Map building through self-organisation for robot navigation

The ability to navigate is arguably the most fundamental competence of any mobile agent, besides the ability to avoid basic environmental hazards (e.g. obstacle avoidance). The simplest method to achieve navigation in mobile robot is to use path integration. However, because this method suffers from drift errors, it is not robust enough for navigation over middle scale and large scale distances. This paper gives an overview of research in mobile robot navigation at Manchester University, using mechanisms of self-organisation (artificial neural networks) to identify perceptual landmarks in the robot's environment, and to use such landmarks for route learning and self-localisation, as well as the quantitative assessment of the performance of such systems.