Robot Manipulation and Learning

Robot manipulation and learning sits at the intersection of control theory and machine intelligence, asking how a robot can reliably pick up, reorient, and work alongside humans in environments that were never precisely engineered for it. At its core, the work spans teaching robots new motions from human demonstrations, estimating where and how objects sit in three-dimensional space, and designing controllers that let a robot's arm yield safely when it meets unexpected resistance or contact. Deep learning has sharpened perception considerably, but translating visual understanding into dexterous, force-aware action remains an open challenge, particularly when objects are partially hidden or when the robot must hand something directly to a person. Active research directions include making learned skills transfer across different objects and settings without retraining from scratch, and building collaborative behaviors that are both efficient and provably safe for the humans working beside the machine.

Works

63,112

Total citations

690,883

Keywords

Robot LearningGraspingDeep LearningHuman-Robot CollaborationObject Pose EstimationDynamical Movement Primitives