Toyota Research Institute

Toyota Research Institute (TRI) is the artificial intelligence and robotics research subsidiary of Toyota Motor Corporation, founded in January 2016 with an initial $1 billion commitment from Toyota over five years. The institute is headquartered in Los Altos, California, with additional research labs in Cambridge, Massachusetts and Ann Arbor, Michigan. It develops AI and robotics technology spanning autonomous mobility, robotics for human assistance and home environments, advanced energy materials discovery, and human-centered AI research. As of April 2026, TRI has published more than 600 research papers, released a body of open-source robotics research including the Diffusion Policy line and Large Behavior Models, and continues to anchor Toyota's industrial-research positioning alongside Woven by Toyota, the Toyota subsidiary that productizes mobility software at scale.

At a glance

• Founded: January 2016 in Palo Alto (relocated to Los Altos), founded with a $1 billion initial commitment from Toyota Motor Corporation over five years; subsequent commitments have continued the funding trajectory.
• Status: Wholly owned research subsidiary of Toyota Motor Corporation. Not independently capitalized.
• Funding: Initial $1 billion over five years (2016 to 2020); subsequent commitments have continued the funding through 2026. Toyota does not separately disclose TRI's specific budget allocation.
• CEO: Gill Pratt, founding CEO and Chief Scientist of Toyota Motor Corporation. Roboticist; former DARPA program manager for the Robotics Challenge.
• Other notable leadership: Russ Tedrake, Senior Vice President of Robotics Research and MIT professor. Eric Krotkov, Chief Science Officer. Wolfram Burgard, former Senior Vice President of Automated Driving (departed 2021). Steve Forgeron, Vice President of Operations.
• Open weights: Yes, partial. Selected research model weights, code, and training data released open-source through GitHub and Hugging Face. The Diffusion Policy line and Large Behavior Models have been principal open-research outputs.
• Flagship outputs: Diffusion Policy (visuomotor learning for robotics), Large Behavior Models (LBMs) for robotics manipulation, autonomous-driving research stack (substantially transitioned to Woven by Toyota), advanced battery and materials AI research, and industrial-research publication output across Robotics: Science and Systems (RSS), CoRL, NeurIPS, and other venues.

Origins

Toyota Research Institute was announced in November 2015 by Toyota Motor Corporation president Akio Toyoda and founding CEO Gill Pratt at a press event in Tokyo, with a $1 billion commitment over five years and an explicit mandate to advance Toyota's AI and robotics research positioning. Pratt was recruited from his role as program manager of the DARPA Robotics Challenge (the multi-year competition that culminated in the 2015 finals at the Fairplex in Pomona, California), and had been a roboticist at MIT and Olin College.

The 2016 to 2020 founding period saw recruitment of senior research talent. Russ Tedrake joined as Senior Vice President of Robotics Research while continuing his MIT professorship; Wolfram Burgard joined as Senior Vice President of Automated Driving from the University of Freiburg; James Kuffner (the originator of RRT motion planning and a former Google senior roboticist) joined and was subsequently elevated to lead Woven by Toyota in 2018. The four research areas defined at founding (mobility, robotics, materials, and human-centered AI) have continued to anchor TRI's research agenda through 2026.

The 2018 transition saw reorganization of Toyota's autonomous-driving research and engineering. Toyota established Woven Planet (later Woven by Toyota) in 2018 as a separate Toyota subsidiary based in Tokyo, with overlap and transitions of automated-driving talent from TRI to Woven. The 2020 to 2022 period continued the Woven productization of mobility software, with TRI's autonomous-driving research substantially transitioning to Woven and TRI continuing its robotics, materials, and human-centered AI research focus.

The 2023 to 2026 period has seen TRI's most consequential public-research output. The Diffusion Policy paper (2023) presented diffusion-based learning for visuomotor robotics control, with subsequent open-source code releases and industry citations. The Large Behavior Models program (announced 2024) has built foundation-model-scale robot manipulation policies through teleoperated demonstrations, with published research and selected open-source releases. The 2024 to 2025 industry interest in robot foundation models has positioned TRI as one of the principal industrial-research organizations on this trajectory, alongside Physical Intelligence, Skild AI, and Google DeepMind's Gemini Robotics line.

