Driver modelling
Supervisor
Suitable for
Abstract
Semi-autonomous driving is intended to improve safety, and involves, for example, taking control of the car from the driver
if a collision is predicted. Probabilistic models of driver behaviour are needed in order to monitor whether the driver is
attentive or distracted, and to predict if the manoeuvre, such as the driver swerving to the neighbouring lane, is safe. Recently,
a driver model was developed and analysed using PRISM (http://www.prismmodelchecker.org/bib-ext.php?search=driver) and validated
against data from the Carsim simulator. This project aims to extend this model in a number of ways, firstly, with more realistic
threats, and, secondly, with active threats; the latter can be achieved using stochastic games and the PRISM-games model checker
(http://www.prismmodelchecker.org/games/). In the longer term, the goal is to validate the model with real traffic data in
collaboration with the Mobile Robotics Group. This project will suit a student interested in modelling.