| Autor: |
Amélie Legeleux, Cédric Buche, Dominique Duhaut |
| Rok vydání: |
2022 |
| Zdroj: |
The International FLAIRS Conference Proceedings. 35 |
| ISSN: |
2334-0762 |
| DOI: |
10.32473/flairs.v35i.130559 |
| Popis: |
Cobots are robots specialized in collaborating with a human to do a task. These cobots needs to be easily re-programmed in order to adapt to a new task. Learning by Demonstration enables a non-expert user to program a cobot by demonstrating how to realize the task. Once the learning is done, the user can only improve the learning by adding new demonstrations or deleting existing ones. In this article, the proposed model gives the possibility to the user to impact the learning by choosing which parts of the demonstration has more importance. This model uses an extended version of Gaussian Mixture Model (GMM) with weighted data coupled with Gaussian Mixture Regression (GMR). This architecture was tested with two different tasks and with two robots. Results indicate better generated trajectory with the proposed approach. |
| Databáze: |
OpenAIRE |
| Externí odkaz: |
|