Mission and strategy

Toyota Research Institute's stated mission is to improve the quality of human life through AI and robotics research, with emphasis on transportation, robotics for human assistance, materials discovery, and other application areas. The strategic premise reflects Toyota's Mobility for All framing under Akio Toyoda, with the institute's research agenda explicitly oriented around long-horizon problems with transformative potential.

The strategy combines four threads. First, robotics research with open-research output (Diffusion Policy, Large Behavior Models, and other visuomotor and manipulation research). Second, autonomous-mobility research substantially transitioned through Woven by Toyota for productization. Third, advanced battery and materials AI research with published output and connection to Toyota's broader battery-development organization. Fourth, human-centered AI research focused on the human-AI collaboration framing.

The competitive premise is that Toyota's industrial-research scale, with continued funding through Toyota Motor Corporation's research-and-development budget, supports a long-horizon research agenda that Toyota's vehicle and robotics product organizations cannot pursue under shorter productization timelines. Industry coverage has noted TRI's open-research output and senior research-talent retention as differentiators among industrial-research organizations.

Distribution channels are predominantly through research publication, open-source code releases, and continued internal collaboration with Toyota's product organizations and Woven by Toyota. TRI does not have a direct commercial-product line; the research output flows into Toyota's vehicle and robotics product roadmap and to the broader research community.

Models and products

• Diffusion Policy. Diffusion-based learning for visuomotor robotics control, published 2023 with open-source code releases and industry citations. The principal academic-publication output of TRI's robotics research line.
• Large Behavior Models (LBMs). Foundation-model-scale robot manipulation policies trained on teleoperated demonstrations, with continued research output through 2024 to 2026. The principal open-research line on foundation-model robotics.
• TidyBot, Mobile ALOHA, and other robot research platforms. Selected open-research robot platforms and demonstrations.
• Open-source materials-discovery research. Active-learning frameworks for battery materials discovery with published output.
• Internal autonomous-driving research stack. Substantially transitioned to Woven by Toyota for productization.
• Research publication output. More than 600 papers published across Robotics: Science and Systems (RSS), CoRL, NeurIPS, ICRA, and other venues.

Distribution channels include research publication, open-source code on the TRI GitHub, open-research-data releases on Hugging Face, and continued internal collaboration with Toyota's product organizations and Woven by Toyota.

Benchmarks and standing

Toyota Research Institute's evaluation framework focuses on robotics-research benchmarks, materials-discovery benchmarks, and continued open-research publication output rather than horizontal foundation-model leaderboards. The Diffusion Policy line and Large Behavior Models have established benchmark positioning on robot manipulation tasks.

The robotics-research community has consistently characterized TRI's research output as among the principal industrial-research outputs on robot manipulation, alongside Google DeepMind (Gemini Robotics), Physical Intelligence, Skild AI, and other organizations. The 2024 to 2025 industry interest in robot foundation models has positioned TRI's Large Behavior Models alongside the principal frontier robotics-foundation-model research output.

The autonomous-mobility research that has substantially transitioned to Woven by Toyota has continued; Woven's Arene operating system and Woven by Toyota's continued mobility-software development represent the productization arm of Toyota's autonomous-mobility research.

The materials-discovery research line, with active-learning frameworks for battery materials and other applications, has produced published output and continued integration into Toyota's broader battery-development organization.

Leadership

As of April 2026, Toyota Research Institute's senior leadership includes:

• Gill Pratt, CEO of Toyota Research Institute and Chief Scientist of Toyota Motor Corporation. Founding CEO since January 2016. Roboticist; former DARPA program manager for the Robotics Challenge; former MIT and Olin College professor.
• Russ Tedrake, Senior Vice President of Robotics Research. MIT professor (Toyota Professor of Electrical Engineering and Computer Science); roboticist with publication record at Robotics: Science and Systems.
• Eric Krotkov, Chief Science Officer. TRI senior leader.
• Steve Forgeron, Vice President of Operations.
• Senior research-and-engineering leadership across the principal research areas (robotics, materials, human-centered AI).

