#Japan #Tokyo #Physical AI #commissure #UMI Technology

commissure Accelerates Mimetic Learning for Physical AI

The Tokyo-based startup, commissure, has been selected for the NEDO's "NEP Yooshin Course 3000" initiative aimed at advancing deep tech talent development. This funding will significantly contribute to the development of a tactile integration platform that accelerates mimetic learning in robots by capturing human interaction data with tools and devices.

Overview of the Initiative

NEDO, the New Energy and Industrial Technology Development Organization, supports initiatives in deep tech to foster entrepreneurship. commissure’s project targets the development of a haptic module that can be retrofitted onto various devices, including robots and Universal Manipulation Interface (UMI) devices. This will enable the collection and utilization of tactile data to help robots learn how to handle tasks akin to human operators.

Significance of Tactile Learning

Every time humans engage in tasks such as inserting connectors or manipulating wires, they rely not only on visual feedback but also on the subtle changes in resistance and contact that inform their adjustments. By equipping robots with the ability to learn from tactile data—alongside visual and kinetic information—commissure hopes to bridge the gap between human expertise and robotic execution. The tactile integration platform aims to synchronize sensory feedback with visual and motion data to create a cohesive learning environment for robots.

Research and Development Focus

The heart of commissure's R&D is connecting the data collected from human demonstrations with the learning capabilities of robots. The startup plans to develop two key components:
1. HaptoAtom: A slim, lightweight tactile module that can be added to both robots and UMI devices, designed to capture pressure distribution across contact surfaces. This will integrate seamlessly with existing systems while permitting tactile sensing enhancements.
2. HaptoConnect: This software will synchronize tactile data with image and motion data, providing a comprehensive framework for robot learning. It will interface with popular learning frameworks like LeRobot and simulation environments such as Isaac Sim, facilitating a smooth transition from data collection to learning and evaluation.

Integrating UMI Technology into Business Operations

As bots take on more manufacturing and operational roles, defining the touch and pressure parameters is vital to ensuring optimal performance. By meshing tactile technologies with UMI data acquisition methods, commissure aims to assist a range of businesses in analyzing and improving key workflows. Their UMI service employs handheld devices equipped with cameras to capture human operational data, paving the way for substantial advancements in automation.

The ultimate goal is to create a robust ecosystem where businesses can conduct comprehensive studies, assess manufacturing success criteria, and enable continuous improvement in robot-trained tasks. This not only heightens efficiency but also addresses the pressing skills gap in many industries today.

Promoting Knowledge Sharing

In line with academic openness, commissure plans to make the collected genuine tactile datasets available to the research community. This initiative aims to foster further studies in tactile learning by enabling third parties to replicate and compare outcomes. By sharing viable models and data, commissure seeks to inspire collaborative research and practical implementation involving tactile feedback in robotic systems.

Entrepreneurs Behind the Project

Leadership at commissure includes CEO Masaki Mizobashi and CTO Arata Horie, both of whom have extensive backgrounds in technology and entrepreneurship. Their joint vision is to refine tactile technology for industry use, ultimately paving the way for more sophisticated robotics that can adapt and learn in real-world environments. Emphasizing hands-on applications, around producing tangible solutions from their research, is central to their mission.

Masaki Mizobashi states, "Our commitment is to empower workers and support businesses by turning our research into practical tools that robots can use to shoulder tasks in environments facing labor shortages."

Arata Horie, with his engineering expertise, remarks, "It's essential that our technology allows robots to learn from human actions effectively. This intersection of haptic feedback with AI will significantly bridge the gap between human skillsets and robotic capabilities."

As commissure embarks on this transformative journey, their work continues to serve as a blueprint for future innovation in the realms of robotics and physical AI, responding to a growing demand for heightened efficiency and skill transfer in the industrial landscape. Through ongoing collaborations and development, they aim to build a future where robotics and humans work seamlessly together in harmony.