Queißer, Jeffrey: Multi-modal Skill Memories for Online Learning of Interactive Robot Movement Generation. 2018
Inhalt
- List of Figures
- List of Tables
- Introduction
- Motivation
- Problem Statement
- Related References of the Author
- Funding Acknowledgements
- Organization of the Dissertation
- Skill Represenation & Skill Learning
- Background: From Theories of Motor Control to High-Level Skill Learning
- A Novel Conceptual Framework for Parameterized Skill Learning
- Parameterized Skills for Kinematic Representations
- Task-Parameterized Skills
- Bootstrapping of Parameterized Skills
- Regularization of Reward
- Experimental Evaluation of Bootstrapping
- Discussion
- Efficient Exploration of Parameterized Skills
- Parameterized Skills for Compliant & Soft Robots
- Parameterized Skills for Control of Complex Robots
- Primitive Based Dynamics Representation
- Parameterized Skills for Dynamic Action Primitives
- Evaluation of the Dynamics Representation
- Interaction in Dynamic Environments by Integration of Kinematics and Dynamics
- Discussion
- Discussion & Conclusion
- Appendix
- Parameter Grid Seach for InverseEquilibrium Model of the Affetto Robot
- Parameter Grid Seach for InverseEquilibrium Model of the UR5 Robot
- Optimization of Human Demonstrations
- Example Task Instances of the Drumming Scenario
- Prototype Spectra of Human Demonstrations
- Interactive Scenario: Joint Angle Trajectories
- Interactive Scenario: Optimized Forward Signals
- Interactive Scenario: Sucessful Generalizations
- References