Intelligent Tutoring System for a Driving Simulator
Sandro Ropelato
Master's Thesis, July 2017
Supervisors: Dr. Fabio Zünd, Dr. Stéphane Magnenat, Prof. Dr. Bob Sumner
Abstract
We show how an intelligent tutoring system (ITS) can be used to assist the process of learning how to drive a car. On a virtual reality driving simulator, we create a detailed model of a city, including artificial intelligence traffic, interacting with the driver. In a mock-up of a car cockpit, the user operates the vehicle through the virtual environment by controlling a steering wheel, pedals and a gear lever.
Using a head-mounted display, the driver observes the scene from within the car. The real- ism of the simulation is supported by the use of a motion system capable of simulating forces experienced when accelerating, braking or turning the car.
During the run on the simulator, the quality of driving is permanently assessed by the system. In order to achieve fast learning progress, a navigation system guides the driver along a path where he or she has to complete car driving maneuvers. The sequence in which these maneuvers occur is dynamically adapted based on the user’s performance. Since the driving simulator offers a safe way of training car driving tasks, dangerous situations can be practised in a controlled environment.
At last, we present the findings of a user study examining the impact of the ITS on the learning progress and evaluating the quality of the driving simulator. The system is assessed with standardized questionnaires used to rate the realism of virtual reality applications.