# Toyota Research Institute

> Source: https://aiwiki.ai/wiki/toyota_research_institute
> Updated: 2026-06-25
> Categories: Embodied AI, Research Organizations, Robotics
> From AI Wiki (https://aiwiki.ai), a free encyclopedia of artificial intelligence. Quote with attribution.

The **Toyota Research Institute** (**TRI**) is the United States-based research and development subsidiary of Toyota Motor Corporation, founded in 2015 with an initial commitment of about $1 billion over five years to advance [artificial intelligence](/wiki/artificial_intelligence), robotics, automated driving, and energy and materials science [1][6]. Headquartered in Los Altos, California, and led by founding chief executive Gill Pratt, a roboticist and former U.S. Defense Advanced Research Projects Agency ([DARPA](/wiki/darpa)) program manager, TRI is best known in the AI community for coining the term [Large Behavior Models](/wiki/large_behavior_models) (LBMs) and for co-developing the influential Diffusion Policy method for robot manipulation [3][4][16]. The institute began operations in January 2016 with roughly 200 employees and maintains a major second site near the Massachusetts Institute of Technology in Cambridge, Massachusetts, plus a former office in Ann Arbor, Michigan, near the University of Michigan [1][3].

TRI was announced on November 6, 2015, at a California press event and incorporated as Toyota Research Institute Inc. [1][2]. Gill Pratt, who ran the DARPA Robotics Challenge, was recruited to lead it [4][7]. In 2023 the team co-developed Diffusion Policy, in subsequent years it championed the LBM framing for robot learning, and in 2025 it published a large empirical study of LBMs and partnered with [Boston Dynamics](/wiki/boston_dynamics) to put a single LBM on the electric [Atlas](/wiki/atlas_electric) humanoid [16][20][21]. TRI is a distinct organization from Toyota's Japan-based Partner Robot Division, which builds the company's older humanoid and assistance platforms such as the T-HR3, the [Human Support Robot](/wiki/human_support_robot), and the CUE basketball robot; for that lineage see the [Toyota (robotics)](/wiki/toyota) article.

## What is the Toyota Research Institute, and when was it founded?

### Founding (2015 to 2016)

Toyota disclosed plans for the institute on November 6, 2015, at a press event in California, framing it as a way to accelerate the company's capabilities where software and AI meet physical systems [1][6]. The announced budget was approximately $1 billion over five years, and Toyota separately pledged roughly $50 million over five years to fund joint AI research centers at Stanford University and MIT, work that had been set in motion earlier in 2015 through the Stanford and MIT collaborations [1][6]. TRI was incorporated as Toyota Research Institute Inc. and opened its headquarters in Silicon Valley, initially described as being in Palo Alto near Stanford, with a second facility near MIT in Cambridge [1][3]. Operations began in January 2016 with about 200 employees [3].

Toyota recruited Gill Pratt to lead the new company [4][7]. Pratt had served as a DARPA program manager from January 2010 through August 2015, where he directed the DARPA Robotics Challenge as well as robotics and neuromorphic computing programs [4]. Before DARPA he was a founding faculty member at the Franklin W. Olin College of Engineering and an associate professor at MIT, where he ran the Leg Laboratory [4]. Pratt earned his bachelor's, master's, and doctoral degrees in electrical engineering and computer science from MIT [4].

At its launch, Toyota set out three primary goals for TRI: improving vehicle safety by reducing the likelihood of crashes, broadening access to driving for people of varying ages and abilities, and adapting Toyota's mobility technology for indoor and non-driving settings, especially to help older people live independently [1][6]. The company also said the institute would apply its work more broadly, for example to improve manufacturing efficiency and to accelerate scientific discovery in materials [1].

