Research Analyzer
← Back ICRA 2026

Placeit! a Framework for Learning Robot Object Placement Skills

Amina Ferrad, Johann Huber, François Hélénon, Julien Gleyze, Mahdi Khoramshahi, Stéphane Doncieux

PDF

AI summary

Key figure (auto-extracted from paper)
Placeit! leverages quality-diversity optimization to automatically generate diverse, stable object placement poses, achieving 90% real-world pick-and-place success.
Quality-Diversity Robot Manipulation Object Placement Evolutionary Computation Sim-to-Real Data Generation

Problem

Acquiring large-scale, high-quality data for robot manipulation skills like object placement remains slow and laborious, while existing automatic methods rely on hand-crafted priors or fail to ensure physical stability across diverse scenarios.

Approach

Placeit! uses quality-diversity evolutionary algorithms to automatically explore the space of possible object poses in simulation, filtering them with domain randomization to archive only diverse and physically stable placements.

Key results

  • Introduction of Placeit!, a versatile QD framework for automatic placement pose generation
  • QD optimization significantly outperforms state-of-the-art baselines across multiple simulation scenarios
  • Development of QDGP, a pick-and-place pipeline leveraging generated poses
  • 90% success rate over 120 real-world deployments using domain randomization

Why it matters

Enables scalable, automated data generation for robotic foundation models and robust real-world manipulation tasks.

Abstract

Robotics research has made significant strides in learning, yet mastering basic skills like object placement remains a fundamental challenge. A key bottleneck is the acqui- sition of large-scale, high-quality data, which is often a manual and laborious process. Inspired by Graspit!, a foundational work that used simulation to automatically generate dexterous grasp poses, we introduce Placeit!, an evolutionary-computation framework for generating valid placement positions for rigid objects. Placeit! is highly versatile, supporting tasks from placing objects on tables to stacking and inserting them. Our experiments show that by leveraging quality-diversity optimization, Placeit! significantly outperforms state-of-the-art methods across all scenarios for generating diverse valid poses. A pick&place pipeline built on our framework achieved a 90% success rate over 120 real-world deployments. This work positions Placeit! as a powerful tool for open-environment pick- and-place tasks and as a valuable engine for generating the data needed to train simulation-based foundation models in robotics.

Index terms

Manipulation Planning Evolutionary Robotics Data Sets for Robot Learning

Related papers