Direct Positioning with Channel Database Assistance

TL;DR

This paper demonstrates that utilizing wall and building location knowledge significantly improves radio source positioning accuracy, especially with fewer antennas, by developing a generalized MUSIC algorithm that accounts for wall reflections.

Contribution

It introduces a generalized MUSIC algorithm incorporating wall reflection parameters as nuisance variables, enhancing localization performance with channel database assistance.

Findings

Wall knowledge reduces location error by up to 100x.

Increased antennas diminish the benefit of wall knowledge.

Without wall knowledge, errors can be substantially higher.

Abstract

When we have knowledge of the positions of nearby walls and buildings, estimating the source location becomes a very efficient way of characterizing and estimating a radio channel. We consider localization performance with and without this knowledge. We treat the multipath channel as a set of "virtual receivers" whose positions can be pre-stored in a channel database. Using wall knowledge, we develop a generalized MUSIC algorithm that treats the wall reflection parameter as a nuisance variable. We compare this to a classic MVDR direct positioning algorithm that lacks wall knowledge. In a simple scenario, we find that lack of wall knowledge can increase location error by 7-100x, depending on the number of antennas, SNR, and true reflection parameter. Interestingly, as the number of antennas increases, the value of wall knowledge decreases.