Departures and arrivals are continuous. James Kuffner departed TRI in 2018 to lead Woven by Toyota and continued Woven leadership through subsequent years; Wolfram Burgard departed in 2021. The 2024 to 2026 period has seen continued senior research-talent recruitment and engineering scale-out across the robotics research program.

Funding and backers

Toyota Research Institute's funding flows from Toyota Motor Corporation. The initial $1 billion commitment over five years (2016 to 2020) was continued through subsequent commitments, with Toyota not separately disclosing TRI's specific budget allocation in its annual reports. Toyota Motor Corporation reported approximately $11 billion in research-and-development spending in fiscal 2024, with TRI representing a portion of that allocation on AI and robotics research.

Toyota Motor Corporation is publicly traded on the Tokyo Stock Exchange and on NYSE (as TM ADR), with a market capitalization above $250 billion as of April 2026. The profitability of Toyota's vehicle business has supported TRI's continued research-funding base. Open questions on near-term resourcing are minimal compared to standalone AI labs, given Toyota's stable revenue base.

Industry position

Toyota Research Institute occupies a structurally distinctive position as one of the largest industrial AI and robotics research organizations globally, with a published-research output, the Diffusion Policy and Large Behavior Models open-research lines, the autonomous-mobility research that has substantially transitioned to Woven by Toyota, and the continued advanced materials and human-centered AI research areas. Industry coverage has characterized TRI's research output as among the principal industrial-research outputs on robot manipulation and foundation-model robotics through 2024 to 2026.

The Akio Toyoda framing of Mobility for All has continued to anchor TRI's research-mission positioning, with TRI characterized in industry coverage as one of the principal long-horizon industrial-research organizations focused on AI and robotics.

Competitive landscape

• Google DeepMind Gemini Robotics. Industrial-research peer with overlap on foundation-model robotics. The Gemini Robotics line is the principal alternative to Toyota's Large Behavior Models.
• Physical Intelligence. Direct robot-foundation-model competitor with a different commercial-startup architecture.
• Skild AI. Direct robot-foundation-model competitor.
• NVIDIA Research. Industrial-research peer with overlap on humanoid-robot infrastructure (Project GR00T) and the underlying compute platform.
• Tesla AI Optimus. Humanoid-robot research and deployment with a different vertical-integration approach.
• Boston Dynamics, Figure, 1X. Humanoid-robot organizations.
• MIT CSAIL, Stanford AI Lab (SAIL), Berkeley BAIR, CMU SCS. Academic peers on robotics research, with overlap on senior research-talent recruitment and academic-collaboration partnerships.
• Adobe Research, Microsoft Research, Apple Research. Industrial-research peers with substantially different application-area focuses.
• Woven by Toyota. Sister Toyota subsidiary; productization arm for autonomous-mobility software substantially transitioned from TRI's autonomous-driving research.

Outlook

• The continued cadence of Diffusion Policy and Large Behavior Models open-research output through 2026 to 2027.
• The integration of TRI robotics research into Toyota's product roadmap and the broader robotics industry.
• TRI's continued open-research positioning relative to commercial robot-foundation-model startups (Physical Intelligence, Skild AI, Figure, 1X).
• The continued advanced battery and materials AI research integration into Toyota's broader battery-development organization.
• TRI's senior research-talent retention through the 2026 to 2027 commercial robotics expansion.
• The cadence of senior leadership transitions, including Akio Toyoda's continued strategic role as Toyota Motor Corporation chairman.

Sources

• Toyota Research Institute official site. Institute reference.
• TRI GitHub. Selected open-source code releases.
• Diffusion Policy paper. Foundational paper from TRI's robotics research line.
• Large Behavior Models announcement. LBM program reference.
• Toyota Motor Corporation Investor Relations. Toyota's research-and-development spending and financial reporting.