Opening Articulated Objects in the Real World

📅 2024-02-27
📈 Citations: 2
Influential: 0
📄 PDF
🤖 AI Summary
This paper addresses the problem of end-to-end autonomous opening of unknown articulated objects (e.g., doors, drawers) by mobile manipulators in unstructured environments without prior knowledge. To tackle the full pipeline challenges—approach, perception, planning, and manipulation—we propose a modular system architecture integrating RGB-D visual perception, articulated object model estimation, and optimization-based trajectory generation, deliberately avoiding end-to-end learning. Key findings reveal that perception—particularly accurate estimation of articulation parameters—is the primary bottleneck; off-the-shelf perception models trained in isolation fail to generalize under robot-centric viewpoints. Modular design significantly outperforms end-to-end alternatives in robustness and performance. We evaluate the system across 13 real-world environments, achieving over 100 successful end-to-end trials on previously unseen articulated objects. All code, trained models, and experimental videos are publicly released.

Technology Category

Application Category

📝 Abstract
What does it take to build mobile manipulation systems that can competently operate on previously unseen objects in previously unseen environments? This work answers this question using opening of articulated objects as a mobile manipulation testbed. Specifically, our focus is on the end-to-end performance on this task without any privileged information, i.e. the robot starts at a location with the novel target articulated object in view, and has to approach the object and successfully open it. We first develop a system for this task, and then conduct 100+ end-to-end system tests across 13 real world test sites. Our large-scale study reveals a number of surprising findings: a) modular systems outperform end-to-end learned systems for this task, even when the end-to-end learned systems are trained on 1000+ demonstrations, b) perception, and not precise end-effector control, is the primary bottleneck to task success, and c) state-of-the-art articulation parameter estimation models developed in isolation struggle when faced with robot-centric viewpoints. Overall, our findings highlight the limitations of developing components of the pipeline in isolation and underscore the need for system-level research, providing a pragmatic roadmap for building generalizable mobile manipulation systems. Videos, code, and models are available on the project website: https://arjung128.github.io/opening-articulated-objects/
Problem

Research questions and friction points this paper is trying to address.

Mobile manipulation systems
Unseen objects and environments
Opening articulated objects
Innovation

Methods, ideas, or system contributions that make the work stand out.

Modular systems outperform learned systems
Perception is the primary task bottleneck
System-level research enhances mobile manipulation
💼 Related Jobs
No related jobs found.
University of Illinois at Urbana-Champaign
A
Arjun Gupta
University of Illinois at Urbana-Champaign
M
Michelle Zhang
University of Illinois at Urbana-Champaign
R
Rishik Sathua
University of Illinois at Urbana-Champaign
S
Saurabh Gupta
University of Illinois at Urbana-Champaign