Master of Engineering - Research
School of Electrical, Computer and Telecommunications Engineering
Wang, Zichen, Doppler spread estimation in high mobility wireless communications, Master of Engineering - Research thesis, School of Electrical, Computer and Telecommunications Engineering, University of Wollongong, 2015. https://ro.uow.edu.au/theses/4425
Wireless communication systems have experienced phenomenal growth over the last three decades. Tremendous research efforts have been devoted, andmany important developments have been made. Future wireless communication systems aim at providing reliable data transmission even under high mobility scenarios. One of the challenging issues is the large Doppler spread, which is caused by the high mobility of a wireless terminal and may lead to severe communication performance loss. Doppler spread is a measure of spectral broadening of the rate of change in a mobile fading channel which is proportional to the mobile speed. As Doppler spread is a significant parameter of themobile channel, it is important to a number ofwirelessmobile applicationswhich require the knowledge of the fading rate of channel variations or the mobile speed. Therefore, the large Doppler spread in high mobility wireless communication systems cannot be negligible anymore and an accurate estimate of Doppler spread is imperative to develop future mobile communication systems for high speed vehicles. The objective of this thesis is to design effective algorithms for Doppler spread estimation in high mobility wireless communication scenarios.
Many approaches have been proposed forDoppler spread estimation. The optimal Doppler spread estimation is based on the maximum likelihood principle, which usually involves exorbitant complexity. Other Doppler spread estimation approaches, such as the level crossing rate based approach and the covariance based approach, are with low complexity, but they need a large number of observations and their performance degradation exists in either low Doppler spread range or high Doppler spread range.
As the knowledge of Doppler spread can be used to characterise the time variation of fading channels, it can be employed to enhance channel estimation performance. Many Doppler spread estimation approaches are based on the channel estimates, while the channel coefficients are assumed to be known at the receiver. The mutual dependence of channel estimation and Doppler spread estimation motivates joint Doppler spread and channel estimation in order to achieve better performance.
We propose a new joint Doppler spread and channel estimation approach for Rayleigh fading channels through an iterative process between a Doppler spread estimator and a channel estimator. The proposed Doppler spread estimator is based on the autocorrelation function (ACF) of the estimated channel coefficients, where we devise an ACF lag selection mechanism to maximise the performance of Doppler spread estimation. We use a Forney-style factor graph to represent the fading channel with a first order autoregressive (AR(1)) model and implement a Gaussian message passing (GMP) based channel estimator for Doppler spread estimation, where the AR(1) model coefficients are optimised. Compared to existing joint Doppler spread and channel estimation approaches, the proposed approach achieves similar or better performance with lower complexity.
We also propose an expectation-maximization (EM) based time-varying channel estimation approach. It has been shown that the EMalgorithmmay be viewed as message passing in factor graphs, and for a linear Gaussian system with unknown parameters, the EM algorithm may be implemented by the GMP techniques. With a Forney-style factor graph representation of the fading channel approximated by an AR(1) model, we devise a low-complexity EM-GMP based approach to joint estimate the channel coefficients and the AR(1) model coefficients, where theDoppler spread can be extracted fromthe estimatedAR(1)model coefficients. The proposed EM-GMP based approach has similar complexity per iteration as the conventional EMbased approach, but it significantly outperforms the conventional EM based approach in terms of convergence rate, especially in low Doppler spread range.