R&D Project
Last updated:
Goal 3 R&D Projects (Selected in FY2026)General-Purpose Autonomous Humanoids Growing through Multi-layered Instruction and InheritanceJPMJMS263D
Project manager (PM)KAWAHARAZUKA KentoLecturer, Next Generation Artificial Intelligence Research Center, The University of Tokyo
Project SiteSummary of the project
For humanoids to adapt to diverse and changing environments, tasks, and embodiments, a fundamental understanding of what they have learned is essential. Just as humans can be said to truly understand something only when they are able to teach it to others, robots, too, need to become capable of passing on their own experiences and skills.
We therefore aim to develop general-purpose autonomous humanoids that grow progressively through multi-layered instruction — hands-on guidance, learning by imitation, and verbal instruction — and through multi-layered inheritance of experience and skills from humans to robots, from robots to other robots, and from robots to humans. We will evaluate their adaptability, versatility, and ability to pass on what they have learned.
Specifically, we pursue R&D on humanoids suited to instruction and inheritance, foundational systems that enable instruction and inheritance, task execution and real-world deployment based on instruction and inheritance, and next-generation community building. By 2050, we aim to realize humanoids that grow together with people and other robots by teaching and learning from one another, and that can take on all physical tasks that could reasonably be entrusted to part-time workers.
*Certain illustrations in this document were created with the assistance of generative AI.
Milestone by year 2030
By 2030, toward realizing multi-layered instruction and inheritance, we aim to achieve the following four milestones with humanoids.
[House-moving tasks with rapid adaptation to unknown environments]
In settings where room layouts and items are unknown in advance, humanoids autonomously move furniture and cardboard boxes from the room and load them onto a truck, following human instructions.
[Unseen household tasks by abstracting and generalizing experience]
Humanoids perform household tasks not learned in advance — such as brewing pour-over coffee, preparing a light breakfast, and cleaning — after only a single human demonstration or verbal instruction.
[Motor-skill transfer across diverse embodiments]
Motor skills such as mat, horizontal-bar, and vaulting exercises are transferred from humans to robots, from robots to other robots, and from robots to humans.
[On-demand work grounded in real-world needs]
Humanoids perform inspection and line work in factories and farms after only a few to a few dozen minutes of on-site instruction.
R&D Items
- [1] Research and development of humanoids suited to instruction and inheritance
- [2] Research and development of foundation systems enabling instruction and inheritance
- [3] Task execution and real-world deployment based on instruction and inheritance
Performers
| R&D Item [1][2][3] | KAWAHARAZUKA Kento | The University of Tokyo |
|---|---|---|
| R&D Item [1][2][3] | MUROOKA Masaki | AIST |
| R&D Item [1][2][3] | KURITA Shuhei | National Institute of Informatics |
| R&D Item [1][2][3] | KANESAKI Asako | Tohoku University |
| R&D Item [1][2][3] | YAGI Satoshi | The University of Osaka |
| R&D Item [1][2][3] | MIYAZAWA Kazuki | The University of Osaka |
| R&D Item [1][2][3] | HONDA Kohei | Nagoya University |
| R&D Item [1][2][3] | HIEIDA Chie | The University of Electro-Communications |
| R&D Item [1][2][3] | KOBAYASHI Taisuke | National Institute of Informatics |
PDF Download
- Summary of the project (328 KB)