Autonomous acquisition of multimodal information for online object concept formation by robots.