Modeling of Dynamic Fabric Motion Using Multiple Oriented Bounding Boxes

Motion planning is the first step when we use a robot to move an object to its goal position. Dynamic fabric motion modeling is essential when a robot handles a piece of fabric in the garment manufacturing process. In this paper, we propose a dynamic fabric motion modeling method using evenly segmented multiple oriented bounding boxes (ES-MOBB) in 3D space. ES-MOBB are computed based on the 3D fabric point cloud captured by a depth camera. Then, the dynamic fabric motion is modeled and predicted as a time series of ES-MOBB using ES-MOBB-Transformer. Experiments are conducted to evaluate the performance of the proposed scheme.