Research on online modeling of multiple demonstration trajectories for robots based on composite movement primitives
ObjectiveAiming at the problems of low parameter estimation efficiency and lack of multi-trajectory modeling ability in traditional robot demonstration learning algorithm, an online modeling method of robot multiple demonstration trajectories based on composite dynamical movement primitives was prop...
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | Chinese |
| Published: |
Editorial Office of Journal of Mechanical Transmission
2025-07-01
|
| Series: | Jixie chuandong |
| Subjects: | |
| Online Access: | http://www.jxcd.net.cn/thesisDetails#10.16578/j.issn.1004.2539.2025.07.019 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | ObjectiveAiming at the problems of low parameter estimation efficiency and lack of multi-trajectory modeling ability in traditional robot demonstration learning algorithm, an online modeling method of robot multiple demonstration trajectories based on composite dynamical movement primitives was proposed.MethodsGaussian mixture model was used to model multi-demonstration trajectory in the composite algorithm. Through dynamical movement primitives, the regression trajectory of the Gaussian mixture model was learned which enhanced the multi-trajectory modeling capability. To improve the efficiency of parameter estimation of the model, an online parameter estimation algorithm based on Welford’s formula and path integral was designed to calculate model parameters incrementally during the trajectory sampling process, and the parameter estimation could be completed at the end of the demonstration. Finally, a six-degree-of-freedom industrial robot was used as the object to design and complete the learning experiment of the depalletizing trajectory demonstration.ResultsThe results show that the parameter estimation time of the proposed algorithm is only 0.033 ms. The demonstration learning algorithm based on improved dynamic motion primitives is efficient in parameter estimation and has the ability of fast multi-trajectory modeling and generalization. |
|---|---|
| ISSN: | 1004-2539 |