This content provided to DVN by Focal Point
(Manuel Del Castillo, Business Development VP at GNSS software company Focal Point Positioning, says the future of autonomous driving depends on trustworthy positioning.)
Waymo is currently testing driverless vehicle technology on the streets of London, in preparation for a commercial launch there. Among the sensing technologies that make driverless vehicles possible is global satellite navigation systems (GNSS). It’s widely used in a range of applications, but its role in autonomous driving is often overlooked. While positioning errors in other applications may be relatively unproblematic, GNSS is a safety-critical capability for trustworthy autonomous driving.
The ‘confident but wrong’ problem
GNSS provide an absolute source of location. A GNSS receiver typically calculates the vehicle’s position and assesses how accurate that position is likely to be. This output is called ‘estimated accuracy’.
In a development or test environment, engineers can compare the GNSS position against a ground truth system and measure the actual error after the fact. A vehicle operating in the real world does not have that luxury; it must make decisions in real time, using the information available at that moment.
The estimated accuracy output thus becomes central to how the vehicle treats GNSS. If the system believes the GNSS position is highly reliable, that information may be used to support map alignment, lane-level localization, or cross-checking against other sensors. But if the GNSS output is judged to be degraded, the vehicle can reduce the weight it gives to that input, or reject it entirely.
The risk comes when the estimated accuracy understates the actual error. In other words, the receiver believes it has a good position estimate, but the real error is much larger than expected. This is the ‘confident but wrong’ problem.
For automated driving, this can be more dangerous than an obvious GNSS failure. A faulty output that appears healthy can be accepted by the wider localisation stack and used as if it were trustworthy.
Cities are a challenge for reliable positioning
Traditional GNSS performance is strongest in open-sky environments. Satellite signals reach the receiver along a clear line-of-sight path, and the resulting position is typically accurate to within a few meters.
Urban and semi-urban environments are far more difficult. Buildings, roadside infrastructure, tree cover and large vehicles can obstruct or reflect satellite signals before they reach the antenna. Instead of receiving only direct line-of-sight signals, the receiver may also track signals that have bounced off nearby surfaces. Because the receiver assumes a direct path when calculating position, these reflected signals introduce ranging errors – known as multipath and non-line-of-sight errors.
The receiver may still appear to be in good operating condition. It may detect several satellites, and the satellite geometry may look favourable. On the surface, the system looks healthy, so the receiver reports a high level of confidence in the calculated position. However, some of the measurements feeding that calculation may already be corrupted by reflections.
This matters because the localization system is not only trying to identify a vehicle’s approximate position. For ADAS and AD, the vehicle needs to know which lane it is in, whether it remains inside the correct operational design domain (ODD), and how its position corresponds to a high-definition map.
The role of GNSS in ADAS
Autonomous vehicles use overlapping sources of information because every sensing technology has limitations. Cameras, radar, lidar, inertial sensors, and GNSS each contribute something different to the vehicle’s ‘understanding’ of the road environment. GNSS provides a scalable source of absolute positioning. Other sensors describe the world around the vehicle, but GNSS helps anchor the vehicle to global coordinates and the digital map. It also provides an independent input that can be used to calibrate or cross-check the wider localisation system.
That input is valuable, if the GNSS output is reliable. If reflected signals cause the receiver to produce an incorrect position with an optimistic confidence estimate, the ADAS may give that input more weight than it deserves.
This can create several operational risks. Map matching can be affected if the vehicle is placed in the wrong lane or aligned with the wrong part of the road geometry. The vehicle may also struggle to reconcile the GNSS output with relative sensor data. In some cases, the system may not receive an obvious warning that the GNSS input has been degraded.
That is why estimated accuracy is part of the safety case for using GNSS in advanced autonomy. A trustworthy receiver needs to support the vehicle in deciding when GNSS should be trusted, when it should be de-weighted and when it should be rejected.
L2+ and L3 raise the stakes
The confidence problem becomes more important as automated driving functions take on more of the driving task. In L2 and L2+ driving, the human driver remains responsible for supervising the system. In higher levels of automation, such as L3, the system assumes responsibility for the driving task under specific conditions. That places a bigger burden on the localisation stack.
This is especially important as manufacturers seek to expand automated driving availability beyond open-sky environments. Wider ODDs require systems that can handle the complexity of real-world environments, including the positioning challenges created by dense urban streets, partial sky visibility and changing roadside conditions.
