Governments around the world are setting ambitious targets to increase the level of autonomy available to drivers. The UK signalled these legal changes in the King’s Speech in 2023, with Transport Secretary Mark Harper even saying he expects drivers to be able to travel without having to watch where they’re going by the end of 2026. Meanwhile, the European Union in July 2022 finalised a legal framework for fully automated vehicles with autonomous driving functions to address the challenges and opportunities these technologies present.
In order for OEMs to safely offer greater levels of autonomy, they must invest in the right technology and systems that can process massive amounts of data. Humans learn to drive with many hours of practice, and so machines too will have to drive in all situations and environments to ensure performance, safety, and ultimately, the public’s trust in self-driving cars.
The autonomy scale
Autonomous implementation is on a scale—not an absolute. While the industry currently operates between Level 2 and L3, it’s in most cases limited to certain environments, such as on motorways. For instance, today’s autonomous capabilities may struggle to account for unplanned interferences like jaywalkers and cyclists.
Moving to L4 is the step that requires more extensive changes due to the increased number of environments to which the vehicles must adapt. Here, cities and roads would need to be smarter, some infrastructure re-designed, legal responsibilities totally revamped, communication made constant between cars (and potentially between cars and their surroundings), and 5G+ rolled out extensively—as well as many changes to the highway code. This would require extensive cooperation and collaboration between all levels of public authorities, car manufacturers, software vendors and road users.
Technology needs time and space to learn
Accurate autonomy starts with GNSS. Signals from several constellations of satellites combine to give a precise triangulated position, with the atomic clock on a fourth satellite constellation providing exact timekeeping for the sensors and devices in the vehicle. However, because GNSS signals are distorted by their 20,000km journey between satellite and receiver, it can only achieve accuracy of around two metres on the ground, and therefore is far from sufficient for self-driving cars.
Correction services for autonomous cars go one step further, providing corrections for the region where the user is located via IP or satellite L-Band. Using this correction technology alongside other sensors enables precise positioning to the nearest centimetre, with multiple levels of redundancy should one sensor fail. This improves the safety of that journey, and by correcting map data and other sensors while on the move, the autonomous system becomes exponentially more accurate.
The next key part is inertial navigation. An initial measurement unit (IMU) is used to track the exact distance something has travelled from a fixed, known location. This is done via core components of inertial navigation—GNSS for absolute location, IMU to help the vehicle in terms of angular and accelerometer changes, and a host of other sources including wheel odometry (rotary sensors that measure the distance a wheel has travelled via how many times it has revolved), standard cameras, and proximity sensors. This is all used alongside maps data stored in the vehicle, which is updated in real-time by the car’s precise position.
It is essential that each component is tech-agnostic. Software needs to work with any hardware, and vice versa, so any vehicle manufacturer can use the best quality tech rather than only what fits or what they’ve produced in-house, maximising speed to market and value for money, and in turn, safety.m This last point is particularly important. Due to how expensive some of this tech is, and the clear relationship between a sensor’s quality and its cost, any unnecessary cost barriers should be removed to ensure the more accurate and reliable tech is used. Furthermore, using the deepest tech stack collects a wider variety of data, which will speed up the machine learning process.
Legal and infrastructure requirements
Moving autonomy past L2 and 3 and making it available to road users in more contexts, such as in urban areas or on smaller roads, will require substantial changes. It throws up challenges for all parties, not just for car manufacturers. Tier Ones, city planners, and the regulators and legal experts behind the Highway Code will have to collaborate to ensure safety and the rule of law. There will need to be a certain amount of flexibility to handle real-world situations and lawmakers will need to rethink existing laws to get ahead of every eventuality, on national and international levels.
Then, there’s the infrastructure, which will need to become smarter and more interconnected to ensure safety in potentially dangerous situations. For example, fog can obscure traffic lights, so an IoT network between said objects and the ADAS system would ensure a car never misses a red signal.
While we need collective ambition to move things forward, governments and business leaders should not lose sight of what is at stake here: public safety. Trust needs to be built, with a 2021 study showing that 74% of people do not trust them or believe that they can outperform their human counterparts. It would, therefore, be in everyone’s interest to take a synchronised approach of transparency and cooperation.
The first step is agreeing on a minimum level of technology. Although current tech is more advanced than most people realise, there are still advances to be made when it comes to enhancing safety. Key to achieving pinpoint accuracy is a global GNSS correction service like RTX, used alongside relative sensor systems, to create an ADAS capable of leveraging different positioning sources to ensure the safety, most reliable data, and redundancy.
In the dynamic world of autonomous driving, many things remain uncertain, or yet to be decided, about how we safely move up the scale, particularly when it comes to legal doctrine. However, a maximalist approach should be taken to the tech stack, as what we do know for certain is that the foundation of safe autonomous systems lies in precision, reliability and redundancy, achieved with accurate data.
The opinions expressed here are those of the author and do not necessarily reflect the positions of Automotive World Ltd.
Dr Rana Charara is Strategic Marketing Leader, automotive & IoT, at industrial technology provider Trimble
The AutomotiveWorld.com Comment column is open to automotive industry decision makers and influencers. If you would like to contribute a Comment article, please contact editorial@automotiveworld.com
