Understanding URDF: A Dataset and Analysis

TL;DR

This paper introduces a comprehensive dataset of 322 URDF files from various sources, analyzes their characteristics, and provides insights into their generation and usage to support robotics research and development.

Contribution

It presents the first dataset of URDF files with metadata, along with analysis tools and insights into URDF usage across different robots and sources.

Findings

Most common mesh file types are used in URDFs

Multiple definitions of the same robot model exist across sources

Insights into how URDFs are generated and utilized

Abstract

As the complexity of robot systems increases, it becomes more effective to simulate them before deployment. To do this, a model of the robot's kinematics or dynamics is required, and the most commonly used format is the Unified Robot Description Format (URDF). This article presents, to our knowledge, the first dataset of URDF files from various industrial and research organizations, with metadata describing each robot, its type, manufacturer, and the source of the model. The dataset contains 322 URDF files of which 195 are unique robot models, meaning the excess URDFs are either of a robot that is multiply defined across sources or URDF variants of the same robot. We analyze the files in the dataset, where we, among other things, provide information on how they were generated, which mesh file types are most commonly used, and compare models of multiply defined robots. The intention of this article is to build a foundation of knowledge on URDF and how it is used based on publicly available URDF files. Publishing the dataset, analysis, and the scripts and tools used enables others using, researching or developing URDFs to easily access this data and use it in their own work.