MUNICH, Germany – Researchers at the Technical University of Munich (TUM) have made a significant breakthrough in the development of autonomous driving software. The team has created an algorithm that takes into account the 20 ethics recommendations from the EU Commission expert group, allowing for fairer risk distribution on the road. The software is considered to be a major step forward in ensuring the safety of automated vehicles by evaluating the varying levels of risk to pedestrians and motorists.
The TUM researchers have tackled the complex ethical questions that arise when developing autonomous driving algorithms. The team’s work is the first to incorporate ethical considerations into the software, instead of relying on binary, either/or decision-making. The algorithm was tested in over 2,000 critical scenarios on various types of streets and regions, including Europe, the USA, and China.
Maximilian Geisslinger, a scientist at TUM’s Chair of Automotive Technology, explains the approach: “Previously, autonomous vehicles were limited to making either/or choices in ethical situations. However, street traffic can’t be divided into clear-cut, black and white situations. Our algorithm takes into account the various shades of gray and makes an ethical choice from thousands of possible behaviors, all in a matter of a fraction of a second.”
The software’s ethical parameters were defined by an expert panel and are based on the principles of prioritizing the worst-off and fairly distributing risk among all road users. The research team classified vehicles and pedestrians based on the risk they pose to others and their respective willingness to take risks. The algorithm was then programmed to not exceed a maximum acceptable level of risk in various street situations and also takes into account the responsibility of traffic participants to obey traffic regulations.
The new software offers more options in critical situations compared to previous approaches, which limited the number of possible maneuvers in unclear situations and often resulted in the vehicle coming to a stop. The researchers’ risk assessment allows for a greater number of possible maneuvers with reduced risk to all parties involved.
Franziska Poszler, a scientist at TUM’s Chair of Business Ethics, comments on the new approach: “Previous methods relied on traditional ethical theories, which often resulted in a dead end as there was no alternative to violating one ethical principle. Our framework puts the ethics of risk at the center, allowing us to take into account probabilities and make more differentiated assessments.”
The TUM team emphasized that even the most advanced ethical algorithms cannot guarantee accident-free street traffic. Further considerations, such as cultural differences in ethical decision-making, will need to be taken into account in the future. The code developed by the TUM researchers is available as open-source software and will be tested on the street using the EDGAR research vehicle.