The chip at the wheel - autonomous driving and acceptance

Interview: Prof. Dr. Matthias Petzold

Head of Fraunhofer Center for Applied Microstructure Diagnostics, Fraunhofer Institute for Microstructure of Materials and Systems IMWS Halle


>>What we have known so far from science fiction films is now real on our streets: Cars communicate with each other, send their owners e-mails, control deviations in the middle of the trail and warn in time of an obstacle. So electronic systems take over, but the responsibility remains with the driver. Experts speak of partially or highly automated driving. In your view, has the future already begun?                                                              



Prof. Dr. Petzold


The future has already begun in any case. The technical innovations that are currently used in the test vehicles have already been developed in recent years. As a result, the performance of various important subsystems has increased dramatically at the present time. These include sensor technology, complete computer performance, communication via the Internet, data management systems and software. In the field of software, the current developments in artificial intelligence will also play a role in the future.

Many of these technical developments have been in operation for 15 or 20 years. The interaction of all these factors in the present time has made this particularly rapid development towards autonomous driving possible.

>>Cameras, sensors, assistance systems - our cars help us more and more, until one day they can do everything themselves. Is electronics responsible for the car's behavior in traffic? Do you think that's ethical?

I would divide the answer in two parts. On the one hand, of course, electronics can never take responsibility. The responsibility for the performance, safety and reliability of electronics that are to operate autonomously in traffic must be borne by the companies and organizations that market or permit such systems. There are already very strict regulations in the car, for example already today governed by a standard for functional safety ISO 26262. In any case, the safety of people and the passenger in the car must have the highest priority.

Regarding the question of ethical acceptability, there is the statement of the Ethics Commission set up by Minister of Transport Dobrindt that automated and networked driving is even ethical when these systems cause fewer accidents than the human driver, so when there is a positive risk assessment.

From the users' point of view, however, the perspective is certainly shaped by the feeling of being dependent on technical systems. However, we already have similar dependencies in many other, indeed critical, places, e.g. in robotic surgery. But also in the area of traffic, for example airplanes largely fly autonomously, except for short phases at take-off and landing. There are also already autonomously operating subways, for example in Nuremberg, in operation. The difference to autonomous driving, however, is the much greater complexity of the traffic situations that can occur on the road. In addition, then almost all people are affected.

>>In many cases, the ethics of automated driving was discussed, such as how the computer in a dilemma between two evils should decide: Driving over pedestrians or the car against a wall and thereby accept the death of inmates. Can a program make such a decision at all?

This is a question that has already been dealt with extensively and intensively. It is, to my surprise, always made very early, usually before the opportunities, benefits and benefits of autonomous driving for saving lives and preserving health are discussed by reducing the risk of accidents. It is still a question that is very strong and needs to be answered.

There is also a very clear statement here: Ultimately, it is a question of algorithms that are programmed for autonomous driving. At first we all assume the software will be programmed so that the system does its best to prevent an accident. Another premise is that, in particular, personal injury must be avoided, which of course has a higher priority compared to property damage. For me the third premise is that there should be no individual personal calculation and no comparison at all between the life and health of the occupants in the vehicle and the life and health of persons on the road or in other vehicles. This is also the statement of the ethics committee: In unavoidable accident situations, any qualification of people is inadmissible. I am not a specialist on this issue, but I assume that this will be implemented accordingly: Avoid accident as best as possible, prevent personal injury as best as possible, and nothing further beyond that.

>>In the near future, the fully autonomous driving should have prevailed in everyday life. Whether this succeeds and whether man is willing to leave his autonomy to technology is currently one of the most frequently discussed questions. Do you see an acceptance of autonomous driving in the population?

I would answer the question with a clear yes. Some of this acceptance is already being created today through the continuous introduction of ever new driver assistance systems, such as traffic jam or parking assistants. As a result, there is a gradual transition in which bit by bit more autonomy is taken over by the electronics or the software. So we're not dealing with a sudden disruptive introduction of a whole new technology that makes people have a hard time making friends. Rather, a gradual habituation to a growing number of autonomous subsystems will occur, which will decrease more and more our driver activities.

However, this incremental innovation over a longer period of time also allows for a continuous technical learning process for each new stage of development separately, reducing risks. By the way, autonomous driving also promises further advantages: While in private transport only the individual driver learns from coping with critical situations, automated driving makes it possible to transfer such a learning experience not just to a single vehicle but to a whole fleet of millions of vehicles, by using updates of the software.

But acceptance will also arise with reference to the usage behavior. For example, the ability to productively use the time released in a self-driving car, e.g. for office work, for communication or entertainment for leisure. Furthermore, mobility is also guaranteed for elderly or disabled persons. Another very important point is that autonomous vehicles are also of particular interest to fleet operators of self-driving Car Sharing cars, autonomous taxis and shared taxis, as well as smaller and larger buses. Depending on requirements, such a vehicle can be ordered by the user via smartphone and software in order to travel comfortably and autonomously from point A to point B. This also corresponds to the current trend - away from your own car, to a "mobility on demand", this means to cars that you call out only when they are really needed. Since I do see many reasons for a high attractiveness, I do not predict any problem. Nevertheless, many users will of course not want to miss the fun of driving a car. As a consequence the transport systems must then continue to adjust to a so-called "mixed traffic", this will be not so easy.

>>Let's talk about the technique. A car that steers, brakes and navigates independently only works with powerful electronic components. But also the electronics have their limits. Can there be safe autonomous driving at all?

