Testing cognitive mapping ideas on a mobile robot
A recent theory of perceptual mapping argues that humans process spatial information in a different way than previously thought. In particular, the theory suggests a process which, unlike SLAM, does not correct perceptual errors and does not need to integrate successive views when computing the map. Such a process could equally be applied for robot mapping. The purpose of this study is to implement the theory on a mobile robot, and see what maps can be produced. Towards this end, an algorithm based on it has been implemented and tested on a mobile robot equipped with a laser sensor. Its performance has been analyzed in relation to the traditional SLAM approaches, and different experiments have been conducted which cover typical problems of robotic mapping. Additionally, a simple method has been shown which enables a robot to autonomously navigate using the map created. The results obtained showed that the generated maps do indeed preserve a good layout of the environment, thus supporting the integral claim of the theory. It was also found that basic navigation with the produced maps is possible. The principal conclusion is thus that the theory shows much promise and could be used as a foundation for further research of human cognition and to develop new algorithms for robot mapping.