Previously, the Institute of Mechatronic Systems developed an autonomously parking vehicle with fuzzy control, based upon a MCC Smart microcar.

The Smart was augmented through additional hardware. A realtime Java system with a CAN-Bus interface serves as the controller, operating actuators for the steering column and the brakes and several additional vehicle functions. A differential GPS (DGPS) system delivers the coordinates to aid in navigation.

The goal of this project is to develop an autonomous vehicle which is able to follow a set path safely without foreign aid. In previous works the authors have developed concepts for car-based autonomous navigation and have implemented a simple algorithm. Furthermore concepts for equipping the vehicle with obstacle detection possiblities were evaluated and a simulation software was written. During the course of this thesis a DGPS-aided navigational algorithm was developed which follows a preset path. While driving a path, both the direction and the speed of the vehicle are constantly adjusted to suit the driving conditions.

Using this algorithm, a different approach to autonomously parking a vehicle was implemented. A path leading into a known parking field is calculated which then can be followed in an analogous manner to normal navigation. To be able to detect parking spaces automatically, a radar sensor from A.D.C. GmbH was added. With this data the controller of the vehicle determines the dimensions and position of the lot.

The highly precise DGPS system employed in this vehicle proved to be very sensitive to loss of signal from the satellites. To solve this problem, a Kalman filter was implemented to perform the task of sensor fusion between the DGPS system and the wheel-driven odometry. Additionally a serial sensor fusion strategy was implemented, so that the performance of the Kalman filter could be tested. A Java-RMI application was developed to remotely control the vehicle?s controller.