At the beginning of the year, the Law Commission of England and Wales and the Scottish Law Commission proposed unprecedented legal reforms on self-driving vehicles that have the potential to revolutionise the future of travel in the UK if accepted by Governments. The published joint report suggests that the legal responsibility for accidents caused by self-driving vehicles should rest not with the person in the driver’s seat, but with the company or body that obtained authorization for the self-driving features used by the vehicle. This adds a moral dimension to the technical aspects of self-driving innovation and has opened a debate on whether it is fair or not to shift responsibility from humans to a machine.
Is it fair? Indeed. However, we should be framing the entire question differently. We should start thinking of “autonomous cars” as proper “A.I. drivers” – or “automatic-drivers” and “robo-drivers” interchangeably – which are not different by any means from a physical driver that takes control of the car.
- Overview of self-driving cars sector – where are we now, when can we expect them to be available
Despite the fact that in 2015 Tesla CEO Elon Musk predicted that self-driving cars with unlimited mobility would have been available by 2018, there are currently no vehicles on roadways that are fully driverless. While Tesla, GM, Google and other auto manufacturers and tech companies have invested billions in developing driverless vehicle technology, at the moment there isn’t even a legal definition of “self-driving” technology in the United Kingdom. Therefore, the legal reforms proposed by the Law Commission of England and Wales and the Scottish Law Commission may only apply to technology that does not yet exist.
So far, the current generation of commercially available systems fall into the category of driver assisting technologies. These include front-crash prevention systems, self-parking technologies and the “level-2” autonomous car, a Drive Pilot which allows the driver to hand over direct control of steering and speed in certain circumstances, while still supervising the overall operation of the car. While these technologies are already revolutionising the automotive sector, it’s not clear when all the pieces will truly fall into place to allow for driverless technologies to safely navigate public roads among traditional cars. While experts agree that there will be a time in the future when this will come true, they disagree on the timeline.
However, a different picture is rendered looking outside the public roads and commercially available solutions. For example, the motor racing sector is ahead of times in terms of innovation and is promoting an unprecedented research effort in self-driving technology. In 2019, the best universities in the world decided to start working intensively on self-driving AI with the aim to compete in the world’s first head-to-head high-speed race with self-driving cars, the Indy Autonomous Challenge. Two years later, the challenge has taken place twice, demonstrating the technical feasibility of full self-driving technologies and bringing closer the vision of a future with self-driving cars on the public roads. In particular, the autonomous motorsport experience has been extremely useful to test limit – or “edge” – conditions, such as limited road-tire adherence, fast movements, complex vehicle dynamics, and complex multi-agent situations. Autonomous car racing can significantly contribute to the safety of autonomous cars of the future, by developing the complex know-how needed to manage edge conditions in critical situations on public roads.
Insights on the functions and their safety – how fair is it to give fault to the vehicle rather than the driver? What responsibilities does that the driver hold in the movement?
Besides the technical aspects though, the legal issues need to be clarified for the self-driving technology to reach the general public with an acceptable standard of safety, which includes legal certainty on the rights, duties and responsibilityes of the user. The legal aspects and the technical ones are intimately connected, as the attribution of responsibility for whatever action depends on the level of autonomy allowed by the specific technology. Traditionally, there are 5 levels of autonomy – with 5 representing the highest a machine can have from the human user, and existing self-driving technologies stand in between level 2 and 3.
Level 2 vehicles are also referred to as “hands-off”, because they have internal systems that take care of all fundamental aspects of driving: steering, acceleration and braking. However, the driver must keep their hands on the wheel at all times and stay alert to intervene if any part of the system fails. This level is still a cooperative type of control between man and the machine, who interact on both a longitudinal and lateral dynamics. The direct cooperation between the driver and the vehicle – which is called “cobotics” – is the most critical aspect of this level, both in technical and legal terms, because it is extremely difficult to attribute responsibility of an action when human and autonomous forces are overlapping. Therefore, so far, the driver legally is responsible for the vehicle and must always remain in control of the car.
Level 3 vehicles are ones instead that can truly be considered autonomous, while not being self-driving yet. In fact, they are often referred to as “eyes-off” vehicles, as the car takes care of everything while the driver is allowed to safely use their phone or relax, although they are still required to intervene if necessary. While technologies for Level 3 do exist, they are still in a grey regulatory area for what concern the obligation to be “eyes-on” while driving on the public roads and the attribution of responsibility in case of accidents. Moreover, even if the regulatory aspects were clarified as suggested by the joint report of the two Commissions and level 3 vehicles became popular, we would face a severe problem in the medium-long term: we can expect that drivers would progressively lose the ability to drive and, despite having a valid driver license, miss the experience required to take control of the vehicle in critical situations. In that case, it would be hard to assess whether an accident caused by the inexperience of the driver is their fault, or attributable to the technology that should not have failed from the start.
- Insights on the general safety – what tests have been carried out so far? What are the results? What needs to be worked on?
Despite Audi A8 was intended to be the first level 3 car authorised on the road with its Traffic Jam Pilot, Audi abandoned the project before receiving any safety approval, stating that they would wait for governments to establish a clear regulatory framework before working for making the system widely available. Contrarily, Mercedes Benz continued working on self-driving features and, in December 2021, became the first automaker officially authorised to produce level 3 vehicles. With the Drive Pilot system installed on car models Mercedes-Benz S-class and Mercedes-Benz EQS, they have received approval from UN-R157, a United Nations regulation body that sets the standards for level 3 autonomous driving technologies in vehicles. In Germany, the first customers of these cars are safely authorized to drive in conditionally automated mode at the speed of 60 kmph in heavy traffic, congested situations or stretches. The automaker is conducting expensive tests of their system in other European countries, the USA and China, and will start rolling out as soon as the other governments allow.
While level 3 is reality in technical terms and is close to a solution on the regulatory aspects, it is still far from the fully self-driving technologies envisioned by researchers in the sector. Level 4 vehicles, referred to as “mind-off” because the driver is not required any form of alert or intervention, are currently in the development stage and have been used only in geofenced areas as part of a test programme. An example of this is Google’s Waymo project, that since last April has been operating level 4 technology on taxi in the state of Arizona, although a test driver was always “mind-on” during the tests. As for level 5 vehicles, that do not require human interaction whatsoever, there is a long way to go before they could be on the roads. So far, the closest existing examples of level 5 vehicles are the Audi Aicon, presented in 2018 but very unlikely to be made commercially available soon, and the racing cars driven by AI that took part in the Indi Autonomous Challenge. While successful on track, it is currently impossible to think their application outside of a controlled environment such as a racing competition.
- Predictions on how the product will evolve in the next few years
We can expect Level 3 technologies to become popular in the new few years, while they are increasingly recognised by governments within their regulatory frameworks. Their mass commercialisation however will not have a disruptive impact on the current automotive industry because it would not modify traditional mobility models. The real disruption will come when Level 4 and 5 technologies will be available to the public, as at that point, traditional mobility models will quickly transition into robo-taxi mobility models. Research demonstrates that the current mobility needs could be fully satisfied by just 1/10 of the vehicles on the roads today. The introduction of cheap, self-driving taxis with no driver fee, will enable a revolution in the automotive industry with two possible outcomes: a gigantic reduction in the number of cars on the public roads, and shift in the automotive market from a B2C approach to B2B.
Author: Sergio Matteo Savaresi, Full Professor in Automatic Control at Politecnico di Milano