Consider generating manipulation actions by a robot. In this case, the robot grasps the target object while avoiding interference with the surrounding environment or sometimes, allowing this interference. In other words, the interaction with the surrounding environment must be considered when grasping the object and machine learning can be used to ensure successful grasping while considering such interactions. In the figure in the left, Deep Learning is used to predict the success or failure of grasping for the problem of picking parts stacked in boxes, which is often assumed to happen in the manufacturing industry.