### Who leads TRI?

| Person | Role | Notes |
|---|---|---|
| Gill Pratt | Chief Executive Officer (founding) | Also a Toyota Motor Corporation Fellow and chief scientist; former DARPA program manager (2010 to 2015) [4][5] |
| Russ Tedrake | Senior Vice President, Large Behavior Models | Holds the Toyota Professorship at MIT (EECS, Mechanical Engineering, and Aero/Astro); leads TRI's robot-learning research [8] |

Pratt was later named a Toyota Motor Corporation Fellow and serves as a chief scientist advising Toyota's broader research, a sign of how closely TRI's leadership is tied to the parent company's technology strategy [5]. Robotics researcher Russ Tedrake, a longtime MIT professor, leads the institute's Large Behavior Models effort and is one of the most visible public faces of TRI's manipulation research [8]. Pratt has summarized the institute's stance on robotics succinctly: "Our research in robotics is aimed at amplifying people rather than replacing them." [29]

### Related Toyota entities

TRI sits within a wider set of Toyota software and venture organizations that are frequently confused with it but are legally and operationally separate:

- In 2017 TRI established a corporate venture arm, originally called Toyota AI Ventures, to invest in early-stage startups in AI, autonomy, robotics, and related fields. The fund was founded and led by managing director Jim Adler. In June 2021 it rebranded as [Toyota Ventures](/wiki/toyota_ventures) and announced two new $150 million funds (a Frontier Fund and a Climate Fund), bringing assets under management above $500 million [9][10].
- In 2018 Toyota set up a separate, Tokyo-based company called Toyota Research Institute - Advanced Development (TRI-AD) as a joint venture with suppliers Denso and Aisin to consolidate its automated-driving and vehicle software. In January 2021 TRI-AD was reorganized into Woven Planet Holdings, which was later renamed [Woven by Toyota](/wiki/woven_by_toyota) [11][12]. Despite the similar name, TRI-AD and Woven are distinct from the U.S.-based TRI; Woven owns Toyota's vehicle operating system work and the Woven City project near Mount Fuji [12].

TRI's research often feeds Toyota's production engineering and these sister organizations, but the institute itself remains focused on earlier-stage research rather than shipping production software.

## What does TRI research?

TRI organizes its work around four broad divisions: robotics, automated and assisted driving, energy and materials, and human-centered AI, supported by core machine learning research that cuts across them.

### Robotics

TRI's robotics division concentrates on home and assistive robots, motivated by aging populations in Japan and elsewhere and by the goal of helping people "age in place" [29]. Rather than pursuing full autonomy as an end in itself, TRI frames its mission as "intelligence amplification" and "mobility for all," using AI to extend human reach, strength, and independence [29]. The division has explored behavior learning from human demonstration, continuous and shared learning across robots, whole-body tactile sensing, simulation-based testing, and unconventional form factors such as a ceiling-mounted gantry robot that descends to perform household chores [29].

A signature line of work is soft, whole-body manipulation. TRI built the Soft Bubble Gripper, an inflatable latex gripper containing a depth sensor and camera so a robot can "feel" what it holds, and released the design openly [13]. Building on that, TRI introduced **Punyo**, a torso-up humanoid research platform whose arms and chest are wrapped in compliant materials and tactile sensors, with air-filled bladders along each arm that can be individually pressurized [13][14]. The name comes from a Japanese word for something soft, cute, and resilient [13]. Punyo is designed to carry and manipulate bulky objects using its whole body, hugging and bracing loads against its chest the way a person might, rather than relying only on fingertip grasps [14][15]. The fuller technical history of Punyo, the Soft Bubble Gripper, and Diffusion Policy is covered in the [Toyota (robotics)](/wiki/toyota) article; this page treats them as part of the institute's research portfolio.

### What are Large Behavior Models?

TRI coined the term Large Behavior Model to describe a single neural network, trained on large and varied robot data, that can perform many manipulation tasks and pick up new ones from a handful of demonstrations, analogous to how a [large language model](/wiki/large_language_model) generalizes across text [16][17]. The approach grew out of [Diffusion Policy](/wiki/diffusion_policy), a 2023 method (developed with Shuran Song's group, then at Columbia University) that represents a robot's visuomotor policy as a conditional denoising diffusion process and won wide attention at the Robotics: Science and Systems conference [16]. TRI positioned Diffusion Policy as a building block toward LBMs and a broader [robot foundation model](/wiki/robot_foundation_model) program, situating its work within the wider field of [embodied AI](/wiki/embodied_ai).

