Time of arrival estimation of LTE signals for positioning: bounds and algorithms

This thesis presents a research work on the estimation of the time of arrival (TOA) of modern cellular signals for positioning purposes. The Third Generation Partnership Project (3GPP) Long Term Evolution (LTE) signals are analyzed, and the underlying orthogonal frequency division multiplexing (OFDM) based physical layer used in the cellular downlink is exploited. The original contribution presented in the thesis is twofold. Firstly, a framework has been developed for assessing the TOA estimation performance achievable with OFDM signals. The signals are realistically modeled, and different power distributions of the available OFDM sub-carriers have been carefully defined. This allowed new exploration of the TOA estimation performance both in the asymptotic and in the threshold root mean square error (RMSE) regions. Moreover, a novel performance metric based on the shape of the Ziv-Zakai bound curve has been defined, and used to precisely evaluate the boundaries between the threshold and asymptotic RMSE regions. The analysis revealed a trade-off between the threshold RMSE, which is related in practice to sensitivity, and the asymptotic RMSE, which determines the ultimate accuracy. This shows that not only the Gabor bandwidth but also the threshold signal-to-noise ratio (SNR) should be considered when designing reference signals. Secondly, a TOA estimation algorithm has been developed and applied to real LTE OFDM signals collected in multipath indoor and outdoor propagation environments. The new algorithm, referred to as ESPRIT and Kalman filter for time of Arrival Tracking (EKAT), combines a super-resolution algorithm, which performs the multipath separation, with a Kalman filter, which tracks the estimated direct path TOA. In addition, techniques have been extended for combining the received LTE pilot tones in the time, frequency, spatial and also cell ID domains. This exploits the intrinsic diversity offered by the LTE cell specific reference signal (CRS), and showed an improvement in the robustness and in the quality of the TOA estimates. The pseudoranges evaluated with the proposed EKAT algorithm have been used to feed a positioning filter, delivering position estimates with an error smaller than 8 m (50% CEP) in the indoor scenario.