Improving Vehicle Localization with Two Low-Cost GPS Receivers

Abstract

A primary concern of intelligent traffic management systems (ITMS) is to collect necessary traffic data. Vehicle position is one of the most important data types to manage traffic effectively. Most current approaches to localize modern vehicles (MVs) fall into three categories. The first category uses standalone reference stations, such as the wide-area augmentation system (WAAS), which are expensive modules. The second category uses multiple expensive localization sensors such as the global positioning system (GPS), global navigation satellite system (GNSS), and inertial measurement unit (IMU). However, such expensive solutions may not be applicable in all vehicles, impacting generalizability. The third category is a software-based approach. As opposed to the abovementioned approaches using expensive hardware, the third category uses software, such as map-matching techniques, to augment noisy localization sensors. In this study, we investigated map-matching software in some case studies and found that it cannot locate the vehicle effectively if the positional data are collected by a low-cost and too noisy GPS receiver. Therefore, this paper analyzes and highlights the impact of GPS receiver's noise in applying self-localization. It also proposes a new methodology by integrating cross-GPS validation, interpolation/best fit, and map-matching techniques to localize a vehicle in the presence of GPS signal noise and investigate it in real traffic data from a metropolitan area. Our proposed methodology is able to identify the more accurate GPS receiver dynamically, by considering the fixed distance between the two GPS receivers. Our evaluations indicate that the proposed methodology can significantly improve vehicle self-localization performance.

Geolocation estimation of target vehicles using image processing and geometric computation

“…Collecting the geolocations of the ego vehicle accurately plays a vital role in accurately estimating the target vehicle's geolocations. Therefore, to enhance the accuracy of the ego-vehicle localization in the GPS coordinate system, the proposed approach in [20], based on cross-GPS validation, interpolation, best-fit, and map-matching techniques, is used.…”

“…One of the cameras mounted on the front window glass looked forward through the window, and another mounted on the back window glass looked backward. The purpose of mounting two cameras on the same vehicle was to improve the ego vehicle's self-location in the pre-processing step [20], and to collect more data for future studies. To estimate the target's geolocation, this study used only the footage collected from the camera mounted on the front window glass.…”

“…Therefore, the main objective of this study is to go beyond the target vehicle localization and estimate the relative angles besides estimating distance and ego vehicle's geolocation to be able to estimate the geolocation of the target vehicle dynamically. To estimate the ego vehicle's geolocation based on a lowcost GPS receiver, we used the proposed approach in [20]. To estimate distance and relative angle, we proposed two new approaches based on object detection and image processing and geometric computation.…”