When is the self-driving car going to come?

15. July 2021

It was only a few years ago that the “robot car” seemed almost to be waiting at the gate – more recently, however, things have noticeably quietened down on the automated driving front. The German government is now seeking to give the technology a boost by changing the law. Exactly what will the “Autonomous Driving Act” make possible?  Which of the five stages of automation are we currently at, and how are things going to proceed internationally? All these questions are addressed in this interview with Katrin Leicht from TÜV NORD. 

#explore: Self-driving buses are being tested in many parts of Germany, albeit always subject to special authorisation. What will the “Autonomous Driving Act”, already passed by both chambers of the German parliament, change?

Katrin Leicht: The “Autonomous Driving Act” sets out the possibility of deploying fully automated vehicles at Level 4 while dropping the requirements for special authorisation and a safety driver. In this case, a control and monitoring centre – the Technical Monitoring system – would be the body of last resort with the duty to intervene in an emergency. The plan with this regulation is for the first vehicles of this kind to be approved as early as 2022. Germany will thus be the first country in the world to allow driverless vehicles onto the road at the national level.

The “Autonomous Driving Act” isn’t intended to offer blanket authorisation for these vehicles; the idea is instead for them to be deployed in limited scenarios. What is the specific thinking here?

The idea is to allow the on-road operation of shuttle vehicles –  that is, people carriers – and trial vehicles. In specific terms, we’re looking at minibuses which take passengers on fixed routes or goods transporters designed to carry goods for the last part of their journey or between two distribution centres. “Limited scenarios” also means that the area earmarked for their deployment is surveyed and assessed in advance by the Technical Service people and the authorities.

Which tasks come within the remit of the Technical Monitoring body?

That depends on the nature of the envisaged deployment area. There needs to be a clear definition of the tasks and situations which the system can cope with independently, where it will run up against its limits and when the control centre will have to step in. For instance, if a sensor fails, the vehicle must be able to make itself safe – by stopping at the roadside, for example. It would then be able to report this fault to the Technical  Monitoring team, which would in turn take the appropriate measures. If there isn’t a safety driver on board, these vehicles will also have to have the communication channels they need to be aware of their environment. After all, there are lots of exceptional situations in which human drivers communicate by making eye contact or using gestures.

 

The Bundesrat, the upper house of the German parliament, has called for a few changes to the law in its report on the draft legislation. Among other things, the law should explicitly state that the vehicles must be capable of responding to traffic police offers and giving way to oncoming fire engines to allow them to get past. From the technical point of view, is this is already possible?

The vehicles need to be able to respond to flashing blue lights and traffic officers, of course. This is something we’re calling for as well. If there’s an accident on the motorway, for instance, they will also have to be able to follow the instructions of the emergency services or at least record them. In technical terms, there are a number of conceivable options here – but we’ll have to wait and see which solution turns out to be the successful one. Another conceivable option would be for the Technical Monitoring system to take over at this point. But if it had to intervene to control a number of vehicles at once, this kind of system could quickly run up against its limits. So, when it comes to designing the Technical Monitoring system, we need to define how many vehicles this kind of control centre can reliably monitor at once.

Does the legislation address the key question of liability?

There are indeed a few open questions here. The “Autonomous Driving Act” does set out the duties of vehicle owners, technical monitoring systems and carmakers. But who exactly is liable when and in which situations, or which criteria are used to determine liability, still needs to be worked out in detail.  

The minibuses which have been tested to date run on virtual tracks and stop when they run up against an obstacle. If they are going to be able to integrate themselves into the traffic flow, they will also, of course, need to be able to overtake. Are the vehicles already capable of doing this?

We’re currently involved in a project to optimise highly automated buses, which also features overtaking manoeuvres. This kind of extension of the operating range requires the perfect dovetailing and seamless interplay of safety concepts, localisation and environment monitoring systems. Here, we’re collaborating, among others, with the technical university of Braunschweig, which is working on these systems and assessing them with a view to getting them approved for road use.

The German government sees the “!Autonomous Driving Act” as an interim solution until such time as standard international regulations are adopted. How far are we away from that point?

Work is already ongoing at the European level on statutory requirements for highly automated Level 4 systems. The final draft legislation is scheduled for presentation in 2022. Once the statutory regulations are in place, we shouldn’t then have too long to wait for the technical concepts.

What level of autonomous driving do we currently find ourselves on?