On July 7, 2025, the institute posted a paper titled "A Careful Examination of Large Behavior Models for Multitask Dexterous Manipulation" (arXiv 2507.05331), credited to the "TRI LBM Team" and more than 80 named authors, with Rares Ambrus listed as the corresponding contact; the study was subsequently published in the peer-reviewed journal Science Robotics [16][17][30]. The models were built on a diffusion-transformer architecture that ingests camera images, proprioception, and language prompts and outputs chunks of robot actions [16][17]. According to the paper and TRI's accompanying materials, the models were trained on roughly 1,700 hours of robot data drawn from internal bimanual teleoperation, simulation, the Universal Manipulation Interface, and the public Open X-Embodiment dataset [16][18]. The team evaluated performance across 29 tasks using more than 47,000 simulation rollouts and about 1,800 real-world trials on dual-arm stations using Franka Panda FR3 arms [16][17][18].

The headline result, emphasized in TRI's communications and widely reported, is that a single pretrained LBM can learn hundreds of tasks and acquire new skills with markedly less data than training a fresh single-task policy, with TRI citing figures of up to 80 percent less data [18][19]. The paper itself states the more measured claim that finetuned LBMs are more robust to distribution shift and need only a fraction of the data of single-task baselines to reach the same performance, reporting roughly 3 to 5 times less data in challenging settings [16][18]. A stated motivation of the work was methodological: to provide a careful, statistically rigorous benchmark for robot foundation models, an area the authors argued had been hard to evaluate honestly [16].

### How does TRI use Atlas? Partnership with Boston Dynamics

In October 2024 (announced October 16), TRI and Boston Dynamics formed a research partnership to combine TRI's Large Behavior Models with Boston Dynamics' electric Atlas humanoid, with Russ Tedrake and Boston Dynamics' Scott Kuindersma co-leading the Boston-based effort [22]. On August 20, 2025, the two organizations released results showing a single LBM driving Atlas through long, continuous sequences of whole-body manipulation and locomotion: walking, crouching, lifting, and tasks such as packing, sorting, tying a rope, unfolding a tablecloth, and flipping a barstool, while recovering from interruptions like a researcher sliding a box away mid-task [20][21].

The technical novelty was that one language-conditioned, end-to-end policy directly controlled the whole robot, treating the hands and the feet as parts of a single action representation rather than splitting low-level walking and balance from arm manipulation as humanoid stacks usually do [21]. Boston Dynamics described the policy as a roughly 450-million-parameter diffusion transformer trained with a flow-matching objective, running at about 30 Hz and predicting action chunks of length 48 (about 1.6 seconds), with training data collected by full-body virtual-reality teleoperation [21]. Russ Tedrake framed the rationale for using a capable humanoid: "Training a single neural network to perform many long-horizon manipulation tasks will lead to better generalization, and highly capable robots like Atlas present the fewest barriers to data collection for tasks requiring whole-body precision, dexterity, and strength." [20] The companies presented the demonstration as a step toward general-purpose humanoids and a way to add new skills quickly through human demonstrations [20]. This work parallels other robot-foundation-model efforts in the field such as [Mobile ALOHA](/wiki/mobile_aloha).

### What is TRI's approach to automated and assisted driving?

For self-driving, TRI publicly pursued a deliberately two-track strategy that it branded "Guardian" and "Chauffeur" [24]. Chauffeur referred to full autonomy, with the human removed from the driving task either everywhere or within a restricted operational design domain, corresponding to higher SAE automation levels [24]. Guardian, by contrast, was an advanced safety system meant to amplify rather than replace the human driver: the person stays in control, while the automated system runs in parallel, watches for hazards (including signs of distraction or drowsiness), and intervenes only when needed to avoid a crash [23][24]. TRI argued Guardian could be deployed sooner and save lives in the near term, and that it could in principle backstop either a human driver or another company's autonomous system [23]. The institute demonstrated both modes on test platforms and unveiled research vehicles such as the P4 automated-driving prototype at CES [24][25]. Much of Toyota's later production-bound automated-driving software work moved to TRI-AD and then Woven, while TRI continued more exploratory driving research [11][12].

### Energy and materials

TRI applies machine learning, robotics, and laboratory automation to accelerate the discovery and design of new materials, particularly for batteries and clean-energy applications [26][27]. Its Accelerated Materials Design and Discovery program partners with universities and companies to use AI to speed up materials research; in May 2021 TRI committed an additional $36 million to the program over four years, and the institute has said it has put well over $70 million into accelerated materials work overall [26]. In a widely cited 2019 result, researchers from MIT, Stanford, and TRI showed that combining large experimental datasets with machine learning could predict the useful cycle life of lithium-ion batteries early and accurately [27]. TRI has also collaborated with institutions such as Northwestern University on algorithms for rapidly synthesizing and screening candidate materials [26].

