Obayashi and Tohoku University establish co-creation institute for physical AI in construction

The institute will operate for five years, from Nov. 1, 2026, through Oct. 31, 2031. Obayashi Corp. and Tohoku University established the Obayashi × Tohoku University Next-Generation Construction Technology Co-Creation Institute within Tohoku University's Tough Cyber-Physical AI Research Center. The partners will pursue research and development of Tough Physical AI technologies, workforce development and the creation of new technologies.

Tohoku University and Obayashi say they will test research combining physical simulation, robotics and AI at construction sites and other locations, with the goal of putting it to practical use. The framework is designed to advance construction-site autonomy on an ongoing basis, from technology development in the laboratory through field testing. The scope of applications and how they are evaluated will help determine the technology's practicality.

The initiative centers not on individual research into robotics or AI, but on making a process that combines physical simulation of worksites, robot control and AI—and tests the results at construction sites—part of the institute's activities. The five-year institute provides a framework for ongoing research and workforce development, but the announcement does not specify investment, the number of robots to be developed or performance targets. To assess its scale, it will be important to look beyond the institute's operating period and see whether the targets for field testing and methods for evaluating results become more specific. The research builds on existing efforts. Obayashi says it has worked to improve construction planning and management through BIM/CIM and digital twins, as well as to develop autonomous and automated operation of construction machinery. Tohoku University says it has a research base in AI, robotics and disaster science. The new element is bringing these capabilities together through work on physical simulation, robotics and AI, and field testing, while creating a collaborative forum that also covers disaster science, energy, law and ethics, in addition to civil engineering and architecture. Rather than replacing existing technology development, the institute appears intended to establish an ongoing link between research and practical application in the field. For autonomy at construction sites, challenges include not only controlling individual robots but also coordinating with workers and responding to changing site conditions. Tohoku University's definition of Tough Physical AI includes stable autonomous operation at disaster sites and outdoors under uncertain conditions. Another effort to test the autonomous movement of multiple transport vehicles in a real-world setting is [Narita Airport to conduct automated logistics road demonstration in fiscal 2026, with Obayashi, Cuebus and NTT Data participating](/news/naritakuukou-jidou-butsuryuu-douro-no-jisshou-jikken-wo-2026-2f90e8). The applications differ, but both underscore the importance of operating and testing equipment in real environments. In construction, differences between simulated environments and actual sites, the division of roles between workers and robots, and safety during disasters are issues that will affect practical deployment. The announcement does not specify a research budget, target equipment, performance metrics or dates for individual field tests. Key points to watch include which tasks will be targeted, what metrics will be used to assess safety and effectiveness, and how information gathered in the field will be used to improve simulations and control systems.

The institute will combine research on physical simulation of construction sites, robot control and AI, then test the results at Obayashi construction sites and elsewhere to move them toward practical use. In addition to physical AI, its areas of focus will include civil engineering and architecture, energy, human-robot collaboration, 3D concrete printers, disaster science and recovery. Tohoku University's announcement describes Tough Physical AI technology as enabling robots, autonomous vehicles and other systems to recognize their environment and physical laws, then operate and control themselves in disaster sites, extreme environments and outdoor settings where conditions are uncertain.