Performance Comparison of Time-of-Arrival Estimation Techniques for LTE Signals in Realistic Multipath Propagation Channels

Abstract

Long term evolution (LTE) signals have the potential to be used for positioning, especially in challenging environments such as in urban area or indoors. Among a number of positioning techniques supported by LTE, the time-of-arrival (TOA)-based technique has attracted attention due to the achievable high positioning accuracy. However, if multipath propagation is present, the TOA measurements obtained with the conventional method may include large biases and cause significant positioning errors. To solve this, advanced TOA estimation algorithms are necessary for dealing with LTE signals. The first main purpose of this paper is to summarize several existing advanced TOA estimators, i.e., first peak detection (FPD) based on two different threshold criteria, information theoretic criteria (ITC), super-resolution algorithm (SRA), and delay-locked loop (DLL). The second main purpose is to evaluate the TOA estimation performances of these techniques with multipath propagation effects by simulations. With respect to the FPD, ITC, and SRA techniques, the probability distribution and statistical results of the TOA estimation errors are obtained for a wide range of signal-to-noise power ratio (SNR) values and several types of realistic multipath propagation channels. The dependence of the estimation accuracy on the SNR condition, channel type, and key parameter setting is also analyzed for each technique. With respect to the DLL, the multipath error envelope metric is specifically evaluated considering different signal bandwidths to indicate the multipath mitigation performance. The narrow pull-in range limits the applicability of the DLL in dynamic environments. Finally, the quantitative root mean square (RMS) errors are compared for revealing suitable TOA estimation techniques in various situations. These techniques provide similar estimation accuracy for medium SNR values, but exhibit significantly different performances under relatively low or high SNR conditions. Moreover, some other performance characteristics of these techniques are also discussed.