Paderborn University research prize awarded
A ball rolls onto the road, a child stands on the pavement - drivers' alarm bells immediately ring. The result: they brake because they assume that the child will run onto the road. Drivers are also able to anticipate the behaviour of pedestrians in other potentially dangerous situations. Autonomous vehicles, which will increasingly characterise the streetscape, cannot do this. Although news technologies react to critical situations, they lack the ability to predict behaviour. A new research project is now being launched at Paderborn University that aims to close this gap and enable autonomous vehicles to recognise the intentions of pedestrians even before they act.
Experimental studies on the decision-making behaviour of pedestrians
The future of transport faces a major challenge: how can autonomous vehicles and pedestrians interact efficiently and safely? This is the question that drives Dr Sandra Gausemeier and Dr. rer. medic. Tim Lehmann. Their idea: autonomous vehicles should recognise intentions by using a combination of AI methods (artificial intelligence) and motion analysis. The approach is new and promising. Dr Gausemeier is an expert in driver assistance systems in the "Control Engineering and Mechatronics" department at the Heinz Nixdorf Institute. Research in the field of model-based development of mechatronic systems is therefore part of the scientist's everyday work. Dr Lehmann is a research assistant in the Training and Neuroscience working area of the Department of Sport & Health. He specialises in researching human motor behaviour and the underlying neurocognitive processes. The two scientists have joined forces and are conducting experimental studies on people's decision-making behaviour for their project. These will later serve as the basis for predictive algorithms in autonomous vehicles.
Bold ideas for science
The scientists have been honoured with the Paderborn University Research Award for their project. The university management presents the award, which is endowed with 150,000 euros, in recognition of exceptional research projects outside the mainstream and thus sponsors bold ideas for science. "By combining artificial intelligence and neurocognitive analyses, the project aims to bring about a paradigm shift in the interaction between humans and autonomous systems. This is not only of great relevance to society, but visionary in the best sense of the word," says Prof Dr Thomas Tr?ster, Vice President for Research and Academic Career Paths at Paderborn University.
More than collision calculations
"Our goal is to develop an AI-based system that can assess pedestrians' future intentions based on their motor skills, predict their behaviour, create risk profiles and thus proactively avoid critical situations," explains Dr. Gausemeier. To this end, experimental studies on the decision-making behaviour of people in real urban scenarios are to be carried out for the first time. "This goes far beyond simulation or laboratory-based studies and addresses the complex and highly dynamic interactions between humans and machines. Autonomous systems should then be able to incorporate not only classic collision calculations, but also the situational awareness and distraction of pedestrians into manoeuvre planning," adds Dr. Lehmann.
Pattern recognition to recognise human movement sequences
The behaviour of other vehicles is determined on the basis of traffic regulations, among other things. The number of possible manoeuvres is therefore limited to a few options. Pedestrians do not have such severe restrictions; both their movement and decision-making options are much more flexible: "This is where machine learning (AI) methods are to be used to understand the complexity of human movement sequences over several seconds using pattern recognition and to predict intentions with a high degree of reliability," explains Dr Lehmann.
Increasing the safety of all road users
The quality of the training data is crucial for pattern recognition using AI. To achieve this, the scientists want to develop a multi-stage process with different data. Dr Gausemeier explains: "To collect the data, test subjects will be equipped with eye tracking, mobile electroencephalography, i.e. the measurement of brain activity, multisensory mobile measuring systems and motion capturing. This will allow us to classify the effects of situational parameters and cognitive cerebral decision-making behaviour with regard to the resulting movement sequences." After training, the autonomous systems will be able to recognise intentions based solely on the onboard camera images and draw conclusions about future movement sequences. "This approach can substantially increase the safety of all road users," says Prof Tr?ster. The team expects initial results at the beginning of 2027.
This text was translated automatically.
