On the 11th, the Gwangju Institute of Science and Technology announced that a research team led by Professors Kim Gyeong-jung and Kim Seung-jun of the Department of AI Convergence has developed a walking assist robot called 'Adaptive Walker' using artificial intelligence (AI) technology in collaboration with the Computer Science and Artificial Intelligence Laboratory (CSAIL) at the Massachusetts Institute of Technology (MIT).
Existing walking assist technologies have continued to evolve, but for elderly individuals with cognitive decline, operation can be difficult, and there are limitations in responding immediately to various indoor and outdoor environments. In particular, maintaining balance on slopes and adjusting walking speed have posed significant obstacles in practical use.
In response, the research team began developing an AI-based walking assist robot that minimizes the user's cognitive burden and can actively respond to the environment. The newly developed 'Adaptive Walker' features two core functions: automatic speed adjustment according to the user's intent and assistance in maintaining a stable posture even on slopes.
First, the armrest of the walker is equipped with a high-resolution tactile sensor developed through joint research between GIST and MIT. This sensor precisely detects the user's arm movements, and the collected data is analyzed by an AI model based on artificial neural networks, predicting the acceleration intended by the user. The predicted information is reflected in the robot's motor control, allowing speed adjustment through the user's natural movements without the need for separate button operations. Experimental results showed accurate estimation of the user's walking speed within a margin of error of 20%.
The robot is also equipped with an Inertial Measurement Unit (IMU) that detects the inclination of the ground in real-time. When entering a slope, the robot detects it and automatically adjusts the lengths of the front or back legs to effectively maintain the user's balance. In actual experiments, it was confirmed that the robot maintained a horizontal position with a margin of error within 1 degree, ensuring a stable posture even on slopes.
Professor Kim Gyeong-jung noted, 'This research presents a technological solution that can provide practical help to the elderly who have difficulty walking,' and added, 'We plan to develop next-generation walking assist devices with even more functions to maximize mobility convenience for the elderly.'
The research results were presented on May 20th at the IEEE International Conference on Robotics and Automation (ICRA), one of the world's premier robotics academic conferences.