For mass-production vehicles, the first Level 3 assistance systems have been permitted at UN level since January 2021. In specific terms, these relate to automated lane-keeping systems, or ALKS for short. In motorway driving, the systems can stay in lane, accelerate and brake independently at a speed of up to 60 km/h. Drivers can do other things during this time but have to take control again as soon as the car tells them to. Honda has launched its first car with Level-3 functionality in Japan. German carmakers haven’t yet applied for the authorisation for such a system, so it obviously hasn’t yet been granted. So, in Germany, we’re effectively still at Level 2.

How much time do drivers have to take back control of the car once told to do so?

They have ten seconds to take up the reins again. This is a challenge to both drivers and the technology. In ten seconds at 60 km/h you can travel 160 metres – which is a fair distance. So, the system needs to be capable of clearly recognising its own limits in advance and remaining fully functional during the transition. 

Such Level 3 systems should in the future be capable of driving at up to 130 km/h. At this speed, you will cover 360 m in ten seconds. And it’s significantly harder to suddenly get your bearings in the traffic situation again. Is it even possible for this to be done safely?

The faster you’re going, the earlier the system needs to recognise its limits and the more of the route ahead it needs to monitor. Connecting vehicles might improve the “distance vision” of everyone involved – but that’s still largely a pipe dream. But the demographic change is also a problem. The number of elderly people is growing. And the older the drivers, the more time they need to get their bearings behind the wheel. As this transfer of control from a human driver to a machine is such a big challenge, some manufacturers would prefer to skip Level 3 and go directly to Level 4. The aim would be to reduce the frequency of situations in which the human driver has to intervene.

What does the increasing level of automated driving mean for testing these systems?

This presents various challenges for testing organizations: The more tasks the vehicle takes on, the more traffic scenarios it must be able to handle—and these, of course, must be tested. The more precisely these scenarios are defined and described, the better they can be tested. That’s why we need standardized catalogs of driving situations so that we and other technical services can conduct comparable tests. As the number of driving situations increases, so do the demands on the test facility. International working groups are therefore currently identifying which test facilities around the world are actually suitable for testing automated systems. But even on an ideal test site, not all scenarios can be replicated. Sensors and systems must function reliably under varying lighting and weather conditions, but these cannot be created at the push of a button. In this respect, there will be no way around virtual testing methods in the future. Such virtual testing methods are being developed and validated as part of a collaboration on autonomous driving between Germany and Japan.

The goal is Level 5 autonomous driving, in which the car takes over all driving tasks. In your view, how far are we from achieving this, both technically and in terms of regulations?

In my view, Level 5 autonomous driving is not yet on the horizon. Work is progressing, and ultimately it’s probably just a matter of time. But for now, it remains a major challenge because such a vehicle must be able to handle all traffic situations safely and independently.

 

A while back, it felt like announcements about self-driving cars were coming out on a weekly basis. In recent years, things have quieted down considerably on this topic. Were manufacturers a bit too bold in their claims about when these systems might be ready for the market?

In fact, after the initial hype, things have indeed quieted down a bit when it comes to autonomous driving. However, this is a pattern that can be observed with most new technologies. Gartner’s Hype Cycle describes this quite accurately: It begins with a technological catalyst—in this case, Google’s developments. From there, it rapidly ascends to the “Peak of Inflated Expectations” and then plunges steeply into the “Trough of Disillusionment,” until it eventually levels off on the “Plateau of Productivity.” Electric mobility was no exception. However, how quickly or slowly such systems make their way onto the roads is also a matter of societal acceptance—and that varies from country to country. We Germans tend to be somewhat more conservative and safety-oriented in this regard, while Asian countries, for example, tend to take a more proactive approach to testing such systems. As for the state of technical development: When it comes to the overall package of vehicles, hardware, and software, Germany, with its automotive industry, is generally well-positioned. When it comes to the core software alone, IT companies like Google have, of course, taken a strong lead. This represents fierce competition—and, at the same time, a major driving force.

 

In Germany, in particular, one often hears the argument against autonomous cars that people don’t want to hand over the wheel to a machine at all. They say this would take the fun out of driving. Does this desire to drive oneself vary from country to country, or are manufacturers simply failing to meet the needs of their customers?

 

The differences exist not only between countries but also between generations. For the younger generation, the car is no longer a status symbol. And individual mobility can increasingly be organized in a multimodal way, at least in major cities. To get from A to B quickly, people use different modes of transportation. The more reliably I can find the right vehicle for each trip, the more the advantages of not having to drive myself come to the fore. A good internet connection and the ability to use their smartphone while on the road, for example, are more important to many young people than sitting behind the wheel themselves.