On December 15, 2020, Markus Ulmschneider successfully defended his Ph.D. thesis Cooperative Multipath Assisted Positioning. The Ph.D. examination committee was formed by Prof. Dr. Matthias Kuhl (Institute for Integrated Circuits) and the two reviewers Prof. Dr.-Ing. Gerhard Bauch and Prof. Dr. Klaus Witrisal (Graz University of Technology).

Abstract

Location-awareness has become an important issue in modern life. In environments with good view to the sky, global navigation satellite systems can provide very accurate location estimates. Though, their positioning performance decreases significantly if the view to satellites is blocked. Indoors, no localization solution may be obtained at all.

Terrestrial signals such as from telecommunication base stations or wireless local area network routers can be used for indoor localization. However, the localization performance may suffer from multipath propagation when traditional propagation delay-based algorithms are applied. The transmit signal is reflected, scattered and diffracted in the environment, distorting the signal and leading to a bias in the delay estimates.

With multipath assisted positioning, a new approach has emerged, that exploits the information in multipath components (MPCs) by treating them as line-of-sight signals from virtual transmitters as illustrated in Figure 1. The MPC arriving at the receiver is regarded as a line-of-sight signal from the virtual transmitter vTx. While the locations of the physical and virtual transmitters are generally unknown, they can be estimated with simultaneous localization and mapping (SLAM). One SLAM based multipath assisted positioning algorithm is Channel-SLAM, where the user location is estimated simultaneously with creating a map of physical and virtual transmitters.

Figure 1: The idea of multipath assisted positioning.

The Channel-SLAM algorithm is limited to single users. There are many settings where multiple users move in the same indoor environment, such as in shopping malls or public buildings. Within this thesis, the Channel-SLAM algorithm is extended by cooperation among users in terms of exchanging maps of transmitter locations. This cooperation can drastically increase the positioning performance of all users.

However, Channel-SLAM is only a relative localization system, as no relation of the user location to an absolute coordinate frame is known. The transformation parameters relating the coordinate systems of a user to the coordinate system of a map the user obtains from a different user are unknown. In addition, the correspondences among transmitters observed by the user and transmitters in the map are unknown. Estimating these transformation parameters and correspondences is referred to as map matching.

A robust map matching scheme is crucial for such cooperation. Since the lack of diversity among virtual transmitters leads to ambiguities in map matching, the transmitter state space is augmented by information from where transmitters are visible. A transmitter is visible, if its signal can be received in a LoS condition, as depicted in Figure 2. It shows the same scenario as Figure 1, but with an obstacle blocking the transmit signal. Two new tracking filters for Channel-SLAM are derived that map not only the locations, but also the visibility information on transmitters. The visibility information increases the robustness of Channel-SLAM by facilitating the detection of loop closures. In addition, it is used to derive a robust data association scheme. Data association in multipath assisted positioning refers to associating signal components with transmitters, leading to robust and accurate long-term SLAM. Finally, the increased transmitter diversity through visibility information improves the positioning performance of Channel SLAM significantly by a more robust data association and by enabling robust user cooperation.

Figure 2: In (a) and (b), visibility areas of the transmitters Tx and vTx, respectively, are shown. From areas colored green, the respective transmitter is visible, while from red areas it is not.