Russell Toris and Sonia Chernova Worcester Polytechnic Institute, - - PowerPoint PPT Presentation

▶
russell toris and sonia chernova
SMART_READER_LITE
LIVE PREVIEW

Russell Toris and Sonia Chernova Worcester Polytechnic Institute, - - PowerPoint PPT Presentation

Aug 09, 2023 115 likes •262 views

Learning of Multi-Hypothesized Task Templates from a Corpus of Noisy Human Demonstrations Russell Toris and Sonia Chernova Worcester Polytechnic Institute, Worcester, MA, USA Motivation Traditional LfD focuses on teaching sequences

slide-1
SLIDE 1

Learning of Multi-Hypothesized Task Templates from a Corpus of Noisy Human Demonstrations

Russell Toris and Sonia Chernova

Worcester Polytechnic Institute, Worcester, MA, USA

slide-2
SLIDE 2

Motivation

  • Traditional LfD focuses on teaching sequences

○ Trajectory learning ○ Keyframe learning

  • Pick-and-place tasks concerned with end-state

○ Pack the items in the box ○ Put away the pallet ○ ...

  • What’s missing: plausible goal states
slide-3
SLIDE 3

The Problem

  • How can we learn goal states based on

high-level task descriptions?

  • Different interpretations lead to different goal

states

slide-4
SLIDE 4

Challenges

  • Multiple plausible goal states possible
  • Learning should be done in an unsupervised

manner

  • Frame of reference unknown
  • Requires large training set

○ Crowdsourcing ○ Noisy or malicious data

slide-5
SLIDE 5

Methodology

  • Data Collection

○ Raw placement locations of items per task

  • Transformation

○ Autonomously extract features and eliminate irrelevant ones

  • Model Training

○ Training templates (states) from transformed data

slide-6
SLIDE 6

The Interactive World

  • Browser based
  • Micro-task markets
  • Intuitive interface
slide-7
SLIDE 7

Domain

World Surfaces Items Points of Interest (POIs)

slide-8
SLIDE 8

Data Collection

“Your job is to place the given object in an appropriate location in the house.”
slide-9
SLIDE 9

Data Transformation

...

Cup in Table Space Plate in Counter Space Fork in Plate Space Plate in Table Space Spoon in Plate Space
slide-10
SLIDE 10

Model Training

  • Each transformed dataset becomes potential model λ∈Λ
  • Search for and rank dense clusters
slide-11
SLIDE 11

Tests and Verification

Multi-Item Template Single Item Placement Multi-Item Random
slide-12
SLIDE 12

Conclusion & Future Work

  • Verified on a series of tasks
  • Test in manufacturing and industrial domains
  • Creates a base knowledge base
  • Room for further, local expert user

collaboration

  • Modify models to personalize the template

○ Human-in-the-loop

slide-13
SLIDE 13

Q&A