RAW (Robot Auto and Wegenwet)

Robot cars are going to enter the market soon, developments with respect to the car itself are in full swing. Subsequently, to deploy these cars on the road, development within the infrastructure and legislation is also of equal importance. In this project, HAN-AR investigates on one of the challenges concerning how the autonomous cars can "know" the Dutch legislation (Traffic) and apply them in practice.

The machine learning based automotive software is, in fact, a 'black box': there is no explicit and transparent link between the software and the laws and regulations (which consists of legislation, case law and other interpretations). It is not to determine whether and how a robot car knows the traffic rules, for example, when it may be to some extent violated to ensure road safety (cross solid white stripe, exceed maximum speed). This has direct consequences on the certification of autonomous cars: when can we say that they earn a driver's license? One must also, for example after an accident, for legal and safety reasons, may request a car: "Why did you do so?"

Background

The arrival of the robot car or self-driving cars on public roads comes with challenges and questions to be answered, but the fact that it is going to happen seems to be inevitable. Robot Cars promise a revolution in the mobility domain. The self-driving car will ensure fewer accidents and ideally less congestion. It also promises mobility for everyone, with and without a license. To achieve that many researches have been initiated with respect to the level of technology; software, sensors, actuators. Besides the development of the cars themselves, there are many infrastructural and social issues that must be considered. In the legality perspective, it not only requires new legislation but also the question of whether and how robotic vehicles abide by the traffic laws needs to be answered, i.e. a 'license' (certification). Certification of car and driver in the context of robotic vehicles is the subject of an exploratory running in the project PRAUTOCOL. The proposed project RAW examines how the Highway Code can be used by the control software of robot cars. It is a first phase (for the study and simulation) as a preparation which can be modeled on a follow-up project in which large parts of the road traffic regulations will be tested in / with a prototype robot car.

Final Outcomes

  • Concise report literature, context and potential effects.
  • Draft legislative models (some models WVW94, including data models) focusing on two specific cases i in the highway pilot domain, showing how such legal models look and can be achieved.
  • Demo simulations showing the effect of different types of legal modeling on two specific use case situations in the field of highway piloting.
  • Final evaluations of legal models and simulations, including the connection of law models on simulations.
  • Recommendations, general, and related specifically to the project partners.
  • Further development and implementation, problem delimitation, boundary conditions and steering in order to further project proposals.

Current Status

  1. Development of scenarios under progress.
  2. Development of decision making architecture on top of the black box model used in self-driving cars.
  3. Investigation on connecting parts of law models with simulation models of the vehicle.