ML channel estimator for mobile radio navigation receivers. Our novel ML channel estimator enables the coherent noise averaging over several hundred codewords for time-variant channel phasors. Compared to the conventional incoherent summation of loglikelihood functions or compared to the conventional timeinvariant log-likelihood functions, we avoid the squaring loss (SL) completely. The novel time-variant log-likelihood function compared to the conventional time-invariant log-likelihood function yields an SNR gain of up to 15 dB for an observation interval of 200 ms. Additionally, since the novel time-invariant log-likelihood functions only require a Slepian subspace of a small dimension, the computational complexity of our novel timevariant ML channel estimation does not exceed the computational complexity of the conventional time-invariant ML channel estimation.

Abstract-The positioning performance of global navigation satellite systems (GNSSs) mass market receivers severely degrades when the received satellite signals are subject to multipath propagation. Therefore, the estimation of several unknown channel amplitudes and taps in a multipath environment is an important approach to mitigate the multipath effects. In professional receivers, viable multipath mitigation approaches are the maximum likelihood (ML) estimator, the expectation maximization (EM) approach and the space alternating generalized expectation maximization (SAGE) algorithm. However, all methods require a high computational complexity when used in spread spectrum systems due to long spreading sequences. Therefore, one contribution of this paper is that we apply subspace methods to decrease the computational complexity before executing the above iterative estimation algorithms. Further, we assess respective performance of the algorithms and the future Galileo navigation system by using the Galileo receiver analysis and design application (GRANADA) simulator. The complexity reduction algorithms are specifically adjusted to the E1 Galileo binary offset carrier (BOC) signal, which superimposes data and pilot signals. Moreover, we adapt the complexity reduction so that it can handle any sampling frequency that is not necessarily an integer multiple of the chip rate.

No abstract