YUBI: Yielding Universal Bidigital Interface for Bimanual Dexterous Manipulation at Scale
YUBI: Efficient data collection for bimanual manipulation, 8434 hours, improves operational efficiency.
Key Findings
Methodology
YUBI employs a fingertip-driven gripper design combined with a VR system for precise 6 DoF trajectory tracking. This design allows YUBI to naturally map human hand movements to gripper actions, supporting large-scale data collection. The system integrates VR sensors and portable data collection devices, ensuring high-fidelity data acquisition.
Key Results
- YUBI collected data for 119 tasks over 8434 hours and 1.20M episodes, significantly enhancing versatility and operational efficiency in complex bimanual tasks.
- A single policy trained on the YUBI dataset transfers across multiple bimanual robot platforms like UR, Franka, and ELEY, validating the data's executability.
- YUBI improves success rates by 20% in precision tasks compared to UMI and reduces completion time significantly.
Significance
YUBI provides a reproducible path for large-scale data acquisition, addressing ergonomic and usability challenges in existing handheld data collection systems for fine manipulation tasks. By releasing hardware and datasets, YUBI offers the research community a scalable platform for advancing bimanual dexterous manipulation research.
Technical Contribution
YUBI overcomes mechanical offset and tracking accuracy issues of traditional UMI systems through its fingertip-driven gripper design and VR integration. It offers a lightweight, scalable solution for high-precision bimanual data collection and enables cross-platform policy transfer.
Novelty
YUBI is the first to implement a fingertip-driven gripper design, eliminating mechanical offset issues of traditional handheld devices. Compared to existing UMI systems, YUBI significantly improves ergonomics and operational flexibility.
Limitations
- YUBI may face challenges in handling extremely complex industrial tasks, especially those requiring high loads.
- The system relies on VR equipment, which may be limited in certain environments.
Future Work
Future work could include optimizing YUBI's design for higher load tasks, exploring more application scenarios, and further automating data collection processes.
AI Executive Summary
YUBI is an innovative fingertip-aligned gripper designed to address ergonomic and usability challenges in existing handheld data collection systems for fine manipulation tasks. By employing a fingertip-driven design, YUBI naturally maps human hand movements to gripper actions, combined with a VR system for precise 6 DoF trajectory tracking.
The YUBI system integrates VR sensors and portable data collection devices, ensuring high-fidelity data acquisition. Experiments show that YUBI outperforms the UMI gripper in versatility and operational efficiency for complex bimanual tasks. A single policy trained on the YUBI dataset transfers across multiple bimanual robot platforms like UR, Franka, and ELEY, validating the data's executability.
YUBI provides a reproducible path for large-scale data acquisition, addressing bottlenecks in existing handheld data collection systems. By releasing hardware and datasets, YUBI offers the research community a scalable platform for advancing bimanual dexterous manipulation research. Future work will include optimizing YUBI's design for higher load tasks, exploring more application scenarios, and further automating data collection processes.
Deep Analysis
Background
Recent advancements in robotics heavily rely on large-scale datasets, particularly in bimanual dexterous manipulation. While traditional UMI systems offer a low-cost data collection pathway, their bulky designs limit applications in fine manipulation tasks. YUBI aims to address these challenges by providing a more efficient data collection method through innovative design.
Core Problem
Existing handheld data collection systems face ergonomic and usability issues in fine manipulation tasks. Bulky designs and mechanical offsets limit operational flexibility and precision, especially in high-precision and high-load tasks.
Innovation
YUBI's core innovation lies in its fingertip-driven gripper design, eliminating mechanical offset issues of traditional handheld devices. Combined with a VR system, YUBI achieves precise 6 DoF trajectory tracking, supporting large-scale data collection.
Methodology
- �� YUBI employs a fingertip-driven gripper design, directly mapping human hand movements to gripper actions.
- �� Integrates VR sensors for precise 6 DoF trajectory tracking.
- �� Supports portable and stationary data collection, adapting to different operational environments.
- �� Open hardware and datasets support reproducibility and scalability for the research community.
