A novel dynamic movement primitives-based skill learning and transfer framework for multi-tool use

Lu, Zhenyu; Wang, Ning; Li, Miao; Yang, Chenguang

Abstract

Dynamic Movement Primitives (DMPs) is a general method for learning skills from demonstrations. Most previous research on DMP has focused on point to point skill learning and training, and the skills learned are usually generalized based on the same tool or manipulator. There is rare research on skill learning and transfer between two or more different tools. For this problem, a new DMP-based skill learning and transfer framework is proposed for the use of multiple tools. It consists of two types of skills: Object Effective (OE) skills and State Switching (SS) skills. OE skills consider the tools' limited forcing areas that can be expressed as constrained inequalities, and extract skills from demonstrations. It can then be generalized along with changes in the shape and range of influence of a new tool. SS skill is used to connect OE skills and implement changes of contact points of the object and tool. Finally, the two skills are integrated and used to realize the transfer of skills from the demonstrated tool to the new tool. An experiment is conducted to verify the effectiveness of the proposed framework, and the procedural solutions and the final manipulation effect are shown in detail.

Citation

Lu, Z., Wang, N., Li, M., & Yang, C. (2022). A novel dynamic movement primitives-based skill learning and transfer framework for multi-tool use. . https://doi.org/10.1109/ICCA54724.2022.9831826

• Conference Name: 2022 IEEE 17th International Conference on Control & Automation (ICCA)
• Conference Location: Naples, Italy
• Start Date: Jun 27, 2022
• End Date: Jul 30, 2022
• Acceptance Date: Apr 6, 2022
• Online Publication Date: Jul 25, 2022
• Publication Date: Jul 25, 2022
• Deposit Date: Aug 15, 2022
• Publicly Available Date: Aug 31, 2022
• Publisher: Institute of Electrical and Electronics Engineers (IEEE)
• Series ISSN: 1948-3457
• ISBN: 9781665495721
• DOI: https://doi.org/10.1109/ICCA54724.2022.9831826
• Keywords: Robotics, Dynamic Movement, Skill Learning, Transfer Framework, Multi-Tool Use, Training, Three-dimensional displays, Automation, Shape, Switches, Trajectory, Task analysis
• Public URL: https://uwe-repository.worktribe.com/output/9840608
• Publisher URL: https://ieeexplore.ieee.org/document/9831826/keywords#keywords