抄録
Visual recognition in multi-robot systems has a peculiar problem in that an observation from a different viewpoint gives a different perspective. Also, a representation of the same target object is highly affected by such a viewpoint or environmental condition. Hence it has become a challenge to realize cognitive sharing of the object or ROI between the robots without a help of artificial markers. To cope with these issues, we have proposed a concept of Hierarchical Invariants Perception Model where multiple representations are dynamically evaluated and selected under the situation of a robot. In this paper, we propose consensus-making algorithms to acquire the viewpoint invariant representation of the geometrical relations. Experimental results show the ability of cognitive sharing significantly improved by proposed method.
