Constructing Skill Trees for Reinforcement Learning Agents from Demonstration Trajectories

  • Konidaris G
  • Kuindersma S
  • Barto A
 et al. 
  • 140

    Readers

    Mendeley users who have this article in their library.
  • 37

    Citations

    Citations of this article.

Abstract

We introduce CST, an algorithm for constructing skill trees from demonstration trajectories in continuous reinforcement learning domains. CST uses a change- point detection method to segment each trajectory into a skill chain by detecting a change of appropriate abstraction, or that a segment is too complex to model as a single skill. The skill chains from each trajectory are then merged to form a skill tree. We demonstrate that CST constructs an appropriate skill tree that can be further refined through learning in a challenging continuous domain, and that it can be used to segment demonstration trajectories on a mobile manipulator into chains of skills where each skill is assigned an appropriate abstraction.

Author-supplied keywords

  • CST
  • LFD
  • learning from demonstration
  • segmentation
  • skill discovery
  • uBot

Get free article suggestions today

Mendeley saves you time finding and organizing research

Sign up here
Already have an account ?Sign in

Find this document

Authors

  • George Konidaris

  • Scott Kuindersma

  • Andrew Barto

  • Roderic Grupen

Cite this document

Choose a citation style from the tabs below

Save time finding and organizing research with Mendeley

Sign up for free