Of course, there can never be absolute security. The question actually is, can driving be safer by the introduction of such autonomous systems? Can we do more to save more lives and protect our health better than we have before? Today, nearly 1.3 million people die each year in traffic worldwide. It is known that about 90 percent of all accidents involving serious consequences for life and health are due to human error. This leads to the expectation that technical systems that adhere strictly to rules and have a much better overview of the current traffic situation thanks to various sensor and information systems will make it possible to significantly reduce the risk of accidents. If this is not achieved, one of the most important reasons for autonomous driving would be eliminated.

However, one look back: If one looks at the number of road deaths in Germany in the past, this has decreased dramatically through passive protection systems and active electronic regulations, from approximately 19,000 in 1970 to just over 3,000 in 2017. This has been achieved, although the number of cars, the number of miles driven, and the speed with which they are driven have increased significantly.

A positive balance for autonomous driving is therefore possible in any case and the required performance of the technology seems already given today, even if questions of reliability under all driving conditions are still to be clarified.

>>Electronic systems and chips can break. In addition, there are high expectations for the service life of 10 to 15 years. What happens if the electronics fails?

There will certainly be security measures. Initially, intensive work is being done on how to increase the life of individual automotive electronic components and systems, e.g. through new, more knowledge-based concepts to ensure reliability under load ("robustness validation"). Furthermore, in the electronic systems themselves so-called mechanisms of self-monitoring are integrated, which immediately display and respond when certain functions no longer operate in accordance with standards. In addition, redundancies must exist. That is, there are several systems that have the same function and thereby not only complement each other, but also monitor. This works, for example, with various camera, laser and / or radar systems that react independently of each other to obstacles in front of the vehicle. Other possibilities are external monitoring systems e.g. by the manufacturer or according to further developed maintenance strategies. In the development of electronics, 10 years to 15 years is a long time, there will certainly be a regular replacement of hardware components to keep up with the technical development.

Nevertheless, there is still a lot to do, especially in this area. In addition to electronics, this mainly affects the software with its enormously high complexity. How can you even test them? At these sites, further research is needed to ensure reliability over a longer period under all conditions.

>>Autonomous driving - do we dare to take a look? How far away are we from the self-driving car and what are the next steps? Which level of autonomous driving will we reach from your point of view? When does fully automated driving or autonomous driving come?

Partial autonomy has already been achieved in road traffic today. There is the possibility that cars park themselves, or drive with traffic jam assistants, without the steering wheel being touched. Already last year, a large German automaker has launched a series-produced vehicle that can drive autonomously on the highway up to a certain speed including overtaking maneuvers - the vehicle itself can take on flashing, starting, accelerating, steering and braking. However, in many cases the legal requirements for this semi-autonomous traffic still have to be created. For Germany this is expected for the coming year.

If we assume a completely fully autonomous car, in which no intervention by the driver is anymore necessary or even possible, then I assume that this can be a reality in the period from 2025 to 2030.

>>The vision of autonomous driving is one of the most important drivers of a longer-term development in microelectronics. Do you find that the current speed in the advancement of technology is too fast?

I have never experienced such a rapid pace in a technical development of this breadth in my professional environment. But a purely technical development can never be too fast. It is only important when and under what circumstances such technical systems are actually released to the consumer. As already mentioned, this requires a very high level of awareness and responsibility for the issues of safety and reliability. Of course, this applies especially to the highest development level of fully autonomous vehicles. Although such completely self-driving cars have already driven several million test kilometers in total, they have so far only been used in special cases for the public and are almost exclusively operated by supervisory staff who can intervene if necessary. This means that we are still in a test and development phase and the time of market introduction is still open. Moreover, with regard to the open questions on safety and reliability, legal approval can only be granted once all legal boundary conditions have been clarified, the state approval procedures have been passed and completed, and corresponding certifications have been issued.

But I also believe that the manufacturers and operating companies themselves have to look very carefully when they can bring with the necessary responsibility such fully self-propelled cars of the highest level of automation in the market. Of course, with regard to life and the protection of persons, this is always an ethical issue, but it also has a market economy effect. Anyone who causes accidents at this point by launching them too early will pay a very high price in the reputation of their brand and cause serious damage to their business.

All three of these factors - the legal requirements for functional safety, ethical responsibility and market economy - point in the same direction, only to bring autonomous cars on the market, if one can assume a very high degree of freedom from accident and safety for the inmates.


>>You just said that you have never experienced such a rapid technical development. Is not such a rapid development dangerous that something is overlooked?

There is no doubt that competition and market demand generate very high development pressure. But there are also more qualitative reasons for the fast pace. In the race for autonomous driving, the world's most capitalized and at the same time most innovative market participants compete or form co-operations with each other. This concerns on the one hand Internet companies, like e.g. Google and maybe Apple, who want to expand their business through integrated mobility services. This leads to specialized mobility providers via fleet models operated worldwide such as Lyft and Uber with high risk capital in the background. This applies not only to the German premium automotive industry, which is also very efficient in this field, but also to car manufacturers worldwide. Added to this there are the highly-developed and innovation-driven companies in the global semiconductor and electronics industry and their suppliers. It can be considered as the global créme-de-la-créme in terms of economic performance, innovation, research and development capacity and capital. These companies can not only attract the brightest minds, they can quickly mobilize thousands of developers to achieve their goals. In addition, they also gather many innovative start-up companies for specialized developments around them and conduct preliminary research with world-leading scientific institutions. This probably unique constellation makes this high tempo possible in the first place.

Of course, in this race and in the enormously high complexity of autonomous driving there is always the danger of doing something wrong. The associated risks must be mitigated by the combination of technical safety measures already mentioned, such as proven reliability in the technical design, self-monitoring and redundancy of the systems in operation, the manufacturer's own responsibility for their products, as well as the regulations on state approval procedures and legal constraints.