GMR based forcing term learning for DMPs
| Autor: | Li Ning, Sujuan Wei, Kui Xiang, Jian Fu |
|---|---|
| Rok vydání: | 2015 |
| Předmět: |
Forcing (recursion theory)
business.industry Mixture model Machine learning computer.software_genre Regression Term (time) symbols.namesake Distribution (mathematics) symbols Artificial intelligence business Unit-weighted regression Gaussian network model Algorithm computer Mathematics Parametric statistics |
| Zdroj: | 2015 Chinese Automation Congress (CAC). |
| Popis: | Dynamic movement primitives (DMPs) is very powerful model to conduct learning from demonstration for robot. In this paper, we put forward a method for forcing term learning based on Gaussian Model Regression (GMR). Specifically, we apply the Gaussian Mixture Model (GMM) to model the jointly probability over data from demonstrations (desired values, positions and velocities from canonical system). Thus we can obtain the generalized prediction by means of the corresponding conditional distribution. The proposed the method has a more fitting precision than LWR (Local weighted Regression) which is a classical regression technique in DMPs. Simulation results on trajectory planning with min-jerk criterion demonstrate the effect and efficient. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|