Procedural Generation of Articulated Simulation-Ready Assets

TL;DR

This paper presents Infinigen-Articulated, a toolkit for procedurally generating realistic articulated assets for robotics simulation, enabling improved training and transfer of policies in simulated environments.

Contribution

It introduces a comprehensive toolkit with generators for 18 object categories, high-level utilities, and an export pipeline for seamless integration into robotics simulators.

Findings

Assets improve movable object segmentation accuracy.

Assets enable training of generalizable reinforcement learning policies.

Assets facilitate effective sim-to-real transfer in imitation learning.

Abstract

We introduce Infinigen-Articulated, a toolkit for generating realistic, procedurally generated articulated assets for robotics simulation. We include procedural generators for 18 common articulated object categories along with high-level utilities for use creating custom articulated assets in Blender. We also provide an export pipeline to integrate the resulting assets along with their physical properties into common robotics simulators. Experiments demonstrate that assets sampled from these generators are effective for movable object segmentation, training generalizable reinforcement learning policies, and sim-to-real transfer of imitation learning policies.