Experiments
The experimental design includes data collection across 22 desks, involving 8434 hours and 1.20M episodes. VR systems are used for high-precision trajectory tracking, and YUBI's versatility and operational efficiency are validated through various tasks.
Results
YUBI outperforms the UMI gripper in versatility and operational efficiency for complex bimanual tasks, with a 20% improvement in success rates. A single policy trained on the YUBI dataset transfers across multiple bimanual robot platforms, validating the data's executability.
Applications
YUBI can be applied in industrial and everyday scenarios for bimanual dexterous manipulation tasks, supporting high-precision data collection and policy transfer, advancing robotic foundation models.
Limitations & Outlook
YUBI may face challenges in handling extremely complex industrial tasks, especially those requiring high loads. The system relies on VR equipment, which may be limited in certain environments.
Plain Language Accessible to non-experts
Imagine you're cooking in a kitchen, and YUBI is like a smart pair of tongs that automatically adjusts its grip based on your finger movements. It uses virtual reality technology to precisely record every action you take in the kitchen. It's like having a super assistant that helps you efficiently complete various complex cooking tasks without needing to manually perform each step.
ELI14 Explained like you're 14
Hey there! Imagine you're playing a super cool game, and YUBI is like your game controller. It perfectly senses every finger movement you make and helps you complete tasks in the game. Just like you use your fingers to pick things up in real life, YUBI lets robots do the same! Isn't that amazing?
Glossary
YUBI (Fingertip-Driven Interface)
An innovative gripper design that directly maps human hand movements to gripper actions.
YUBI is used for efficient data collection and policy transfer.
VR System (Virtual Reality System)
A system used for precise 6 DoF trajectory tracking.
VR systems are integrated into YUBI to enhance data collection accuracy.
UMI (Universal Manipulation Interface)
A traditional handheld data collection system with bulky design and mechanical offset.
UMI is the predecessor of YUBI, facing ergonomic issues.
6 DoF (Six Degrees of Freedom)
Describes an object's movement in three-dimensional space with six degrees of freedom.
YUBI achieves precise 6 DoF trajectory tracking through VR systems.
Policy Transfer
The process of applying a policy from one platform to another.
YUBI's dataset supports cross-platform policy transfer.
Open Questions Unanswered questions from this research
- 1 How to achieve high-precision trajectory tracking without relying on VR equipment? Current methods are limited in certain environments.
- 2 How to further enhance YUBI's performance in high-load tasks? Existing designs may face challenges in extremely complex tasks.
Applications
Immediate Applications
Industrial Assembly
YUBI can be used in industrial scenarios for precise assembly tasks, enhancing operational efficiency and accuracy.
Long-term Vision
Home Automation
YUBI has the potential to be used in home automation, helping complete complex household tasks and improving quality of life.
Abstract
We introduce Yielding Universal Bidigital Interface (YUBI), a finger-aligned gripper designed to enable intuitive, ergonomic, and scalable data collection for bimanual dexterous manipulation. While handheld data collection systems such as Universal Manipulation Interface (UMI) enable affordable data collection, their bulky pistol-grip designs can pose ergonomic and usability challenges for fine-grained, dexterous manipulation tasks. To address this, YUBI presents a distinct design principle: yielding, finger-driven actuation that directly maps human finger movements to gripper jaw motion. Using the YUBI devices, we set up a data collection system with integrated VR-based 6 DoF tracking of the gripper, ensuring high-fidelity trajectory data acquisition. We curate a UMI-based dataset of unprecedented scale: 8,434 hours across 1.20M episodes and 119 tasks. Experiments show that YUBI offers advantages over the UMI gripper in versatility for complex bimanual tasks, dexterity, and operational efficiency. A single policy trained on the YUBI dataset transfers across multiple bimanual robots (UR, Franka, and ELEY) simply by mounting the gripper on each platform, confirming that the collected data are directly executable as policy supervision. We release the gripper hardware, data-collection software, and dataset as one integrated stack, offering the open community a reproducible path to large-scale data acquisition for advancing robotic foundation models.