Future of autonomous driving

GNSS, cybersecurity, and software-defined vehicles: A conversation with Automotive World

We recently discussed the technological future of autonomous driving in a panel discussion with Automotive World. In this article, we detail the top three themes covered.

Your car and your mobile phone are starting to have a lot in common. We’re witnessing the last generations of mechanical vehicles driven by internal combustion engines, and the first generations of electric vehicles whose functions are increasingly handled by software.

Boston Consulting Group sees software-defined vehicles (SDVs) as a $650bn market by 2030, as space freed up by electrification is given over to powerful in-vehicle computing platforms. The shift to SDVs is a necessary part of the transition to hands-free driving, but there are many hurdles still to be overcome before L2+ driving is approved for smaller roads and urban canyons.

To discuss those hurdles—and how to tackle them—Manuel del Castillo, VP, FocalPoint took part in a recent webinar hosted by Automotive World on the Mobex platform with James Tidd, VP of Systems Engineering at Swift Navigation, and Boubeker Belabbas, navigation expert at Bosch.

The panellists discussed three hot topics driving the industry forwards: extending hands-free driving to more roads; emerging cybersecurity and privacy concerns, and the continued role of GNSS in ADAS and autonomous vehicle systems.

1. Extending hands-free driving onto more roads

Autonomous driving is fast becoming a reality, with numerous vehicles approved for hands-free driving on certain highways, and even “hands-free, eyes-off” driving starting to get the green light from regulators.

But extending approvals to secondary and tertiary roads is a different matter, and dense urban environments present particular challenges. The vehicle’s ability to ‘see’ and navigate in complex conditions will rely on batteries of sensors of different types, with high levels of redundancy to compensate for different sensors’ failure modes.

James Tidd offered the example of driving on a secondary road in the snow. Here, camera vision on its own may not be able to make sense of its surroundings, impeding the vehicle’s ability to detect and navigate obstacles and position itself accurately on a digital map.

Camera vision therefore needs to be complemented by other sources of positioning data, such as ‘dead reckoning’ inertial measurement units (IMUs) and global navigation satellite systems (GNSS) signals. Yet these also have failure modes: IMUs can only measure accurately for short distances before starting to drift, and GNSS signals are susceptible to radio frequency (RF) interference and can be blocked or reflected by obstacles in the environment.

The more sensors in the vehicle, the more they can pick up the slack from each other and enable the vehicle to drive safely in complex environments. But at the same time, more sensors means higher manufacturing costs, potentially making hands-free prohibitive for the mass market. So trade-offs must be found that optimise safety, cost, aesthetics and the user experience.

One solution will be to get more use out of lower-cost sensors. GNSS is a case in point, as it’s relatively low cost, and technologies are fast emerging that address its failure modes. FocalPoint’s own positioning solution, S-GNSS Auto with Supercorrelation, for example, is a software-based enhancement to a GNSS chipset that can significantly boost the accuracy and reliability of the vehicle’s GNSS receiver.

2. Cybersecurity and data privacy considerations

As vehicles become more software-defined, they introduce additional security and privacy considerations. Boubeker Belabbas highlighted the need to address the SDV’s reliance on AI as part of this evolution.

Solutions like AI edge computing and anonymization can confine sensitive data within the vehicle, ensuring privacy before sharing data with cloud-based AI. Similarly, sensor fusion algorithms can use edge-based encryption to protect location data, enhancing both security and user trust.

As machine learning and AI advance, the individual steps taken by these algorithms become harder to discern, and the vehicle’s reasoning becomes opaque. Regulators, manufacturers and standards bodies must collaborate to ensure that AI use is transparent and safe, and that standards keep pace with technological developments and new automotive manufacturing processes.

Another concern relates to data privacy. Real-time mapping and personalised navigation services will rely on vehicles sharing their view of their surroundings by streaming sensor data to the cloud. Vehicle users will need to consent to this use, which may be a challenge. The industry will have to work hard to stress the security and anonymity of the shared data, and its benefits to vehicle users in terms of making journeys easier and more enjoyable.

Sensor vulnerabilities are a third concern. While redundant sensors will enable the vehicle to navigate even if one or more sensors fail, improving the robustness and resilience of individual sensors must also be a key goal.

Manuel offered the example of GNSS spoofing; a type of cyber-attack that involves broadcasting fake GNSS signals to throw a vehicle off course. Spoofing can be combated with software (like ours) that detects and rejects spoofed signals, and it can be part of a regular SW update.

As this is just one example of many, the ability for SDVs to receive continual over the air (OTA) software updates will be critical. Precise positioning relies on the vehicle having line-of-sight to GNSS satellites, which can be a challenge in built-up areas where signals reflect off buildings. However, a software solution like S-GNSS® Auto can filter out these ‘multipathed’ signals and focus only on line of sight signals. Together, such solutions mean GNSS can be used as a reliable, accurate source of position data even in dense urban areas—an essential step on the road to ‘anywhere, anytime’ hands-free driving.

3. The continued importance of GNSS

A final theme of the discussion related to the continued importance of GNSS as a sensor in the software-defined and increasingly automated vehicle. As Boubeker emphasised, GNSS stands out as the only sensor capable of providing absolute localisation, enabling the vehicle to determine its precise position on the Earth's surface efficiently and reliably. Without it, a vehicle can only know where it is by observing its surroundings and matching them to a digital map that’s updated in real time.

While it is technically possible to navigate purely using this method, known as simultaneous localisation and mapping (SLAM), it will likely require a continuous low-latency connection to a digital map held in the cloud. It also requires hefty onboard compute power to apply the necessary algorithms to camera and LiDAR data to place the vehicle accurately on the map. This is an extremely costly solution compared to GNSS, a low-cost and near-ubiquitous source of absolute position data.

A far more economical solution is to address the limitations of GNSS so that it can be used reliably everywhere. One of its traditional limitations is positioning accuracy, but this can be addressed with technology that correct errors in the GNSS signal, like Swift Navigation’s Skylark® Precise Positioning Service, which delivers centimetre-level accuracy.

Precise positioning relies on the vehicle having line-of-sight to GNSS satellites, which can be a challenge in built-up areas where signals reflect off buildings. However, a software solution like S-GNSS Auto can filter out these ‘multipathed’ signals and focus only on line of sight signals. Together, such solutions mean GNSS can be used as a reliable, accurate source of position data even in dense urban areas—an essential step on the road to ‘anywhere, anytime’ hands-free driving.

Watch the webinar for more insights on the future of autonomous driving

As the automotive industry moves towards electrified, software-defined and increasingly-autonomous vehicles, it’s imperative that OEMs, Tier 1s, suppliers, regulators and standards bodies work together to ensure a safe, enjoyable driving experience.

For more insights from the three panellists on what needs to happen next, watch the pre-recorded webinar: The changing vehicle mapping and navigation landscape.