[4G3-OS-24d] 日常生活知識とAI

[4G3-OS-24d-05] Proposal of a new force sensor principle based on externally observable shape change information

〇Ikeya Ryuichi¹, Nishida Yoshifumi^1,2 (1. Tokyo Institute of Technology, 2. National Institute of Advanced Industrial Science and Technology)

Keywords: Force Sensor, Smart Home, deep learning

In recent years, the proliferation of smart homes has led to the presence of numerous measurement devices in everyday living spaces. Among them, cameras are the most basic element and are widely used in various applications. However, force sensors are still used as measurement devices to measure the physical functions of the elderly. Several types of force sensors have been proposed, but since they are basically physical sensors, they have a limited measurement range and require special wearable equipment. In this study, we propose a new principle of force sensor that estimates external force based on shape deformation information that can be measured externally using a camera. A dataset that maps shape deformation information to external force was created using FEM, and the relationship was learned by deep learning. The external force was estimated by inputting the shape deformation information extracted from actual images into the deep learning model. Multiple validations using the dataset confirmed that external force estimation based on shape deformation information is feasible.