In Pursuit Of Precision

Satellites have brought great benefits to safety and navigation, but there are limits. Robert Gough considers how existing technology can be harnessed to take mobility to the next level

Legislation published in October 2019 by the EU requires all newly registered vehicles in categories M, N and O to incorporate a host of advanced driver assistance features. This means that, from this year, any vehicle with four or more wheels used to transport passengers or goods will need to have an autonomous emergency braking (AEB) system, intelligent speed assist, lane keep assist and vulnerable road user (VRU) detection, among many other features.

That such a wide range of advanced driver assistance systems (ADAS) features are now mandatory for type approval only eight years after Euro NCAP started testing AEB car-to-car, shows how rapid the pace of ADAS test development has become.

Whereas historically, dynamic measurements such as acceleration, slip angle, wheel speed or attitude were key to vehicle development, accurate location is increasingly becoming the key measurement due to this rapid growth in ADAS development.

Twenty-two years ago, selective availability of GPS was discontinued. Selective availability was a deliberate degrading of the civilian GPS signals to only allow accuracies up to 100m. Turning this system off suddenly allowed cheap commercial receivers to achieve accuracies of a few metres. This made possible an entirely new industry of satellite navigation with the market for in-car satnavs expanding rapidly throughout the early 2000s.

More advanced techniques such as real-time kinematic (RTK) allowed centimetre-level position accuracies and again, it is no coincidence that, a couple of decades on from the year 2000, systems such as lane departure warning and forward collision warning became increasingly common. This is because the development and testing of these systems required accurate position of both the vehicle under test (VUT) and the target to be known extremely precisely.

In so-called open sky conditions on a test track or proving ground, GNSS (global navigation satellite systems, the umbrella term for systems such as GPS, GLONASS etc) is able to provide all of the position data that the development, test and validation of these systems require. However, the increasing requirements to verify the robustness of ADAS systems necessitates testing in different lighting and weather conditions, as well as on the open road, in tunnels and under cover etc.

When testing on the open road, dense tree cover or urban canyons caused by tall buildings can reduce satellite coverage. Worse, reflected satellite signals can trick a GNSS receiver into calculating an incorrect position that may be hundreds of metres out from reality. These issues can be mitigated by tight coupling of an inertial measurement unit (IMU) to GNSS data. This allows poor satellite data to be filtered out, while even data from single satellites can be used to reduce IMU drift and speed-up RTK reacquisition after brief outages.

However, even tight coupling still ultimately relies on the availability of satellite coverage, making it clear that any system able to truly navigate anywhere needs to integrate with other position-aiding sources in addition to GNSS.

As an example, park assist has gone from a niche and expensive optional feature to a commonplace system available on many mainstream cars. However, it must be possible to test these systems in the environment in which they will be used – and given the ubiquity of covered and underground car parks, this necessarily means testing in conditions with very poor or even zero satellite coverage. As automated valet parking systems develop whereby vehicles will drive themselves a short distance to/from a parking space, the need to verify safety is paramount. This necessarily requires knowing the precise location of the VUT, as well as any VRU or vehicle objects used in the test scenario, without access to GNSS satellites.

OEMs and test houses that have dedicated facilities for testing these systems can now use ultra-wide band (UWB) technology to calculate position to within a few centimetres with no satellites when used in conjunction with an inertial navigation system (INS). This offers a very cost-effective method of calculating position without the need for visible infrastructure that could bias the testing. As far as the vehicle sensors are concerned, a facility with a UWB system installed would look no different to any other car park or facility.

Testing indoors with UWB opens up many other options, allowing other ADAS systems to now be tested in indoor environments with controlled light, temperature and weather, etc. UWB scales very well and can be used for both small, low-speed applications and much larger, higher-speed environments.

However, even here, the requirement to have permanently installed infrastructure is a limitation that means users are not truly navigating ‘anywhere’. The ultimate goal of a system that can navigate anywhere is to do so without any external infrastructure.

One way to achieve this is to use LIDAR. This is a very powerful surveying tool and can quickly build up a very precise point cloud that can be used to generate centimetre-accurate maps of structures, objects, vehicles and lane lines, etc. This ability can also be used in a technique called SLAM (simultaneous localisation and mapping). Here, the point cloud currently being generated by LIDAR is simultaneously used to localise against the surroundings being mapped.

All this means that with a LIDAR system linked to an INS, an accurate position can be maintained without the need for satellites or other external infrastructure. Such a system could truly be said to be able to navigate ‘anywhere’.

Robert Gough is a product engineer at OxTS