Learning Affordance Space in Physical World for Vision-based Robotic Object Manipulation | IEEE Conference Publication | IEEE Xplore