Perception-based control for intelligent systems

Intelligent systems theory tries to study the most amazing feature of living creatures: intelligence. One active research area with many promising applications is autonomous navigation of unmanned vehicles which relies heavily on intelligent systems theory. The purpose of this dissertation is to apply an ambiguous concept in intelligent systems, called perception, in robot navigation. Several approaches have been used to model perception for robot navigation. A learning framework, equipped with a perception-based task control center, has been proposed. A statistical approach for uncertainty modeling has been investigated as well. In addition, a spatial knowledge model was used to model robot navigation. Finally, an optimization approach toward perception was used to model robot design and navigation. Several case studies of robot design will be presented. An unmanned ground vehicle, called the Bearcat Cub, was designed and developed for the Intelligent Ground Vehicle Competition (IGVC). This robot was used to demonstrate spatial knowledge modeling. In another design, a soil sampling survey robot was developed to measure the soil strength in remote areas. And finally, the design and development of a snow accumulation prevention robot will be presented. This autonomous robot can prevent accumulation of snow in areas such as driveways and small parking lots. The implementation of unique hardware and software systems in several robotic systems, as well as promoting a multifaceted view of perception modeling, are significant contributions made by this dissertation. The proposed framework uses optimization approach; it has learning capability, and is able to handle uncertain situations that are common in robot navigation.