The International Journal of Robotics Research Study, Ahead of Publish.
Robot understanding for deformable item control– such as fabrics– is commonly performed in simulation as a result of the present constraint of understanding techniques to comprehend towel’s contortion. Because of this, the robotics area is constantly on the look for even more reasonable simulators to lower as high as feasible the sim-to-real space, which is still rather huge specifically when vibrant movements are used. We offer a towel dataset including 120 top quality recordings of numerous fabrics throughout vibrant movements. Making Use Of a Movement Capture System, we tape-record the place of key-points on the towel surface area of 4 sorts of textiles (cotton, jeans, woollen and polyester) of 2 dimensions and at various rates. The situations thought about are all vibrant and entail quick trembling and turning of the fabrics, accidents with frictional items, solid hits with a lengthy and slim stiff item and also self-collisions. We clarify thoroughly the situations thought about, the gathered information and exactly how to review it and utilize it. On top of that, we suggest a statistics to utilize the dataset as a criteria to evaluate the sim-to-real space of any kind of towel simulator. Ultimately, we reveal that the taped trajectories can be straight implemented by a robot arm, allowing understanding by presentation and various other replica finding out techniques.Dataset: https://doi.org/10.5281/zenodo.14644526Video: https://fcoltraro.github.io/projects/dataset/
发布者：Franco Coltraro，转转请注明出处：https://robotalks.cn/tracking-cloth-deformation-a-novel-dataset-for-closing-the-sim-to-real-gap-for-robotic-cloth-manipulation-learning/
