On Methods for Assistive Mobile Robots

This thesis makes a contribution to the field of autonomous vehicles, especially for use on assistive wheelchairs. Methods and technologies for topological signature maps, obstacle avoidance, door passing, localization, and navigation are presented. One method is called circle sector expansion (CSE), which can be used to generate a skeleton in given two-dimensional data. The method is useful for semi-autonomous driving in an unknown environment. The output can be used for path planning and map building. The thesis also presents a way to find fingerprints, called signatures, in the environment that surrounds a vehicle. The signatures are used to build a topological graph, and to detect when the vehicle re-enters a region. The topological map is a hybrid map since the edges store metric properties and represent locations where the vehicle is able to move. A way to model a doorway for a door passing algorithm is also presented. A Matlab toolbox, used to interface sensors directly into the Matlab environment, is also presented in the thesis. The toolbox is a useful tool for developing and testing new algorithms. It has support for logging and replaying of data. Most of the work was done on our experimental platform; the MICA wheelchair, the Mobile Internet Connected Assistant, which is equipped with different types of sensors and an embedded PC. The wheelchair is connected to the Internet via wireless LAN. The wheelchair can be controlled, and sensor readings can be requested from the onboard mounted sensors over the Internet.