Disaster-response robots and field evaluation
I develop robots for information gathering, inspection, and manipulation in disaster sites and industrial plants that are difficult for people to access. Using the tracked robot Onix, I study mechanisms that combine mobility and usability, robotic manipulation, multi-robot cooperation, teleoperation support, and operator training. Field evaluations and competitions such as RoboCup Rescue and the World Robot Summit provide practical feedback for improving robot design and control.
Autonomous rough-terrain locomotion considering robot capabilities
I study contact-aware motion control and reinforcement learning for traversing rubble, steps, and narrow passages. My research combines terrain elevation with robot state information, including attitude, joint states, and actuator loads, to select paths and motions according to each robot’s traversal capabilities. By combining simulation-based learning with physical robot evaluations, I aim to develop autonomous navigation that remains useful on unfamiliar terrain and under degraded communication.
Autonomous exploration and manipulation with foundation models
I investigate autonomous exploration and manipulation in uncertain environments, where robots must understand their surroundings and choose actions appropriate to their objectives. My research explores visual navigation integrating camera images, terrain information, and robot states, as well as foundation models connecting language, vision, and action. I aim to connect spatial and mission understanding with physical control to support recovery from immobilization, maneuvers in confined spaces, and robotic manipulation.
Kojima, S. et al. (2022). Heterogeneous robots coordination for industrial plant inspection and evaluation at World Robot Summit 2020. Advanced Robotics, 36(21), 1102–1119. https://doi.org/10.1080/01691864.2022.2111230
Kojima, S. et al. (2020). Wall Deadlock Evasion Control Based on Rotation Radius Adjustment. IEEE Robotics and Automation Letters, 5(2), 1358–1365. https://doi.org/10.1109/LRA.2020.2967332
Kojima, S. et al. (2020). Stable Autonomous Spiral Stair Climbing of Tracked Vehicles Using Wall Reaction Force. IEEE Robotics and Automation Letters, 5(4), 6575–6582. https://doi.org/10.1109/LRA.2020.3015463
Suzuki, T., Ohno, K., Kojima, S. et al. (2021). Estimation of articulated angle in six-wheeled dump trucks using multiple GNSS receivers for autonomous driving. Advanced Robotics, 35(23), 1376–1387.
https://doi.org/10.1080/01691864.2021.1974942
Since August 2026, I have held a cross-appointment between F-REI and Tohoku University. I aim to develop robot intelligence that learns from experience in real environments and uses that experience to improve decision making and control. I also apply technologies developed for disaster-response robots to industrial plant inspection and the automation of construction machinery.