Robot cars with social skills

Transparenz: Redaktionell erstellt und geprüft.
Veröffentlicht am und aktualisiert am

The Fraunhofer Institute is working on an AI system that can recognize the behavior of pedestrians.

Das Fraunhofer Institut arbeitet an einem KI-System, welches das Verhalten von Fußgängern erkennen kann.
The Fraunhofer Institute is working on an AI system that can recognize the behavior of pedestrians.

Robot cars with social skills

Communication with other road users plays a crucial role in the development of autonomous vehicles. The robot cars must be able to recognize which passers-by could become relevant in order to then record and interpret their behavior. The Fraunhofer Institute for Optronics, System Technology and Image Analysis IOSB in Karlsruhe has now presented the prototype of a system that is intended to achieve exactly this using artificial intelligence.

Does the pedestrian want to cross the street or not? This is a key question for Fraunhofer researcher Manuel Martin. What a human driver usually perceives intuitively and without thinking, for example based on location, line of sight and gestures, must first be taught to an autonomous vehicle. Only then can it safely act independently in a residential area or in front of a school. Artificial intelligence (AI) can analyze video images for this purpose, but must process enormous amounts of training data in order to draw the right conclusions. This is exactly what Martin is working on as part of the research project “Intelligent human-technology communication in mixed traffic”, or “Initiative” for short.

“We have now implemented a research prototype that estimates whether a pedestrian wants to cross the street by analyzing their gestures and thus creating the basis for the interaction,” explains Martin. The system consists of a stereo camera that can see spatially and thus record the exact position of passers-by, as well as an AI algorithm that records the positions of the limbs and draws conclusions from them. “We successfully demonstrated this system at a project presentation at the halfway point of the initiative,” said the IOSB researcher. “Now it’s a matter of further training the AI ​​and refining the system overall so that it can recognize pedestrians’ intentions as accurately as possible in all conceivable situations.” If the car notices a pedestrian trying to cross, it should be able to use an unmistakable light indicator to inform them that it is about to stop or wants to drive past.