### Human-centered AI

TRI's human-centered AI work brings together researchers from behavioral science, machine learning, and human-computer interaction to model human behavior and build AI that augments human decision-making rather than supplanting it [28]. This division reflects the institute's overarching philosophy, shared across its robotics and driving work, that technology should keep people engaged and in control, a stance Toyota frequently summarizes as "mobility for all" [28].

## Why is TRI significant?

TRI is one of the more academically engaged corporate AI labs, publishing openly, releasing datasets and hardware designs, and seeding a venture fund, while remaining tied to a large automaker's long-term priorities around safety, aging, and clean energy [3][29]. Its robot-learning research, especially Diffusion Policy and the Large Behavior Models program, has been influential in the shift toward learned, generalist robot policies, and its 2025 collaboration with Boston Dynamics is among the most visible demonstrations of a single learned model controlling a full humanoid for whole-body manipulation [20][21]. The fact that the LBM study was peer reviewed and published in Science Robotics, rather than circulated only as a preprint, underscores TRI's emphasis on rigorous, reproducible evaluation of robot foundation models [16][30].

## Related

- [Toyota (robotics)](/wiki/toyota)
- [Human Support Robot](/wiki/human_support_robot)
- [Large Behavior Models](/wiki/large_behavior_models)
- [Diffusion policy](/wiki/diffusion_policy)
- [Boston Dynamics](/wiki/boston_dynamics)
- [Atlas (electric)](/wiki/atlas_electric)
- [Robot learning](/wiki/robot_learning)
- [Robot foundation model](/wiki/robot_foundation_model)
- [Embodied AI](/wiki/embodied_ai)
- [Mobile ALOHA](/wiki/mobile_aloha)
- [Woven by Toyota](/wiki/woven_by_toyota)
- [Toyota Ventures](/wiki/toyota_ventures)

