Type: Bachelor, Master

Date: Immediately

Supervisor: Niclas Kannengießer, Maximilian Schrapel, Jianxin Zhao

As autonomous vehicles (AVs) become increasingly integrated into transportation systems worldwide, ethical concerns regarding their decision-making processesin crash scenarios are gaining attention. These concerns are particularlypressing when AVs must make split-second choices involving vulnerable roadusers such as pedestrians and cyclists. While AV systems are typically trainedto minimize harm using large datasets and are tested for safety and performanceunder controlled conditions, their ability to make ethically sound decisions incomplex real-world situations remains limited and opaque.

One of the core challenges arises in scenarios where AVs must choose betweenmultiple harmful outcomes. For example, if a pedestrian falls onto theroad in front of a moving AV, the vehicle can choose to brake hard to avoidhitting the pedestrian. However, this action could cause a rear-end collisionwith another AV following closely behind, potentially injuring its passengers.Without communication between the vehicles, the first AV acts in isolation,unable to consider the broader consequences of its decision.

Advances in vehicle-to-everything (V2X) communication offer a new opportunityto address this issue. Through real-time data exchange, AVs could sharetheir preferred local strategies and collaboratively negotiate coordinated globalstrategies to mitigate harm based on shared ethical principles [1, 2].

However, two significant challenges remain. First, ethical considerations inharm minimization are often abstract and difficult to translate into actionableand effective strategies. Second, reaching a consensus on such strategies underreal-time constraints and potentially unreliable network conditions is a complextechnical problem. While V2X communication enables AVs to coordinate,the design of consensus mechanisms that support ethics-informed, time-criticaldecision-making has yet to be fully explored. 

Within the context of designing ethics-informed, coordinated decision-makingby AVs to mitigate harm, several topics for Bachelor/Master thesis are available,as listed in the following:

• Crash Simulation: Identify and simulate crash scenarios involving various types of road users and AVs. Develop harm functions in relation to trajectory alternatives, including adaptive behaviors, and evaluate these functions in simulated crash scenarios.
• Ethical Risk Analysis of Autonomous Driving Strategies: Conducta risk-benefit analysis of commonly used AV strategies (e.g. emergencybraking, evasive swerving) from an ethical perspective. Classifyrisks not only by severity, but also by distribution between different groupsof human actors. Use this to design more balanced trajectory planning algorithmsand develop a harm function that integrates the most significantfactors.
• V2X Protocol Design: Investigate driving scenarios where V2X communicationcan improve ethical outcomes. Develop and evaluate a consensusprotocol for ethically aligned decision making based on inter-vehiclecoordination.•
• Reinforcement Learning Design in Longitudinal Control: Autonomousvehicles in longitudinal driving require complex control algorithmsto be more resilient in the face of unknown situations on the road.Existing works use simple metrics, such as distance and speed similaritybetween vehicles, but more complex considerations, such as ethical driving,can be a factor in designing reinforcement learning algorithms. Onlinereinforcement learning during driving can utilize V2X communicationamong vehicles in longitudinal formations.

Interested in one of the topics, or do you have your own ideas that fall into the umbrella topic? Do not hesitate to reach out to us. We offer interdisciplinary collaborative supervision from two leading universities: Technical University of Munich (Campus Heilbronn) and Karlsruhe Institute of Technology.

Contact

Niclas Kannengießer (niclas.kannengiesser(at)tum.de)

Maximilian Schrapel (maximilian.schrapel(at)kit.edu)

Jianxin Zhao (jianxin.zhao(at)kit.edu)

References

[1] Galina Sidorenko, Johan Thunberg, and Alexey Vinel. “Cooperation for Ethical Autonomous Driving”. In: 2024 20th International Conference on Wireless and Mobile Computing, Networking and Communications (WiMob). IEEE. 2024, pp. 391–395.

[2] Galina Sidorenko, Johan Thunberg, and Alexey Vinel. “Ethical V2X: Cooperative Driving as the Only Ethical Path to Multi-Vehicle Safety”. In: 2023 IEEE 98th Vehicular Technology Conference (VTC2023-Fall). 2023, pp. 1–6. doi: 10.1109/VTC2023-Fall60731.2023.10333432.