How Robot Cars Talk to Us

**Autonomous Driving** Even nonverbal communication between human drivers is complicated. Communication with other road users is a particularly challenging task for machines.

No voice, no facial expressions and no body language - robot cars will be the most difficult road users to figure out in the near future. To ensure that the mute machines do not pose a safety risk at intersections and crossings, they will have to learn a comprehensible language with which they can communicate with pedestrians, cyclists and drivers. What this might look like is currently being negotiated.

It has long been clear to developers and manufacturers that robot cars need to communicate with their environment. Only in the "how" was there a lack of orientation. Each prototype and each study pursued its own approach: Audi experimented with projections, light symbols that the car casts onto the road by means of the headlights, where they are read by other road users. The Swedish supplier Semcom relied on a screen face on the radiator grille that smiles friendly at passers-by when they should cross the road. Similarly humanizing: the giant electric eyes from Jaguar. Including eyelid, iris and pupil, they should make eye contact with road users. As coherent and interesting as the ideas are in each case: Wild growth does not work in road traffic. When autonomous cars hit the road in large numbers, they will need a uniform language. Each car manufacturer establishing its own solution is not a solution.

> 17 Rules for New Language

Engineer Stephan Cieler is looking for the universal form of communication for robot cars. The developer at the supplier Continental is working in the ISO standardization committees on the standardization of human-machine interfaces. He is concerned with how robot vehicles and living road users can exchange information most safely in the future. At crosswalks, for example, where the most important question is whether the car will actually stop when you step onto the road. Or at bottlenecks, where two drivers have to agree on the right of way.

Cieler and his team initially developed 17 rules for human-machine interfaces. Something like the basic set of rules for communication between the car and its animate environment. The car's signals must be unambiguous and, moreover, consistent with the vehicle's other behavior. If the car indicates to another road user that it is waiting at an intersection, it must not accelerate briefly to the stop line afterwards.

Furthermore, once a message has been set, it cannot be changed again in the course of communication. In addition, it must be ensured that signals are always clearly recognizable for other road users under all weather and lighting conditions.

The result is an amazingly simple language mechanic. "We believe that communication must be visual, very, very simple, and that only a few messages are necessary or meaningful," says Cieler. That would also reduce the risk of misunderstandings. "For example, if the car is supposed to signal I saw you, that may still make sense to an individual passerby. But who does the vehicle mean in a group or a crowd of people?" the engineer gives an example.

He uses the same argument against communicating instructions or requests - in such cases, it is not always possible for other road users to clearly determine who is meant as the addressee in the urban bustle. In the end, therefore, just two messages are needed that the robot car sends out into the environment: "Autonomous driving active" and "I grant you priority.

> Light Band around Car

Cieler has a suggestion for the concrete design of the robot car HMI. It, too, is impressively simple: A turquoise light band around the vehicle indicates that it is in autonomous mode. When it flashes, it tells it to stop and give the other person the right of way. Turquoise because, Cieler says, unlike blue or yellow flashing lights, the color has not yet been used in traffic. And because it does not imply any instruction - such as a red stop light or a green that suggests free travel.

Whether Cieler's approach will gain acceptance among industry, science and politics remains to be seen. It is not entirely unlikely, as the researcher says that the approaches of other bodies are also moving in a similar direction.

They all want to be simple, intuitive and easy to understand. But Cieler believes it will be another ten years or so before there is a European or even global standard. This is a time horizon that many experts also estimate for the market penetration of autonomous driving.


Basler Zeitung, 2022-08-16, p. 25

DOT FROM lambda-browsing