Experts at u-blox discuss how they’re creating a hybrid positioning system for automated vehicles using GNSS and terrestrial radio ranging
By David Bartlett, senior principal engineer, Product Center Positioning, and
Stefania Sesia, head of Application Marketing, Automotive, u-blox
There’s so much discussion around automated vehicles in the mainstream press these days, that it’s easy to forget some of the critical enabling technology needs to mature significantly before large numbers of people are being whisked from A to B by completely driverless cars.
An area demanding particular attention is high-precision positioning. The Society of Automotive Engineers published a six-level automation scale. For vehicles at the higher end of the scale to become reality, they need to be able to reliably pinpoint their location to within centimeters, at all times.
Society of Automotive Engineers’ six-level automation scale. (Image: SAE International)
The positioning systems in most modern cars — which typically use GNSS receivers coupled with an inertial measurement unit (IMU) and the odometer — can’t get close to this level of accuracy. Even in the most favorable conditions for GNSS satellite signal reception, accuracy is between 2 and 5 meters horizontal circular error probable (CEP) without a correction service. In more challenging environments, such as urban areas or indoors, this is significantly reduced.
Using UWB and V2X to complement GNSS
Various solutions are being developed to address this GNSS shortcoming, but all currently have their limitations or don’t offer a solution that’s workable in all environments. Future autonomous vehicles will therefore invariably need to rely on hybrid solutions that blend multiple technologies.
One area where relatively little research has been done to date is in combining GNSS with terrestrial radio signals to enhance automotive positioning accuracy. Cellular vehicle-to-everything (C-V2X), IEEE 802.11p V2X, its successor 802.11bd and ultra-wideband (UWB) can all be used for short-range distance measurements. V2X ITS communications technology is listed as a potential positioning solution in EN 302890 (Intelligent Transport Systems), while UWB technology is gaining momentum for indoor applications, as well as by vehicle manufacturers for keyless entry.
These technologies are all ripe for further investigation as complements to GNSS and IMUs, to ultimately support higher levels of vehicle autonomy. U-blox recently ran a study to evaluate the terrestrial-ranging strengths and weaknesses of IEEE 802.11p V2X and UWB as part of a hybrid solution with GNSS for automotive navigation. Our aim was to establish their feasibility for this application, and identify where further research needs to happen for this type of hybrid navigation solution to become part of future autonomous vehicles.
Photo: jonathange/iStock/Getty Images Plus/Getty Images
How terrestrial ranging works
A terrestrial-ranging system requires a network of fixed ground stations (typically referred to as roadside units, or RSUs, in V2X systems) at known locations. V2X or UWB signals sent out by the vehicle are returned by the RSUs, enabling the vehicle to measure the roundtrip time, and consequently calculate the distance between itself and the anchor point. Do this for three or more RSUs that are geometrically dispersed relative to the vehicle, and you can determine its position.
The need to simulate
Mass deployment of the RSUs required for this type of solution has not yet happened. Installing a suitable network of ground stations in an urban setting on public land wasn’t feasible for our research, in part because the regulatory landscape around UWB in this context is still evolving.
Instead, we set up anchor points around various private estates, from open fields to areas representative of urban environments, such as a business park. We took extensive measurements of the UWB and V2X signals’ behavior in these environments, which enabled us to extract performance statistics such as noise, and subsequently create a behavioral simulation model for the ranging performance.
Our test methodology
Having established our behavioral simulation model for different types of environments, rural, urban and indoor settings, we did a number of real-world test drives. These covered a wide range of driving conditions. We took in high-speed sections of open road, dense urban areas, start-stop congested traffic, numerous corners, and places with limited or no GNSS reception such as tunnels.
During these drives, we collected both GNSS measurements and ground truth. For the former, we used a u-blox NEO-M8L module with built-in IMU. To establish the ground truth, we used a high-grade real-time kinematic (RTK) receiver, GNSS augmentation data service and a high-spec IMU.
We classified each section of the test drives based on the environment — dense urban, tunnel, open countryside and so on — to enable us to apply the appropriate noise models in our simulation.
Next, we allocated RSU positions based on chosen density and placement rules, and added 2 m of random height variation, to ensure we avoided a fully planar deployment. We tested with various numbers of RSUs, to help understand how many would be required to achieve the necessary levels of location precision.
We then set additional simulator variables, such as the accuracy of the timestamp on the ranging measurements.
Having done all of this, we generated simulated ranging measurements between the RSUs and the truth position for every ranging epoch. To these, we added noise on a sample-by-sample basis, and merged the resulting noisy simulator measurements with the GNSS measurements we recorded en route.
