A new project at IMC Krems – University of Applied Sciences in Austria is developing realistic crash scenarios and simulations of future traffic situations involving multiple autonomous vehicles.
To begin with, the project team will extract driving scenarios from publicly available databases which will then be fed into specialist driving simulation software. Following this stage, an optimization process will be conducted, based on advanced algorithms which are capable of creating novel driving scenarios with increased criticality and severity.
IMC Krem’s project is part of a European Union-funded international research program which seeks to develop safety mechanisms for AVs to reduce accident-related consequences by using advanced manufacturing technologies.
The team believes that in the future, AVs could reduce the number of traffic accidents, but this will require the safety of such vehicles being enhanced beyond present day capabilities. Achieving this is viewed as a difficult task as simulating crash scenarios of mixed traffic situations – such as involving cars with different levels of autonomy – is hampered by a lack of relevant data. The project carried at IMC Krems aims to change this by harnessing the power of purpose-developed machine learning algorithms which will extract and analyze data from existing car crash databases.
“Autonomous driven cars are still not capable of avoiding accidents in all possible situations”, said Prof. Alessio Gambi, project leader at the Department of Science and Technology at IMC Krems. “And as they react autonomously and hence in other ways than human controlled cars, crashes will look differently to the ones seen so far. But currently we don’t exactly know what they will look like. This lack of knowledge is a hindrance to improve the safety of future mixed traffic situations.”
Dr Gambi and team will now contribute to a knowledge base for addressing this problem in an international research project funded with €4m (US$4.35m) by an EU grant. Using data from existing sources – such as large databases recording car crashes – the team will select a set of reference driving scenarios as a base for the next stage of the project.
“It is noteworthy”, explained Dr Gambi. “That worldwide only one openly accessible database exists that actually records the level of autonomy of cars involved in a crash, California’s DMV autonomous vehicle collision report database. That is a very limited base for simulating future car crash scenarios. Our project will help to widen that base.”
The team will feed BeamNG.tech (a state-of-the-art driving simulation software) the extracted reference scenarios from existing sources including CARE, GIDAS, STRADA and ZEDATU. Furthermore, an online, open simulation platform will be developed which follows multi-player paradigms similar to those from video games where remote players interact with each other and with an artificial intelligence. Through the use of this platform, the team will study virtual live interactions between human drivers and (simulated) AVs and be able to generate an additional set of traffic scenarios which are not based on previous accidents but on real interaction.
By using the two sources, specialized search algorithms will now calculate virtual crash scenarios that anticipate possible actions by AVs.
“We actually develop these algorithms in-house and hence we can easily increase the virtual criticality and severity of the simulated crashes,” said Gambi. “This will expose car structures chiefly involved in severe crashes and predict their behavior in such situations.”
At present, the project is running in parallel with a larger EU-project called Flexcrash, with IMC Krems providing the basic work package. The results from the advanced simulation will provide data for improving the design of future autonomous cars. The final goal is to use hybrid manufacturing technology for applying surface patterns using additive manufacturing onto preformed parts. This will greatly help reducing accident-related fatalities, injuries, pollution and manufacturing costs in future.