Interview with Forschungsgesellschaft Kraftfahrwesen mbH on automated trucks

Automotive IQ sat down with Jens Kotte,Manager Active Safety at Forschungsgesellschaft Kraftfahrwesen mbH, and discussed the latest developments within automated commercial vehicles ahead of the upcoming conference Automated Trucks where Mr. Kotte will be presenting. 1) What are the major technical issues in creating multi-brand platoons and how do we overcome it?

Multi-brand platoons are already feasible today. There are some challenges, which are currently in the scope of research activities. One of these is for example the break performance. If we talk about small gaps between the trucks, we have to make sure, that the following truck comes to halt behind the leading truck, even if the leading truck is doing an emergency brake. Therefore, the order of the platooning trucks has to be determined based on the braking performances of the trucks. To do so it is necessary that all OEMs agree on a standard about the content of messages that will be exchanged between platoon trucks. Another issue is a cooperative and efficient behavior of platoons towards other traffic participants. Therefore we need pilots to observe what other traffic participants expect from platoons.

2) What are the challenges and benefits in the area of C-ITS for platooning and automated driving systems?

The challenges are to ensure a secure, stable and trustful communication with a standardized dictionary so that vehicles understand each other. The basis for that is for example set with the SAE J2735 standard (“Dedicated Short Range Communications (DSRC) Message Set Dictionary”), but has to be extended for new uses cases such as e.g. platooning. The benefits are that C-ITS enables vehicle to get information about the environment outside the visible area of today`s on-board sensors and it enables a faster transmission of information about events. The first point is especially important for level 3 systems since it takes a situation dependent time period until the driver is able to take over control. With on-board sensors it is difficult to look ahead for longer distances than 250 m. The second point is important for safety and platooning systems. On-board sensors can only detect actions of the vehicle, which are results of driver inputs. This means with C-ITS we can already react on the actions of the driver, when he/she e.g. pushes the brake pedal. On-board sensors can just detect the deceleration of the vehicle so that a vehicle with a C-ITS system can react early. This enables the smaller gaps in platooning systems.

3) How far are we in platooning regulations and the legal framework, and what can/needs be done to be done for a connected Europe/USA?

If we talk about platooning, we have to distinguish between different systems. Level 1 platooning systems (also called C-ACC), in that the system just takes over longitudinal control, can be actually operated in real traffic today, if the gap between the trucks is large enough. Since platooning for efficiency reasons just makes sense if the platoon operates with gaps lower than the allowed ones today, the regulations for allowed gaps between vehicles have to be adapted. If we talk about level 3/level 4 platooning systems we face the same legal problems as other automated driving functions do.

4) How can interoperability, data analysis & processing be used for better path planning and control?

Interoperability, data analysis & processing enables a new way of path planning in that not only the own path is optimized, but the paths of several vehicles around. Thus, the efficiency of the overall traffic system can be increased.

5) How machine and deep learning can be used to improve control behavior for highly automated driving?

Machine and deep learning has a great potential to accelerate the development of automated vehicles, but it is necessary to analyze reliably for what part of the functions it is really useful and applicable. A validation and safety assessment of an end-to-end learning approach is for example today not possible. But, for tasks such as object classification and behavior prediction, machine and deep learning is really powerful and already today usable.

6) What are the challenges in developing simulations, platform testing and related methods for automated driving?

We face many challenges in developing testing methods for automated driving. If the driver is not responsible for the vehicle anymore, the system has to be able to handle every situation that can occur. But how can we test it? Do we already know which situations can occur? Also new situations occur if automated vehicles drive in real traffic. It`s also not clear today, what can be tested in simulations and what has to be tested in real traffic. We currently work together with other partners in the project Pegasus on answers to these questions. Within Pegasus for example a database of relevant scenarios for highway scenarios is generated. This database will allow to collect relevant scenarios and apply them to several evaluation methods.