ION GNSS 2012
Session D3: GNSS Algorithms & Methods 1: Signal Processing

Title: Probabilistic Multipath Mitigation for GNSS-based Vehicle Localization in Urban Areas
Author(s): M. Obst, C. Adam, G. Wanielik, Chemnitz University of Technology, Germany; R. Schubert, BASELABS GmbH, Germany
Date/Time: Thursday, September 20, 2012, 11:26 a.m.
Room: 204 (NCC)

For a large number of advanced driver assistance systems (ADASs) it is necessary to constantly gather accurate and reliable knowledge of the position of the host vehicle. As off-the-shelf GPS receivers are nowadays deployed in most standard commercial vehicles, satellite-based localization seems to be the most promising technology to solve this task.

Beside the accuracy of a navigation solution, the reliability and availability in terms of integrity are two major aspects of intelligent transportation systems (ITSs), as well. While satellite-based positioning reaches rather descent performance-e.g. street level-under good geometric constellation and open-sky reception conditions, there are often situations with degraded accuracy, too. Especially when the focus of such applications is put into dense urban environments, GNSS-specific problems like satellite signal outage or signal reflection, commonly known as multipath, have to be considered. Even if the full blockage of GNSS signals cause by buildings or foliage may be directly detectable-the expected measurements are not received at all-for reflections this is more difficult. Normally, GNSS observations which are subject to multipath and used throughout the localization process, lead to an unmodelled bias in the final position estimate and therefore violate the promised protection/integrity level. Moreover, multipath in dynamic ITS applications is a rapidly changing phenomenon which makes it hard to detect and predict.

Considering the given limitations for classical GNSS-only localization in urban areas, it can be summarized that reliable positioning is still a challenging task. Having ITS applications like tolling, which charge the user on road or lane-level usage, or green driving assistants for efficient path planning in mind; it is easily understandable that further work on this area is still needed.

In literature, there already exist different approaches which try to detect and even mitigate multipath effects on GNSS observations. First of all, the usage of high-quality antennas (e.g. with choke ring design) has proven to autonomously detect multipath affected measurements and exclude them from the positioning process. Unfortunately, such antennas are expensive in price and large in dimensions, so that an introduction to mass-market automotive applications seems to be unlikely. The same restriction applies to antenna arrays which are another possible solution for multipath detection. In contrast to receiver hardware improvements, algorithmic methods are proposed as well. For example, the Receiver Autonomous Integrity Monitoring (RAIM) algorithm, initially developed for safety reasons in aviation, is able to detect one possible outlier from a given set of GNSS observations. Even if there are extensions like NIORAIM, which can handle more than one outlier, the concept of exclusion seems to be too strict for road applications. On the other side, the promising concept of Bayesian RAIM requires high computational demands. Another interesting approach for efficient multipath detection is the use of an infrared camera installed on the roof top of a vehicle to visually detect blocked lines of sight between the GNSS antenna and the satellite. However, the requirement of an additional sensor lowers the chances for wide mass-market introduction. Recent work introduces 3D environment models of urban areas to directly predict multipath situations. Unfortunately, such map based approaches heavily depend on the quality and accuracy of the provided map data (this also includes the problem of outdate maps).

The aim of this work is to propose a generic online GNSS positioning algorithm for vehicular applications with integrated probabilistic multipath detection and mitigation using the Bayes Filter framework. The algorithm can be used to autonomously increase localization accuracy and integrity without additional hardware sensors. Even if the algorithm is not restricted to urban areas, its biggest impact can be expected in strong multipath-affected situations.

Basically, classical GNSS positioning is done by simultaneously measuring several pseudoranges between the receiver antenna and the satellites in space at one epoch. From this set of measurements a positioning estimate is generated through a non-linear least squares solution. The algorithm presented in this work now introduces the assumption that the measurement set {z} consists of two subsets {z}^c and {z}^m which contain the correct pseudoranges and the erroneous pseudoranges (with multipath), respectively. Unfortunately, it is not known which measurements are correct and which are subject to multipath. Thus, the proposed algorithm defines the set of all possible association hypotheses. From a probabilistic point of view, these hypotheses are represented by discrete association events A_i^m which are grouped into subsets. Such a subset A_m contains all association events which are based on the assumption that exactly m measurements were measured without multipath. For each of these hypotheses, now a Bayes filter update and correct step-which only includes the true measurements-using the last system state is performed. The result is a Gaussian mixture representation which is finally approximated through one single Gaussian probability density function.

The positioning algorithm described will be tested and validated with real data recorded during a test drive under typical urban conditions. To assess the absolute positioning performance, the results will be compared to a high reliable reference sensor system (NovAtel SPAN System with RTK and IMU). Furthermore, the results will be compared to classical RAIM solutions and post-processing methods. Moreover, the multipath detection performance will be assessed in a software simulation, where multipath signals are generated for an urban scenario. Within this work, the multipath simulation software and its properties will be explained as well.

Within this paper, a generic GNSS positioning algorithm with integrated multipath detection and mitigation was proposed and validated. The algorithm was used in combination with a vehicular motion model to show respect to the physical constrains of typical legal manoeuvres. It could be shown, that even without additional sources like 3D models or expensive antenna hardware, the positioning solution could be autonomously improved and a more appropriate integrity estimate can be provided.

Previous Abstract Return to Session D3 Next Abstract