The ability of autonomous systems to see may eventually be so good that it will be better than people, even with their experience. 10 questions for RWTH Aachen University professors Lutz Eckstein and Maximilian Schwalm.
02 Aug 2016 | 3454 Views | By Autocar Pro News Desk
The ability of autonomous systems to see may eventually be so good that it will be better than people, even with their experience. What are the consequences of this? 10 questions for RWTH Aachen University professors Lutz Eckstein and Maximilian Schwalm.
Researchers and developers throughout the world rack ...
This is an Autocar Pro Plus article. Subscribe to continue reading.
The ability of autonomous systems to see may eventually be so good that it will be better than people, even with their experience. What are the consequences of this? 10 questions for RWTH Aachen University professors Lutz Eckstein and Maximilian Schwalm.
Researchers and developers throughout the world rack their brains when faced with the big question of when automated vehicles will surpass people and become the better road users. At the Institut für Kraftfahrzeuge (Institute of Motor Vehicles at the Rheinisch-Westfälische Technische Hochschule Aachen (RWTH Aachen University), Prof. Lutz Eckstein and Prof. Maximilian Schwalm are considering different aspects of automated driving in their areas of expertise. Here are the most important assertions by the engineer (Prof. Eckstein) and the psychologist (Prof. Schwalm).
1. What are the strengths of a person in comparison to a machine when considering a traffic situation?
Prof. Lutz Eckstein: As the result of thousands of years of evolution, a person’s perception is very good. Cameras are not as well developed, especially in terms of long distances. People supplement the information that they perceive with knowledge and experience. Based on this, they have the ability to predict situations. This is precisely the problem that technology is fighting against.
Prof. Maximilian Schwalm: We humans are able to separate the important from the unimportant. Particularly in a traffic situation, the multitude of perceptions challenges us to filter out the things that are important for our actions from this information. The interesting question is: How can we teach a machine to correctly understand a situation?
2. What are the strengths of machines when detecting a traffic situation?
3. How can machines make up for the head start that people have in terms of experience?
Prof. Lutz Eckstein: My vehicle must be allowed to use knowledge about, for example, critical intersections that other vehicles have collected. The intelligence to master complex situations will not be completely accommodated in the vehicle; instead, it is derived from many sources. While it is driving, my car ‘retrieves’ this knowledge and benefits from the experiences that other vehicles have made in this context.
Prof. Maximilian Schwalm: The goal is to develop collective knowledge. If we are able to combine the content learned by all vehicles that we equip with such systems, it may actually be true that such a system will have more experience than a person and therefore acts more safely in such situations.
4. Are people or machines affected more by sources of interference such as rain, snow or fog?
5. What role will the interaction between man and machine play in the future?
Prof. Maximilian Schwalm: That depends on the level of automation. If we allow the driver to be inattentive, but they need to intervene in an emergency, latency can be expected. Cognitively, this is a challenge.
Prof. Lutz Eckstein: The question of how the driver gains confidence is important. We need some sort of guidance display on which the driver can see how the vehicle will behave so that they do not suddenly interfere when they think they have spotted a dangerous situation. This is the critical point.
6. Can the system take on total control, perhaps as a kind of monitoring of the person?
7. What obstacles on the path to fully automated driving still exist?
Prof. Maximilian Schwalm: One prerequisite for these autonomous systems is that we in point of fact actively take this responsibility from the driver.
Prof. Lutz Eckstein: Or that we return it only gradually. This is completely underestimated; we sometimes speak of an automation gap. We cannot simply become progressively better if an unplanned intervention by the driver makes the situation more critical. We will need steer-by-wire, in other words a purely sensor-controlled, decoupled steering wheel without mechanical power transmission, so that we are technically able to decouple the driver in the first place. The system will ultimately have to decide: How much of what the driver is doing right now should I take into account?
8. Will automated driving cause people to forget how to drive manually?
9. The path to fully automated driving is divided into five stages. Is level 5, autonomous driving, in the far future – or is it already being put into practice?
Prof. Lutz Eckstein: We will see preliminary systems between 2020 and 2025, but only in restricted areas. These involve automated means of transport for people and goods. On the other hand, we have the classic motor vehicle which anyone can buy and which is supposed to work everywhere – in this regard, we are still relatively far away from level 5. We are well underway at level 3 on sections of highways and for parking.
Prof. Maximilian Schwalm: We have to be careful that we do not destroy the acceptance. For certain technologies – we have seen this in the past – one or two very prominent examples are enough to suddenly eliminate any trust that has built up with a lot of effort.
10. Are there any areas in which people trust technology more, i.e. where automation is already highly accepted in society?
Interview courtesy: ZF