Dynamic Movement Primitives (DMPs) are a framework for learning a trajectory from a demonstration. The trajectory can be learned efficiently after only one demonstration, and it is immediate to adapt it to new goal positions and time duration. Moreover, the trajectory is also robust against perturbations. However, obstacle avoidance for DMPs is still an open problem. In this work, we propose an extension of DMPs to support volumetric obstacle avoidance based on the use of superquadric potentials. We show the advantages of this approach when obstacles have known shape, and we extend it to unknown objects using minimal enclosing ellipsoids. A simulation and experiments with a real robot validate the framework, and we make freely available our implementation.

Dynamic movement primitives: volumetric obstacle avoidance

Michele Ginesi;Daniele Meli;Andrea Calanca;Diego Dall'Alba;Nicola Sansonetto;Paolo Fiorini
2019-01-01

Abstract

Dynamic Movement Primitives (DMPs) are a framework for learning a trajectory from a demonstration. The trajectory can be learned efficiently after only one demonstration, and it is immediate to adapt it to new goal positions and time duration. Moreover, the trajectory is also robust against perturbations. However, obstacle avoidance for DMPs is still an open problem. In this work, we propose an extension of DMPs to support volumetric obstacle avoidance based on the use of superquadric potentials. We show the advantages of this approach when obstacles have known shape, and we extend it to unknown objects using minimal enclosing ellipsoids. A simulation and experiments with a real robot validate the framework, and we make freely available our implementation.
2019
Dynamic movement primitives, obstacle avoidance, superquadric potential, learning from demonstration
File in questo prodotto:
File Dimensione Formato  
root (2).pdf

accesso aperto

Descrizione: Articolo principale
Tipologia: Documento in Pre-print
Licenza: Dominio pubblico
Dimensione 1.21 MB
Formato Adobe PDF
1.21 MB Adobe PDF Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11562/1018548
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 33
  • ???jsp.display-item.citation.isi??? 30
social impact