Vision-based Mobile Robot Navigation: A Qualitative Approach

The goal of mobile robotics research has been to establish effective ways for a mobile robot to navigate from one place to another. This thesis documents an experimental investigation into the use of vision for the navigation of mobile robots. The research presented in this thesis was directed at achieving qualitative navigation in an unmodified, unknown and dynamic environment. This thesis describes the research involved in the development of a number of key vision primitives: a set of high performance real-time vision processes for the support of navigation, and a suite of visual behaviours which fully utilises the vision processes. From the vision primitives we are able to construct a set of basic visual behaviours, Collision avoidance, Obstacle avoidance and Goal seeking. These behaviours allow a mobile robot system to function in a dynamically changing environment. Building from this foundation, a range of abilities were realised, Wandering, Target tracking, Target pursuiting and Foraging. Demonstrating the flexibility and diversity of our approach, a number of applications were realised: Landmark-based navigation, a Soccer playing robot, and a Vacuum cleaning robot. The ability to resolve conflicting situations for an autonomous mobile robot is addressed in this thesis. Addressing such problems enables a mobile robot to achieve autonomously, goal-oriented navigation. Experimental results and an analytical study of our approach is also presented as part of this thesis. In the final part of this thesis we present a paradigm for robot control, Supervised Autonomy. Supervised Autonomy is a framework, which facilitates the development of human robot systems. The components of this framework have been devised in a human-oriented manner, to augment users in accomplishing their task. Experimental results of applying this framework to the use of a mobile robot teleoperation system are presented. The system we have developed makes extensive use of the vision processing components we developed earlier, in producing an intuitive human robot system.