Integrating Vision Derived Bearing Measurements with Differential GPS and UWB Ranges for Vehicle-to-vehicle Relative Navigation.

E.A. Abolfathi, K. O'Keefe

Abstract:	Vehicle positioning is an important component of intelligent transportation systems. Due to the relative low cost and low complexity of Global Navigation Satellite System (GNSS), the automotive industry has adopted this technology to provide vehicle position. In many current applications, this involves only absolute position, typically at the road level in open sky outdoor environments. However, in various cases, position solutions are needed in urban canyons or under dense foliage where there is poor GNSS availability. Even in some circumstances where a GNSS solution is available, the solution may not be accurate enough for some applications. In these cases, other sensors can be used in order to improve the accuracy of GNSS positioning. In many applications, like vehicle navigation systems that are studied in this paper, the absolute positioning solution is not the main interest. Instead, relative positions are required. For example, cooperative driving on highways or safety related applications such as blind spot warn or collision avoidance. In these cases additional sensors can be integrated with GNSS to increase both availability and accuracy of the relative position solution. Recently, peer-to-peer (P2P) cooperative positioning has been proposed where peers obtain pseudorange measurements from satellites through their GNSS receiver and use their wireless interface both to communicate with each other and obtain terrestrial range measurements to other peers. In this method, by fussing these measurements, relative position can be estimated accurately even when insufficient GNSS observations are available. Results have shown that P2P cooperative positioning can outperform traditional GNSS-only positioning in terms of both accuracy and availability. In this paper, a similar P2P network using both differential GNSS and direct ranging is demonstrated but with the addition of adding bearing measurements to the system. In cases of three or more vehicles in order to determine a horizontal relative positioning solution in the absence of GNSS measurements, vehicle to vehicle range measurement and bearing measurement are both required. In this work, Ultra-Wideband (UWB) ranging transceivers are used to provide ranges between the vehicles. With range observations only, the orientations of the vehicles are still ambiguous. In previous studies the benefit of adding bearing measurements was shown to improve the relative positioning error especially in the across-track directs. However, these results were based on simulations. In this paper, vision sensors are used in order to extract real bearing measurements. Vision-based navigation has received increasingly more attention especially when it comes to intelligent navigation. Since cameras are relatively low cost, low power and capable of providing textual information, there is a preference to use them instead of other expensive sensors. Many research projects have been conducted regarding vision based navigation in several fields like unmanned aerial vehicles, indoor positioning and robot control. The fundamental prerequisite in this step is to determine which features should be extracted and tracked in order to have accurate and available navigation information. Research in extracting good-and-easy to track features for vehicle navigation systems is currently lacking. In this paper , identifying features with image processing and tracking them with Kalman Filter based video processing are tested on real data to determine which features should be extracted and which methods should be used in order to track these features. Since the goal is to determine the bearing from one vehicle to another, candidate features include license plates, wheels and front and back windows. The errors in determining inter-vehicle bearing produced by using feature identification and feature tracking are compared to each other. Some existing methods involve using two cameras to determine a bearing measurement. Although more features can be extracted by using two cameras, the additional cost, complexity and the problem of managing asynchronous measurements obtained from each camera, have motivated the use of a single camera in this work. A data set has been collected for a scenario involving three vehicles under different conditions in a test trajectory in Calgary. Each vehicle is equipped with GPS receivers and UWB ranging radios while one of the cars is additionally equipped with a high definition video camera. First, the identification and tracking methods are evaluated in order to extract bearing measurements of the other two vehicles in the body frame of the camera-equipped vehicle. The estimated bearing measurements are then compared to those obtained from carrier-phase GPS/INS reference trajectories of the three vehicles in order to evaluate bearing measurement errors. In next step, these bearing measurements are integrated with the differential GPS and UWB range data. All these data are then processed and the performances of three different methods in terms of positioning error, and solution availability are compared to each other: GPS-only, GPS + bearing measurement and GPS + bearing measurement + UWB range. The bearing measurements obtained from the reference trajectory are also applied to draw a comparison between expected and real improvements resulting from adding bearing measurements. Results show that using bearing measurements can improve the positioning solution availability, especially when the GPS-only solution is not accurate enough in some environments such as under dense foliage and in deep urban canyons.
Published in:	Proceedings of the 26th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2013) September 16 - 20, 2013 Nashville Convention Center, Nashville, Tennessee Nashville, TN
Pages:	762 - 771
Cite this article:	Abolfathi, E.A., O'Keefe, K., "Integrating Vision Derived Bearing Measurements with Differential GPS and UWB Ranges for Vehicle-to-vehicle Relative Navigation.," Proceedings of the 26th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS+ 2013), Nashville, TN, September 2013, pp. 762-771.
Full Paper:	ION Members/Non-Members: 1 Download Credit Sign In