pp 1-7 | Cite as
Assessment of a GNSS/INS/Wi-Fi Tight-Integration Method Using Support Vector Machine and Extended Kalman Filter
- 40 Downloads
Wi-Fi derived positions have been used in the past few years as a complementary source of positioning information for GNSS and Inertial Systems (INS). Ubiquitous positioning that transitions from indoors to outdoors and vice-versa is currently a hot topic of research. In this context, this study aims to analyze the potential of directional antennas sequentially tracking Wi-Fi signals on the 11 channels around the 2.4 GHz frequency in order to serve as an integrated signal for GNSS and INS positioning. Considering, as an example, a single point positioning (SPP) strategy coupled with an INS, the use of directional antennas can be beneficial in order to provide absolute directions of travel by the means of a Support Vector Machine (SVM) lane matching. In order to test the given hypothesis, real-world experiments were performed in areas with and without obstruction in an urban environment. Using a post-processed, smoothed in both forward and backward modes, and finally edited post-processed kinematic (RTK) solution as a reference, the solution integrating SPP GNSS, INS and Wi-Fi was assessed in terms of accuracy. Preliminary results show that such a combination of the directional antennas along with GNSS and INS and their respective SVM and EKF filters, can provide sub-meter accuracy at all times without the need of precise orbits or differential corrections, increasing solution availability, reliability and accuracy on a scalable and cost-effective way.
KeywordsGNSS INS Sensor integration SVM Vehicle navigation Wi-Fi
Sensor integration techniques are a contemporary topic of research, since mobile platforms can now achieve the computational power required for such tasks. The use of Global Navigation Satellite Systems (GNSS) as a source of positioning, albeit widespread in applications, has limitations particularly noticeable in urban environments. The main damaging effects on GNSS positioning in urban environments are signal obstructions and reflections, causing problems with both signal quality (usually yielding low signal-to-noise ratios), and low number of visible satellites. Several studies have been performed in order to quantify, analyse, and overcome such limitations using different techniques, such as solutions using new GNSS signals (Hsu et al. 2015), novel mathematical models to constrain the accumulating INS errors (Grejner-Brzezinska et al. 2001), sensor integration techniques, and signal-of-opportunity concepts (Groves 2011). With the growing demand for accurate and reliable urban positioning fueled by the advent and popularization of autonomous vehicles, improvements in this area are not only of academic value, but also of immediate practical applications. In this context, the cost and processing power requirements of the solutions are of paramount importance. Accurate and ubiquitous positioning equipment, such as GNSS/INS integrated NovAtel SPAN®, may have comparable costs to a semi-autonomous vehicle, such as the Tesla Model 3, therefore, not being capable of reaching mass-market applications.
With the aforementioned situation in mind, and considering the challenges of positioning in urban areas, the integration of cost-effective sensors and improvements in mathematical models are viable and tested alternatives to overcome such challenges (Grejner-Brzezinska et al. 2001; Groves 2008). In this study, a combination of Wi-Fi signals recorded by directional antennas, an INS and a GNSS (GPS + GLONASS) receiver is studied in order to asses how information from directional Wi-Fi antennas can be integrated in a GNSS/INS solution and what is the benefit of it. The remainder of this paper is divided among the following sections: a brief review of traditional GNSS/INS integration techniques, an overview of Wi-Fi positioning techniques, the development of an integration technique between the three systems, experiment design, and, finally, results and conclusions.
2 GNSS/INS Integration Techniques
3 Outdoor Wi-Fi Positioning
4 Integration Between GNSS, INS and Wi-Fi
Given the rapid diverging characteristic of INS position propagation, in the absence of GNSS for a considerable period of time – situation common in urban areas – any system relying on this information may suffer from unreliable position estimates. On an autonomous vehicle scenario, this situation may cause a disengagement of the auto-pilot or possibly accidents.
5 Experiment and Results
For the SVM training algorithm, data from several previous surveys in the area were utilized. With a rate of 90% of training data for 10% of test, the accuracy of the classifier was 82% (N = 7,422) on the correct lanelet. On a more detailed analysis of the misclassifications (about 130 points out of 7,422), it was found that they happened in situations where the vehicle was stopped at street crossings, and the classifier could not differentiate between the end of three or more lanelets. The misclassifications in this stance do not represent a risk for the system integrity, since sudden “jumps” from one lanelet to another can be easily ruled out by the INS measurements.
In this section, for brevity’s sake, one particularly harsh section of the full experiment was selected to assess the method.
Horizontal component RMS of the techniques evaluated versus post-processed RTK
From Table 1 and Fig. 10, it is possible to see the already explored in literature effect of an external directional constraint on navigation. This paper explores the novel possibility of integrating an SVM classification on Wi-Fi RSS data to generate such directional constraints to be integrated in the filter. Future version of this study will explore other methods of integrating yaw measurements, and explore more fine system calibration techniques to improve the already promising results achieved.
- Angrisano A (2010) GNSS/INS integration methods. Ph.D. thesis, Universita’ Degli Studi Di NapoliGoogle Scholar
- Bender P, Ziegler J, Stiller C (2010) Lanelets: efficient map representation for autonomous driving. In: IEEE intelligent vehicles symposium, proceedings (Iv), pp 420–425. https://doi.org/10.1109/IVS.2014.6856487
- Falco G, Campo-Cossío M, Puras A (2013) MULTI-GNSS receivers/IMU system aimed at the design of a heading-constrained tightly-coupled algorithm. In: 2013 International conference on localization and GNSS, ICL-GNSS 2013, June. https://doi.org/10.1109/ICL-GNSS.2013.6577263
- Grejner-Brzezinska DA, Yi Y, Toth CK (2001) Bridging GPS gaps in urban canyons: the benefits of ZUPTs. Navigation 48(4):216–226. https://doi.org/10.1002/j.2161-4296.2001.tb00246.xGoogle Scholar
- Groves PD (2008) Principles of GNSS, inertial, and multisensor integrated navigation systems, vol. 2. Artech House, LondonGoogle Scholar
- Hsu Lt, Gu Y, Chen F, Wada Y, Kamijo S (2015) Assessment of QZSS L1-SAIF for 3D map-based pedestrian positioning method in an urban environment. In: proceedings of the 2015 international technical meeting of the institute of navigation, Dana Point, California, pp. 331–342Google Scholar
- Lu H, Zhang S, Dong Y, Lin X (2010) A Wi-Fi/GPS integrated system for urban vehicle positioning. In: IEEE conference on intelligent transportation systems, proceedings, ITSC, pp 1663–1668. https://doi.org/10.1109/ITSC.2010.5625268
- Rogers RM (2007) Applied mathematics in integrated navigation systems, 3rd edn. https://doi.org/10.2514/4.861598
- Steinwart, I., Christmann, A.: Support Vector Machines. Information Science and Statistics. Springer New York, New York, NY (2008). https://doi.org/10.1007/978-0-387-77242-4
Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.
The images or other third party material in this chapter are included in the chapter's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the chapter's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.