30hp – Modelling Safety of Autonomous Driving using Stochastic Processes i Sodertalje

Scania genomgår nu en transformation från att vara en leverantör av lastbilar, bussar och motorer till en leverantör av kompletta och hållbara transportlösningar.

Thesis project at Scania is an excellent way of making contacts for your future working life. Many of our current employees started their career with a thesis project.

In experimental projects around the world, autonomous driving has been demonstrated for different kind of vehicles and scenarios. Scania is among the companies aiming to sell fully autonomous vehicles a few years from now. However, to guarantee safety of the autonomous driving is still a huge challenge. Notably, this is also indicated by the fact that all of the reported experiments have had a human driver present ready to take over in a situation when autonomous driving fails.

Clearly, safety of autonomous driving in general and in particular for Level 3 needs deeper analysis. In this project, autonomous driving is to be analysed using semi-Markov processes. The reason to use semi-Markov processes instead of just Markov processes is that the former is more general since it allows transition probabilities that are dependent on time spent in a state; something that is useful when modelling vehicle and driver behaviour in the context considered. By using the models, the aim is to answer questions such as: how does reliability and response time of individual systems affect the overall safety? In a situation where driver actions are needed, how does driver alertness affect overall safety? How does the accuracy of a neural-network based perception algorithm affect overall safety?

Furthermore, it should be investigated how the current methods of the functional safety in the standard ISO26262 can be modified and complemented using semi-Markov process models.

– Extend a tool within MATLAB Simulink/Stateflow for semi-Markov modelling and analysis, to support scenario-modelling related to autonomous driving.

– Use the tool to model different safety concepts of autonomous driving. Draw conclusions on the properties of different safety concepts.

– Propose how to modify the ISO26262-method to utilize semi-Markov modelling of safety concepts.

Specify education or specialization: M.Sc. in Engineering or similar with an interest for mathematics.

Number of students: 1-2.

Start date: January/February 2019.

Estimated time needed: 6 months.

Carl Bondesson, system designer, 08-553 70 125, .

Mattias Nyberg, technical manager, 08-553 83 736, .