Street Parking for Cars: a Control System for Autonomous Parallel Parking

Parallel parking is a difficult manoeuvre for drivers. The aim of this research was to develop a system to perform this manoeuvre autonomously, requiring no driver input at all. The stages of this development were analysis of the kinematics of vehicles, the design of a simulator and prototype robot to demonstrate parking and the development of parking algorithms to achieve reverse- parallel parking at or above the standard of the learner drivers¿ logbook. Equations were derived to describe the motion of a vehicle. These were tested in a flexible Matlab simulation. A remote control car was modified to be controlled via command line inputs or uploaded control law. A complete and successful set of laws has been documented and used to demonstrate the prototype robot in a parking scenario. The final algorithms were able to successfully park the vehicle utilising only an ultrasonic sensor and encoders on the rear wheels. Calculations and parking are performed in real-time, taking substantially less time than a human driver would. This was only possible due to development of a highly optimised transform to analyse the sensor data.