1>RoboCOIN: Advances in Bimanual Coordination for Robots through Unified Data

Published on January 06, 2026 | Translated from Spanish
>
Illustration of two coordinated robotic arms manipulating objects in an industrial environment, with overlaid diagrams of the capabilities pyramid and trajectory annotations.

RoboCOIN: Advances in Bimanual Coordination for Robots through Unified Data

Bimanual coordination remains one of the most complex challenges in robotics, where data heterogeneity has hindered progress in advanced skills. 🤖

Overcoming Fragmentation with RoboCOIN

Researchers have developed RoboCOIN, a dataset that aggregates over 180,000 demonstrations from 15 distinct robotic platforms, spanning domestic, commercial, and industrial environments. This resource includes 421 tasks organized systematically, facilitating the training of models capable of generalizing across diverse hardware configurations.

Key Features of RoboCOIN:
The public availability of RoboCOIN and its associated framework accelerates research in bimanual coordination by providing a unified standard.

Capabilities Pyramid and CoRobot Framework

The capabilities pyramid forms the innovative core, organizing annotations into three levels: trajectory concepts, subtask segmentation, and per-frame kinematic data. Alongside this, the CoRobot framework incorporates RTML | Robot Trajectory Markup Language, enabling quality evaluation, automatic annotation generation, and efficient management of heterogeneous data. 🔄

Advantages of the Unified Approach:

Impact on Multi-Robot Learning

Experimental evaluations confirm that RoboCOIN produces consistent improvements in bimanual learning models across diverse platforms. Notable advances were observed in tasks requiring dynamic coordination between limbs, such as manipulating deformable objects and assembling complex components. This publicly available resource eliminates compatibility obstacles and fosters research collaboration. 🚀