Development of a Physical AI Foundation Model Specialized for Soft Suits (Apr. 2026 – Dec. 2030)

This project aims to develop a lightweight physical AI foundation model specialized for soft suits that assist users’ physical capabilities through human–robot interaction. Current soft suits support only a limited range of actions and require additional AI training and manual configuration to assist user-desired movements, leading to usability challenges.

To overcome these limitations, this project will develop a physical AI model capable of real-time inference of the user’s state and surrounding environment, and of naturally assisting movements aligned with the user’s intended actions. Since the model must perform low-latency action decoding in on-device environments with constrained computational resources, research on model lightweighting and optimization will also be conducted.