Abstract
We propose a novel system to represent and estimate topological relations between objects in a dynamic environment. The system can represent topological relations between objects extracting specific features of the co-occurrence matrix obtained from disparity images with the stereo vision system. Furthermore, to infer time-invariant relations between objects, the habituation concept was employed, and the habituation value was estimated at every occurrence of the event triggered by the motion vector. The proposed system enables the robot to acquire spatial situation awareness autonomously and make a decision in an unknown environment.
