Adapting Object-Centric Probabilistic Movement Primitives with Residual Reinforcement Learning.
João CarvalhoDorothea KoertMarek DanivJan PetersPublished in: Humanoids (2022)
Keyphrases
- reinforcement learning
- object model
- d objects
- complex objects
- data objects
- generative model
- state space
- optimal control
- function approximation
- uncertain data
- building blocks
- probabilistic model
- optimal policy
- low level
- machine learning
- high level
- visual patterns
- user centric
- complex scenes
- deformable objects
- target object
- body movements
- geometric objects
- object segmentation
- spatial relations
- object tracking
- keypoints
- dynamic programming
- data model
- learning process
- object recognition