Performance Evaluation of Information Theoretic Criteria-based Path Number Estimators for Cellular-based Positioning with Dense Multipath

Cellular-based localization has emerged as an important means to provide independent positioning and navigation solutions or be integrated with other navigation sensors for enhancing urban and indoor positioning capability. Taking into account the low heights of both the cellular signal transmitters and receivers as well as the complex environments in urban and indoor areas, severe multipath propagation effect originating from the specular reflection and distributed diffuse scattering mostly occurs. This increases the difficulty in accurately extracting the delay and angle measurements for the received cellular multipath signals, with the aim of either combating or benefiting from the multipath effects. In general, path number estimation needs to be first performed, and followed by the multipath parameter estimation algorithms in the cellular positioning receivers. Information theoretic criteria (ITC), such as the classical minimum description length and Akaike information criterion, are commonly used as the path number estimators. However, due to the contribution of dense multipath component (DMC) from distributed diffuse scattering, the noise condition in the channel measurements might violate from the white noise characteristics and degrade the performance of ITC-based path number estimators. This paper first presents the tailored procedures of two kinds of ITC, originally developed for source enumeration in white and colored noise, for path number estimation in cellular positioning receivers. The performance of these ITC under multipath conditions in the absence or presence of DMC is comprehensively assessed by numerical simulations in terms of the probabilities of correct estimation, overestimation, and underestimation. The impact of the key design parameter in these ITCs and the DMC intensity on the path number estimation performance is also discussed. The statistical performance results clearly indicate their performance behavior under different multipath channel conditions, and motivate the necessity of designing more suitable path number estimators for cellular localization applications.