Here at OxTS we are acutely aware of the challenges facing ADAS and autonomous car developers. Systems and test scenarios are becoming more complex as vehicle trials are moving away from the sterile confines of the proving ground and out onto the open road. As a result, the collection of precise position data is becoming ever more difficult – but no less crucial.
Issues on the open road
Real-world urban test environments – as opposed to those simulated at the test track – present a number of obstacles to the collection of reliable position data with a GNSS-aided INS. Tall buildings, bridges, underpasses and urban canyons – streets hemmed in by tall buildings along each side – can interrupt or obscure satellite signals and cause multipath errors, making it hard to provide a robust, precise position solution.
OxTS’ gx/ix technology was conceived both to enhance position data in clear-sky environments and to provide more consistent, more accurate data in the challenging real-world scenarios that are now a part of ADAS and autonomous vehicle development. Gx/ix’s tightly coupled integration of GNSS and IMU data isn’t brand new, but it is, like all OxTS products, subject to continual development.
Initially released as a post-processing package but later updated with RTK functionality, gx/ix is actually two pieces of processing software working in tandem. The gx element functions when four or more satellites are in view. Instead of using the GNSS receiver to calculate position data, gx processing takes raw satellite signal information and calculates its own solutions, which are then combined with data from the IMU.
When fewer than four satellites are visible, ix processing takes over. Again, it uses raw signal data from whichever satellites are visible – even just one – but deals with each signal individually to provide a useful input to the navigation solution. Gx/ix switches seamlessly between modes to provide the most accurate data stream, either in real-time or during post-processing.
The streets of San Francisco
The effectiveness of gx/ix is clear to see in a number of datasets we have collected around the world. One such trial was conducted on an extended route around the San Francisco Bay Area of California, using an RT3003 GNSS/INS. The route took in a range of urban and highway-type roads across the Palo Alto, San Mateo, San Francisco and Oakland areas, and it included both the upper and lower sections of the Bay Bridge.
Analysis of the data showed that, in its standard differential corrected mode, the RT3003 output a corrected position (differential, float and integer combined) for 82.8% of the overall route, but this improved to 93.8% using gx/ix. Moreover, the RT3003 median position accuracy improved from 0.087 m without gx/ix to 0.043 m with it. Focusing specifically on the more challenging urban-only section of the route, including a two-way pass over San Francisco’s Bay Bridge with its covered lower deck, the RT’s output improved from 77.67% to 91.06% in corrected mode.
Now with added GLONASS
In keeping with the theme of continual development, earlier this year gx/ix was updated to support GLONASS satellite data. Having two satellite constellations available further enhances RTK lock and increases the chances of maintaining a precise centimeter position solution.
To test the effectiveness of updated firmware with GLONASS support, a further series of datasets has been collected, including one around the city of Turin. The dense urban environment of the Italian city is, in some ways, quite different from the more modern, high-rise layout of downtown San Francisco, but it is one which nonetheless presents its own challenges and is usefully representative of many European cities.
The Turin test data was collected using an RT3003 using GPS-only firmware and compared with that collected from an RT3003 using updated firmware featuring GLONASS support. Subsequent analysis of the datasets showed significant gains in the percentage of time spent in the most accurate position mode: gxInteger.
Using the GPS-only gx/ix solution over the full length of the test, the system was able to hold gx Integer position mode for 38.9% of the time. With GLONASS support added, that figure rose to 68.3% – a significant increase in the percentage of time spent in the most accurate position mode. These figures translate into a considerable improvement in horizontal position accuracy, from 40.7 cm with the GPS-only firmware to 21.8 cm using the latest gx/ix processing with GLONASS support.
Dense city and urban environments undoubtedly pose the biggest challenges to the growing number of developers of autonomous driving technology, both for the complex driving scenarios they present and for the data-collection headaches they pose. Our gx/ix processing continues to provide an accurate, consistent and cost-effective measuring capability under just such conditions. And as the industry’s demands continue to develop, so we will strive to offer the best possible position solutions.