This thesis discusses how the problem of parking a passenger car can be solved. There ane numerous obstacles when creating a fully or semi automated system for assisting the parking maneuver. The obstacles range from unobtrusive Man Machine Interface issues to robust algorithms for finding features in sensor data.
Such a system will also have to cope with ever changing environments, most of which will not be discovered during the design phase. Misuse will also be common, since the function is aimed at a mass market - the car buyers. These interesting problem areas is also coupled with the customer value. A car buying customer today aims at feature content and comfort. So for a parking system to survive it needs to be extremely intuitive; the customer value must be discovered during a short test drive with the vehicle. A learning period before the user can operate the system is not feasible.
The work presented here proposes an interesting algorithm for finding parking space features in sensor data collected using ultrasonics. The algorithm, which is based on the well known Hough transform have been proved to be robust in real world experiments.
Furthermore it is also investigated how critical the mount ing of the environment sensing subsystem is. Where should the sensors be placed and where should they be facing to maximize the performance.
Emphasis have also been put on designing a functional architecture that fits in the car development process of today. The implementation of a system in a car must fit in the platform and component reuse is critical to keep cost down.
To test theories developed during the work a prototype car has been used. The prototype is essential to analyze the robustness of algorithms in different parking scenarios. It has also been an invaluable tool when conducting customer surveys to find where the customer value of these kind of systems is.
Linköping: Linköpings universitet , 2007. , 42 p.