生活支援ロボットにおける衣類と折りたたみボードの構造特徴の認識に基づく衣類折りたたみ作業の実現

抄録

In recent years, life support by robots is expected because of advancing of ageing. Although recognizing and manipulating flexble objects is one of the most impotant task for robots performing as housework supporter, it is not easy task for them including folding clothing. When human fold clothing, we use structual feature points, such as sleeves, shoulders and hems, as grasping point or reference point to fold. We recognize these structual feature points with deep learning and focus on that a folding board, a tool to fold clothing, has similar structure, it means both are deformable and have similar way to be deformed. To get the better trained model, we have to prepare the more data with proper dispersion for the task. We use 3DCG simulator to create large amount of data which allow the deep learning model to learn and recoginize various case of design and shape pattern of clothing and folding board.