Reliable GNSS can therefore support autonomy in two ways: It can improve the accuracy of the position estimate itself, and it can help the system understand the reliability of that estimate in real time.
Fixing the problem earlier in the chain
One way to respond to poor GNSS performance is to try to correct the position after it has already been calculated. However, by that point, reflected signals may already have influenced the underlying measurement set. The receiver has processed contaminated inputs and the navigation engine has generated a position from them.
Focal Point Positioning’s approach is different. S-GNSS® Auto is designed to operate at the measurement level, before those measurements are passed into the navigation engine. Rather than attempting to repair a bad position after calculation, it aims to prevent corrupted measurements from distorting the calculation in the first place.
This simple firmware upgrade uses Focal Point Positioning’s patented Supercorrelation® technology to create a synthetic directional antenna in software. This allows the receiver to distinguish more effectively between direct line-of-sight satellite signals and reflected non-line-of-sight signals.
Once those reflected signals are identified, they can be suppressed. Cleaner GNSS measurements are then passed to the navigation engine, whether that engine is running on-chip or externally.
This is important because the navigation engine is only as good as the measurements it receives. By improving the measurement set before the position is calculated, S-GNSS Auto provides a more honest, reliable output.
By adding this capability to the ADAS you reduce false alarms, where unnecessary disengagements are triggered. This helps improve the driver experience and customer satisfaction. More importantly from a safety point of view, adding S-GNSS Auto means fewer missed detections. If a disengagement is not triggered when necessary, it is a safety of life issue. More trustworthy GNSS mitigates this risk.
Safety naturally remains central to the deployment of autonomous vehicles. Regulators have noted when outlining the regulatory framework for self-driving technology, that safeguards must include protection from hacking and cyberthreats. GNSS signals, like any radio-based system, can theoretically be spoofed by transmitting counterfeit signals, which is why techniques that strengthen signal integrity and authentication are increasingly important.
By reinforcing the authentic line-of-sight signal and suppressing delayed or inauthentic signals, Focal Point’s Supercorrelation enhances both the reliability and integrity of GNSS under hostile RF conditions. This provides an additional layer of resilience against spoofing and meaconing attacks, without the need for modifying hardware, making it suitable for deployment in mass-market platforms.
A software-defined route to deployment
For OEMs and tier-1s, any improvement to localisation also has to be practical to integrate. Automotive programmes are shaped by platform decisions and hardware constraints. A solution that demands a major architecture change may be difficult to deploy, even if the underlying performance benefit is attractive.
This is why a software-defined approach is important. S-GNSS Auto is designed to improve urban GNSS performance without requiring a complete redesign of the vehicle’s positioning architecture. It can provide enhanced measurements to the navigation engine and help existing systems make better use of GNSS in challenging environments.
S-GNSS Auto has been integrated onto STMicroelectronics’ Teseo devices, bringing signal-level GNSS intelligence to automotive-grade platforms. For vehicle manufacturers and tier-1 suppliers, this creates a pathway to improved GNSS reliability through software and firmware integration on qualified silicon.
That deployability is important for the next phase of automated driving. OEMs need technologies that can improve safety and availability while fitting within realistic production constraints. A GNSS enhancement that can be integrated into automotive-grade silicon gives engineers a practical route to more reliable positioning performance in the environments where it is most needed.
Trustworthy positioning for wider autonomy
Advanced autonomy cannot scale safely if vehicles are unable to trust their own location data. GNSS remains indispensable because it provides absolute positioning at scale, but its role in automated driving depends on reliability as much as accuracy. The confident but wrong problem shows why this distinction matters.
As ADAS and automated driving functions expand into more complex operational domains, localisation systems will need to become better at recognising uncertainty. They will need to know when GNSS is trustworthy, when it is degraded and when the wider vehicle stack should rely more heavily on other information.
Solving the confident but wrong problem and providing trustworthy GNSS positioning will be key to the future of autonomous driving. By addressing reflected signals at the measurement level, Focal Point Positioning’s S-GNSS Auto is designed to make GNSS more honest, more reliable and better suited to the demands of automated driving.
For further information, contact Ramya Sriram by email or write to
Focal Point Positioning Ltd
1-3 Chesterton Mill,
French’s Road,
Cambridge, CB4 3NP
United Kingdom
