Note: This content is accessible to all versions of every browser. However, this browser does not seem to support current Web standards, preventing the display of our site's design details.

  

Using diffeomorphic matching for robot learning from demonstration

Author(s):

Q. Censier
Conference/Journal:

Master Thesis, FS16
Abstract:

In this Master thesis, I present a new robot Learning from Demonstration (LfD) approach based on a diffeomorphic matching algorithm. From a single demonstrated trajectory, the proposed method extrapolates a velocity vector field that generalizes the demonstration into a skill. Used in a controller, it results in a behavior that can reproduce the demonstration with appropriate responses to perturbations or changes in the initial configuration. The method is easy to implement and relies on very few parameters. It is also efficient and has good mathematical properties (the learned vector fields are guaranteed to be globally asymptotically stable), which makes it a possible alternative to existing LfD techniques. I show experimental results for grasping tasks on two different robot arms: a 6-Degrees of Freedom (DoF) CrustCrawler arm, and a 7-DoF arm of the robot Baxter (Rethink Robotics).

Supervisors: Nicolas Perrin, Stephane Doncieux, Florian Dörfler

Year:

2017
Type of Publication:

(12)Diploma/Master Thesis
Supervisor:



File Download:

Request a copy of this publication.
(Uses JavaScript)
% Autogenerated BibTeX entry
@PhdThesis { Xxx:2017:IFA_5636
}
Permanent link