Key findings
The output of the simulator enabled us to generate performance statistics that facilitated a comparison between the hybrid GNSS + V2X and GNSS + UWB solutions and a conventional GNSS + IMU solution, similar to those found in mainstream vehicles today.
The table below shows performance of the three solutions.
|
UWB |
V2X (IEEE 802.11p) |
GNSS+IMU |
| Ranging update rate |
0.67 Hz
(1.5 s interval) |
10 Hz (0.1 s interval) |
n/a |
| Horizontal accuracy |
0.1 – 2.5 m (Hybrid) |
1.1 – 4.2 m (Hybrid) |
1.2 – 5.5 m |
| Height accuracy |
0.4 – 5 m (Hybrid) |
5 – 10 m (Hybrid) |
2 – 7 m |
| Frequency of operation |
6.5 GHz |
5.9 GHz |
n/a |
| Signal bandwidth |
500 MHz |
10 MHz |
n/a |
Performance of the three navigation solutions on test.
At a very high level, we found that the GNSS+V2X (IEEE 802.11p) system achieved performance similar to a conventional GNSS+IMU(DR) solution using standard positioning. In situations where there is no GNSS reception, or where this is seriously degraded, an IMU also loses its value, given its reliance on continual GNSS reception to remain aligned. Here, a V2X-based positioning solution would be of value for navigation guidance.
However, more work will need to be done, including into the role of the IMU in high-integrity, high-accuracy positioning, to achieve the levels of accuracy and integrity that autonomous applications require.
The GNSS + UWB hybrid system delivered significantly better performance, approaching the levels that can be achieved using an RTK-based GNSS augmentation service. Our test system ran at 0.67 Hz, and was able to deliver precision close to 10 cm, though we would expect future production systems to align with the more common 10-Hz refresh rate broadly used in V2X.
By pairing a 10-Hz UWB ranging system with a high-accuracy GNSS system using correction data, it should be possible to achieve 10 cm-level accuracy in most situations. GNSS with correction data is already proven to be capable of delivering this level of precision in open areas and motorways. A network of RSUs deployed in urban environments would enable UWB to complement high-accuracy GNSS in situations where satellite reception is challenging.
However, the limited range of UWB, coupled with current regulatory restrictions around outdoor use, limit its usefulness at the present time. That said, micro-navigation in indoor areas, such as parking garages, could be a good fit for this technology.
Other lessons learned
The research brought to light a number of other important findings. First, having even just two RSUs visible, in addition to GNSS, provided significant benefit in the hybrid solution.
Second, height variation in the RSUs is essential if the navigation system is to determine the vehicle’s height accurately, particularly with V2X technology. This will be particularly important when it comes to enabling vehicles to safely operate where there are different levels of road one above the other, such as at multi-level junctions.
Third, we were successfully able to build a hybrid filter to process the signals from the V2X, UWB and GNSS systems, and seamlessly handle the transition between areas with GNSS only (where there were no RSUs deployed) and terrestrial ranging only (such as tunnels).
Fourth, despite the promise it showed for this application, terrestrial ranging is far from immune to environmental effects and multipath. Even UWB would sometimes suffer from non-line-of-sight signal propagation.
Finally, accurate time alignment between the GNSS and terrestrial ranging measurements also emerged as a critical factor. Where we had initially anticipated that alignment to within a few milliseconds would be sufficient, in reality we found we needed to be below 100 microseconds.
What next?
This research has shown the potential of using terrestrial-radio ranging to complement the existing positioning technologies and services being deployed in vehicles today. That said, more needs to happen, not least on the regulatory front, for this technology to genuinely become one of the enablers of future autonomous vehicles.
Outdoor UWB use needs to be permitted for this application, for example, and there needs to be widespread deployment of UWB-capable RSUs. Moreover, when RSUs of any kind are being deployed, thought needs to be given to their possible use as positioning anchors, rather than simply as communication devices.
In addition, more spectrum and wider channels need to be allocated to V2X. And we need to see positioning primitives and signals incorporated into the V2X standards. (Positioning primitives allow a car to know in what direction it is headed — up/down/left/right — relative to a point of reference. It uses signals from the sensors to calculate these values.)
A related area that merits further investigation is the use of UWB ranging to protect vulnerable road users such as people walking, wheeling and cycling. With modern smartphones and cars both now including UWB technology, there are opportunities to use this to make autonomous vehicles more aware of the position of people in their surroundings.
If you’d like to find out more about the research, our methodology, or the results, we’d be delighted to discuss these with you. Please email David.Bartlett@u-blox.com to get in touch.
Gilla detta:
Gilla Laddar in …
