Abstract
Global navigation satellite systems have revolutionized the way people are navigating vehicles in unknown environments, and improved the driving experience in complex traffic areas. Satellite based navigation systems however experience their limitations in densely developed areas with high buildings, as satellite signals get attenuated, blocked and reflected by objects in the line of sight between the user and the satellite. Unfortunately, areas with degraded satellite conditions correlate well with areas of more complex navigation requirements, leading to less reliable navigation capabilities where it is needed most.
Current in-dash navigation systems integrate additional sensors such as inertial and wheel sensors, with the drawback of vehicle-specific, inaccurate and often biased measurements. With the recent introduction of advanced driver assistance systems, optical systems become attractive additional sensors. Although cost-effective visible light cameras can provide useful measurements, their usage in combination with satellite navigation systems has not attracted much interest up to this date.
This research reached the goal of accurate navigation in urban environments by deeply integrating vision sensors with satellite navigation systems, while intentionally omitting inertial sensors. The proposed system was tailored specifically for road vehicles, where the test vehicle was post-equipped with a stereo camera and a custom multi-constellation satellite navigation receiver. The error characteristics of the visual sensor system were simulated in software and tested in a laboratory environment. Furthermore, extensive testing in challenging environments was performed, which shows the advantages and limitations of the proposed system in real environments.
The navigation performance of different GNSS only, GNSS/INS and GNSS/vision implementations was compared. The GNSS/vision system provided a continuously available navigation solution, and despite the fact that the used IMU was of cost-intensive tactical grade, the integrated GNSS/vision system consistently outperformed all other systems.
