Sim-to-Real Visual Grasping via State Representation Learning Based on Combining Pixel-Level and Feature-Level Domain Adaptation | IEEE Conference Publication | IEEE Xplore