Research Analyzer
← Back ICRA 2026

Efficient Learning of Object Placement with Intra-Category Transfer

Adrian Röfer, Russell Buchanan, Maximilian Argus, Sethu Vijayakumar, Abhinav Valada

PDF

AI summary

Key figure (auto-extracted from paper)
Robots can learn complex object arrangement tasks from as few as five demonstrations and generalize to novel object instances using canonical class mappings.
Few-shot learning object placement intra-category transfer canonical frames robotic manipulation pose inference

Problem

Deep learning methods for robotic manipulation require vast datasets, making sample-efficient skill acquisition difficult. Current approaches also struggle to transfer learned object placements to new instances of the same category without extensive training or explicit language cues.

Approach

The method maps observed objects to a known canonical class frame to learn relative pose distributions, enabling intra-category transfer. It optimizes model complexity to filter out spurious correlations, allowing accurate placement predictions from minimal demonstrations.

Key results

  • Few-shot learning of relative object poses from ≤5 demonstrations
  • Canonical class mapping scheme for intra-category transfer
  • Model complexity optimization to remove spurious distractor correlations
  • Real-robot table setting achieving 73.3% of human baseline quality

Why it matters

It provides a sample-efficient, cue-free framework for robots to learn and generalize complex arrangement tasks, advancing practical autonomous manipulation.

Abstract

Efficient learning from demonstration for long- horizon tasks remains an open challenge in robotics. While significant effort has been directed toward learning trajectories, a recent resurgence of object-centric approaches has demonstrated improved sample efficiency, enabling transferable robotic skills. Such approaches model tasks as a sequence of object poses over time. In this work, we propose a scheme for transferring observed object arrangements to novel object instances by learning these arrangements on canonical class frames. We then employ this scheme to enable a simple yet effective approach for training models from as few as five demonstrations to predict arrangements of a wide range of objects including tableware, cutlery, furniture, and desk spaces. We propose a method for optimizing the learned models to enable efficient learning of tasks such as setting a table or tidying up an office with intra-category transfer, even in the presence of distractors. We present extensive experimental results in simulation and on a real robotic system for table setting which, based on human evaluations, scored 73.3% compared to a human baseline. We make the code and trained models publicly available at https://oplict.cs.uni-freiburg.de.

Index terms

Learning from Demonstration Probabilistic Inference

Related papers