Sensor Fusion, Navigation, and Control of Autonomous Vehicles

Abstract

The development of completely autonomous mobile vehicles has been the topic of a great deal of research over the past few decades. Spurred by interests as diverse as space exploration and land mine removal, research has focused on the mechanical requirements, sensing and computational requirements, and intelligence required for autonomous decision making.

This thesis focuses on developing the software required for autonomous control, while building upon previous research into appropriate mechanical designs and sensing technologies. The thesis begins by giving an overview of the problem, and then moves on to reviewing the literature relevant to the task of fusing diverse, and often conflicting, sensor data into a usable representation. Literature relevant to the task of using that data to make intelligent decisions in an autonomous manner is reviewed. The focus then shifts to developing a working platform, called Navigator, which tests the theory in the setting of the Intelligent Ground Vehicle Competition. The theory required to control Navigator, along with the dynamic analysis used for controls testing, is developed. Image processing techniques useful for extracting features from the course are discussed, and the required mathematical relationships are derived. The thesis then discusses modifications to the Vector Field Histogram technique, which enable Navigator to fuse data from both the image processing and laser rangefinder. Development of the navigation decision-making algorithm is discussed. The information in this thesis is presented in such a way that it can serve as a reference to those who follow in the task of building autonomous vehicles.