Are you excited about a future where robo-taxis are easily accessible to all, especially the elderly and the mobility challenged? Would you like to help make that future a reality? Join us while we create the first “driving competence exam” for automated vehicles!
InformationFor the near future, driving will remain a form of social interaction because it is performed by people. When we join traffic, we have a clear understanding of the social norms of driving. We know what to expect from other road users and what they expect from us, and we can intuitively recognize who around us drives according to these expectations. Moreover, safety in traffic is the result of interactions among drivers who have the knowledge and training to drive competently, and that are licensed to do so.
But now that more self-driving cars are joining traffic (e.g., Teslas with Full Self-Driving capabilities, or Waymos) the questions arise: how do we know these vehicles can indeed drive competently among us? Can their understanding of our complex social interactions be assessed or measured? What is indeed competent driving and how can we assess it?
Here is where you come in! Within the EU project CERTAIN (Resilient and continuous safety assurance methodology for CCAM and its HMI components) you will join the Dynamics and Control section at the
Mechanical Engineering Department and help create and implement an objective, scalable, data-based, methodology to assess driving competence. That is, you will:
- Conduct a detailed literature analysis on objective assessment of driving competence.
- Formalize driving behaviors in mathematical relationships among kinematic and environmental variables, together with traffic rules and infrastructure information.
- Process naturalistic driving data to determine how these relationships are maintained by competent human drivers.
- Develop a method to assess whether a drivers (human or otherwise) drive competently by comparing their driving data to that of competent drivers.
- In close collaboration with TNO and several of the other 23 international CERTAIN partners, implement the assessment methodology in TNO’s Carlabs for demonstration in one of the key use cases of the CERTAIN project.
There are several potential strategies that could be used for assessment. One is the well-established “scenario-based testing” methodology, in which a vehicle under test is offered, via simulations, multiple instances of different driving scenarios. The responses to these scenarios are then used for assessment. Another is a more system-theoretic approach would be to fit a model to the self-driving algorithm of the vehicle under test, which is then compared to a reference model of a competent driver. You, as a PhD candidate, will have the freedom to explore these and other options and provide a theoretical foundation for the strategy you select.