Learning movement primitives for force interaction tasks

Kinesthetic teaching is a promising approach to acquire robot skills in an intuitive way. This paper focuses on learning skills that do not solely rely on kinematics but also need to take into account interaction forces. We present three novel concepts towards learning such force interaction skills. Firstly, we determine segments from a small number of continuous kinesthetic demonstrations using contact information. Secondly, we associate each segment with a movement primitive, and determine its composition, i. e. , the control variables and reference frames that allow to reproduce the demonstrated task. Lastly, we propose a concept to determine the transitions between the primitives during reproduction. The proposed methods are evaluated on a box pulling and flipping task, and show very good generalization abilities for objects with different geometries, and situations with different object arrangements.

Authors

• Jens Kober
• Michael Gienger
• Jochen J. Steil

Keywords

• Force
• Robot sensing systems
• Standards
• Kinematics
• Switches
• Convergence
• Interaction Forces
• Movement Primitives
• Control Variables
• Reference Frame
• Learning Skills
• Object Placement
• Number Of Demonstrations
• Scaling Factor
• Combined Set
• Variety Of Tasks
• Control Mode
• Gravitational Force
• Motion Capture
• Overall Variance
• Object Size
• Pseudo-inverse
• Large Objects
• Force Control
• Coordinate Frame
• Linear Sequence
• Movement Sequences
• Dynamic Time Warping
• Segmentation Points
• World Coordinate
• Cylindrical Coordinates
• Contact Events
• Velocity Limits
• End Of Segment