PRORETA 5: CONTINENTAL AND UNIVERSITIES ARE JOINTLY RESEARCHING AI FOR AUTOMATED DRIVING IN CITIES
>> Continental continues its long-standing PRORETA research cooperation with international universities under the slogan "urbAn drIving"
>> By the end of 2022, algorithms based on artificial intelligence (AI) should be developed and tested for the entire chain of effects of automated driving
>> Enhanced interdisciplinary and inter-university collaboration covers the entire range of technology for the relevant functional steps of automation
Frankfurt, Germany. Technology company Continental is continuing the series of its PRORETA research projects, together with Technical University (TU) Darmstadt, the University of Bremen (Germany) and TU Iaşi (Romania). PRORETA 5 is dedicated to one of the most challenging tasks for automated driving: recognizing complex traffic situations in inner cities and how algorithms from sensor data can deduce the correct driving decisions in these situations. At an unregulated intersection, for example, it is a challenge to correctly interpret all objects related to the intended direction of travel – including their direction of movement, intention and priority – without human intervention. Artificial Intelligence (AI) plays a key role in this. AI methods are to be tested when the implementation of traditional approaches becomes too complex or reaches its limits. The main advantage of the AI is that, following a training period, it is able to draw its own correct conclusions based on its learning, even in unknown situations. This element is reflected in the slogan “urbAn drIving.”
Division of Tasks: More than the Sum of the Parts
The PRORETA 5 project (2019–2022), which is based on a three-and-a-half-year term, is examining the algorithms of cognition, behavior prediction and decision making in a demonstration vehicle built and equipped by Continental. At the end of the project in September 2022, the aim is to assess the performance of the new AI-based automation for the Society of Automotive Engineers (SAE) Level 4, using the widest possible range of inner-city scenarios and thus demonstrate the potential for future deployment. The algorithms based on artificial intelligence should be able to correctly recognize and interpret these types of complex traffic scenarios, resulting in correct driving decisions being made. A sub-area of this will be to observe how a human driver reduces and assesses the complexity of the environment. The learning-capable algorithms of the PRORETA 5 project will be trained in accordance with similar principles in order to achieve a driving performance comparable to that of humans.
In order to cover the individual processing steps along the active chain of automated driving with new solutions in an optimal and time-efficient manner, the current PRORETA project has been extended to an inter-university and international level. The long-standing, proven cooperation between Continental and TU Darmstadt, which has dedicated itself to individual sub-tasks of driver assistance and automation, is forming the basis for the integration of further universities in the current research cooperation.
“PRORETA is a successful program. Its roots go back to 2002. With the current expansion and internationalization, we are taking on the biggest challenge in the field of automated driving – inner-city driving,“ says Karsten Michels, Head of Research & Advanced Engineering within Holistic Engineering and Technologies at Continental. “The interdisciplinary, international and inter-university team of PRORETA 5 brings together outstanding expertise in all sub-areas of the task.”
Inter-university and International Research
At the University of Bremen, PRORETA 5 will add special expertise in the area of environmental detection through sensor data fusion. Amalgamated under the technical term “cognition”, these are all processes associated with perception and recognition. In turn, TU Iaşi in Romania is focusing on predicting the behavior of other road users. The team at TU Darmstadt is focusing on the topics of systems and safety engineering, trajectory planning and control technology. In the meantime, PRORETA has reached the second milestone: The first measurement campaign to record training data in the prototype vehicle with Continental software and hardware is currently underway in Bremen. The vehicle, which was initially trained in Bremen, will be handed over to TU Darmstadt for further test drives during the course of the project.
“Today, the interface between automation and passengers in the vehicle also plays an integral role. Information, communication and driver observation in the context of automation and psychology are inextricably interconnected,” says Michels.
The project coordinator for the universities, Professor Hermann Winner, Director of the Institute of Automotive Engineering at TU Darmstadt, confirms the importance of interdisciplinary collaboration: “The teamwork between Continental industry experts, doctoral candidates and students offers the opportunity to develop state-of-the-art future technology for the mobility of tomorrow and do so very realistically with regard to vehicles. This cooperation is valuable to both parties.”
The long-standing cooperation between Continental, the Institute of Automotive Engineering at TU Darmstadt, under the management of Professor Winner, and the Control Methods and Robotics Lab, under the management of Professor Jürgen Adamy, also forms an essential foundation for PRORETA 5. In addition to the TU of Darmstadt, the following universities and leading experts are involved in PRORETA 5: at the University of Bremen, the Working Group for Cognitive Neuroinformatics, headed by Professor Kerstin Schill, who focuses on elementary cognitive abilities such as self-localization as well as object recognition and object tracking. From the Gheorghe Asachi Technical University in Iaşi, Professor Florin Leon of the Faculty of Automatic Control and Computer Engineering is coordinating the project contributions with regard to the behavioral prediction of other road users.