## References

1. "Toyota Will Establish New Artificial Intelligence Research and Development Company." Toyota Motor Corporation Global Newsroom, November 6, 2015. https://global.toyota/en/detail/10171645
2. "Toyota Will Establish New Artificial Intelligence Research and Development Company." Toyota USA Newsroom, November 5, 2015. https://pressroom.toyota.com/toyota-establish-artificial-intelligence-research-development-company/
3. "Toyota Research Institute." Wikipedia. https://en.wikipedia.org/wiki/Toyota_Research_Institute
4. "Dr. Gill Pratt." Toyota Research Institute. https://www.tri.global/about-us/dr-gill-pratt
5. "TRI CEO Dr. Gill Pratt appointed Toyota Motor Corporation (TMC) Fellow." Toyota Research Institute. https://www.tri.global/news/tri-ceo-dr-gill-pratt-appointed-toyota-motor-corporation-tmc-fellow
6. "Toyota Invests $1 Billion in AI and Robots, Will Open R&D Lab in Silicon Valley." IEEE Spectrum, November 6, 2015. https://spectrum.ieee.org/toyota-to-invest-1-billion-in-ai-and-robotics-rd
7. "The man that Toyota gave $1 billion to set up a robotics lab." Quartz. https://qz.com/610485/the-man-that-toyota-gave-1-billion-to-set-up-a-robotics-lab
8. "Dr. Russ Tedrake." Toyota Research Institute. https://www.tri.global/about-us/dr-russ-tedrake
9. "Toyota AI Ventures announces rebrand as Toyota Ventures and an additional $300M to invest in emerging technologies and carbon neutrality." Toyota USA Newsroom, June 2021. https://pressroom.toyota.com/toyota-ai-ventures-announces-rebrand-as-toyota-ventures-and-an-additional-300m-to-invest-in-emerging-technologies-and-carbon-neutrality/
10. "Rebranded Toyota Ventures invests $300 million in emerging tech and carbon neutrality." TechCrunch, June 3, 2021. https://techcrunch.com/2021/06/03/rebranded-toyota-ventures-invests-300-million-in-emerging-tech-and-carbon-neutrality/
11. "Toyota Research Institute - Advanced Development, Inc. announces it will expand and improve its operations by forming Woven Planet Holdings and two new operating companies, Woven CORE and Woven Alpha." Toyota Motor Corporation Global Newsroom, 2021. https://global.toyota/en/newsroom/corporate/33786527.html
12. "Woven by Toyota, Inc." Wikipedia. https://en.wikipedia.org/wiki/Woven_by_Toyota,_Inc.
13. "Meet Punyo, TRI's Soft Robot for Whole-Body Manipulation Research." Toyota Research Institute, Medium. https://medium.com/toyotaresearch/meet-punyo-tris-soft-robot-for-whole-body-manipulation-research-949c934ac3d8
14. "Punyo is a soft robot from TRI designed for whole-body manipulation." The Robot Report. https://www.therobotreport.com/punyo-soft-robot-from-tri-designed-for-whole-body-manipulation/
15. "Toyota's Bubble-ized Humanoid Grasps With Its Whole Body." IEEE Spectrum. https://spectrum.ieee.org/humanoid-robot-tri-punyo
16. "A Careful Examination of Large Behavior Models for Multitask Dexterous Manipulation." arXiv:2507.05331, July 7, 2025. https://arxiv.org/abs/2507.05331
17. "A Careful Examination of Large Behavior Models for Multitask Dexterous Manipulation." TRI project site. https://toyotaresearchinstitute.github.io/lbm1/
18. "TRI: pretrained large behavior models accelerate robot learning." The Robot Report, July 2025. https://www.therobotreport.com/tri-pretrained-large-behavior-models-accelerate-robot-learning/
19. "Toyota Research Institute unveils breakthrough in Large Behavior Models that requires '80 percent less data'." Robotics and Automation News, July 14, 2025. https://roboticsandautomationnews.com/2025/07/14/toyota-research-institute-unveils-breakthrough-in-large-behavior-models-that-requires-80-percent-less-data/93039/
20. "AI-Powered Robot by Boston Dynamics and Toyota Research Institute Takes a Key Step Towards General-Purpose Humanoids." Toyota Research Institute, August 20, 2025. https://www.tri.global/news/ai-powered-robot-boston-dynamics-and-toyota-research-institute-takes-key-step-towards-general
21. "Large Behavior Models and Atlas Find New Footing." Boston Dynamics, August 2025. https://bostondynamics.com/blog/large-behavior-models-atlas-find-new-footing/
22. "Boston Dynamics and Toyota Research Institute Announce Partnership to Advance Robotics Research." Toyota USA Newsroom, October 16, 2024. https://pressroom.toyota.com/boston-dynamics-and-toyota-research-institute-announce-partnership-to-advance-robotics-research/
23. "Toyota Research Institute Bets Big in Vegas on Toyota Guardian Autonomy." Toyota Motor Corporation Global Newsroom. https://global.toyota/en/newsroom/corporate/26069806.html
24. "TRI Releases Video Showing First Demonstration of Guardian and Chauffeur Autonomous Vehicle Platform." Toyota Research Institute. https://www.tri.global/news/tri-releases-video-showing-first-demonstration-guardian-and-chauffeur-autonomous-vehicle
25. "Toyota Research Institute Rolls-out P4 Automated Driving Test Vehicle at CES." Toyota Motor Corporation Global Newsroom. https://global.toyota/en/newsroom/corporate/26057919.html
26. "Toyota Research Institute Commits $36 million in Funding Over the Next Four Years for Accelerated Materials Design and Discovery with University Partners." Toyota USA Newsroom, May 2021. https://pressroom.toyota.com/toyota-research-institute-commits-36-million-in-funding-over-the-next-four-years-for-advanced-materials-design-and-discovery-with-university-partners/
27. "MIT, Stanford and Toyota Research Institute Use AI to Accurately Predict the Useful Life of Batteries." Toyota USA Newsroom, 2019. https://pressroom.toyota.com/mit-stanford-and-toyota-research-institute-use-ai-to-accurately-predict-the-useful-life-batteriesie/
28. "Human-Centered AI." Toyota Research Institute. https://www.tri.global/our-work/human-centered-ai
29. "Toyota Research Institute Showcases Latest Robotics Research Aimed at Amplifying Human Ability in the Home." Toyota USA Newsroom. https://pressroom.toyota.com/toyota-research-institute-showcases-latest-robotics-research-aimed-at-amplifying-human-ability-in-the-home/
30. "A careful examination of large behavior models for multitask dexterous manipulation." Science Robotics, 2025. https://www.science.org/doi/10.1126/scirobotics.aea6201