Camera based Navigation by Mobile Robots: Local Visual Feature based Localisation and Mapping

The most important property of a mobile robot is the fact that it is mobile. How to give a robot the skills required to navigate around its environment is therefore an important topic in mobile robotics. Navigation, both for robots and humans, typically involves a map. The map can be used, for example, to estimate a pose based on observations (localisation) or determine a suitable path between to locations. Maps are available nowadays for us humans with few exceptions, however, maps suitable for mobile robots rarely exists. In addition, to relate sensor readings to a map requires that the map content and the observation is compatible, i.e. different robots may require different maps for the same area. This book addresses some of the fundamental problems related to mobile robot navigation (registration, localisation and mapping) using cameras as the primary sensor input. Small salient regions (local visual features) are extracted from each camera image, where each region can be seen as a fingerprint. Many fingerprint matches implicates a high likelihood that they corresponding images originate from a similar location, which is a central property utilised in this work.

Henrik Andreasson received the Master's degree in mechatronics from the Royal Institute of Technology, Stockholm, Sweden, in 2001, and the Ph.D. degree in computer science from Örebro University, Örebro, Sweden, in 2008. He currently holds a post doc position with the Applied Autonomous Sensor System (AASS) Research Center, Örebro, Sweden.