Learning to sequence movement primitives from demonstrations

We present an approach for learning sequential robot skills through kinesthetic teaching. The demonstrations are represented by a sequence graph. Finding the transitions between consecutive basic movements is treated as classification problem where both Support Vector Machines and Gaussian Mixture Models are evaluated as classifiers. We show how the observed primitive order of all demonstrations can help to improve the movement reproduction by restricting the classification outcome to the currently executed primitive and its possible successors in the graph. The approach is validated with an experiment in which a 7-DOF Barrett WAM robot learns to unscrew a light bulb.

Authors

• Simon Manschitz
• Jens Kober
• Michael Gienger
• Jan Peters 0001

Keywords

• Switches
• Robot sensing systems
• Merging
• Vectors
• Training
• Movement Primitives
• Support Vector Machine
• Gaussian Mixture Model
• Light Bulb
• Sequence Graph
• Learning Algorithms
• Graphical Representation
• Recurrent Neural Network
• Directed Graph
• Nodes In The Graph
• Acyclic Graph
• Output Vector
• State Machine
• Sequence Of Points
• End-effector
• Switching Behavior
• Global Representation
• Toy Example
• Real Robot
• Global Graph
• Local Graph
• Probabilistic Classification
• Hand Velocity
• Imitation Learning
• Bayesian Information Criterion
• Dangerous Behaviors
• Reward Function
• Feature Space
• Dimensionality Reduction