Law meets engineering for trustworthy Autonomous Vehicles (LawIng4AV)

In autonomous vehicle (AV) decision making, ensuring the conformance of decisions with existing regulation documents is a crucial factor. Without assuring rule conformance, regulators cannot admit an AV onto the roads of this world. Especially in the case of an accident involving an AV, the AV’s decisions must not only be correct from the AVs point of view, but must also be justifiable, e.g. with respect to existing regulation documents. Such a correct and justifiable decision-making is highly dependent on the AVs capability to understand and reason about traffic rules and context-dependent conflict situations. Suppose an AV damages a traffic sign. It hence departed from the road. This may be "justifiable" if it gave way to an ambulance and the road was slippery. A machine-readable version of rules is necessary, to integrate regulation conformance and reasoning into an AV. Justifications for AV decisions must also be made understandable towards diverse stakeholders: regulators must understand system decisions to decide an AVs admission onto our roads and lawyers must decide liabilities in tort claims in court.

We observe a list of challenges to be discussed within this workshop:

• How can the justifiability of AV decision-making algorithms be validated, given the highly uncertain real-world environment the AV navigates in?
• How can we ensure and explain correct functionality towards diverse regulatory stakeholders with different levels of understanding about the system and the environment?
• Existing court decisions can be a source for law-conforming AV behavior in complex maneuvers. How can we encode such knowledge into the AV?
• In recent research, various situation-dependent and global conflicts have been identified between traffic rules, specific traffic situations, and an agent’s goals. How can we identify and analyze these conflicts methodically?
• How can we empower technicians to implement traffic rules? How can autonomous vehicles obey traffic rules given the unreliable perception of the vehicle’s environment?
• AV regulation processes and traffic regulation documents vary greatly around the globe. What is the impact of these differences? How can we encode culturally and locally differing views on regulation and traffic rules into our systems?