The International Journal of Robotics Study, Ahead of Publish.
Dual-arm control of everyday living items is vital for robotics running in home atmospheres. Knowing from presentation is an appealing technique for mentor robotics dual-arm control abilities. It typically needs a dataset of individuals showing the series of control activities to accomplish an objective. That series can be stood for making use of symbolical or trajectory encoding. The symbolic-encoded series is referred to as the job strategy. The primary restrictions of existing datasets are that the majority of have a tendency to neglect dual-arm control abilities and leave out the official grammar utilized to annotate the job strategies. This paper presents BiCap, an unique bi-modal dataset of dual-arm control activities on everyday living items paired with a bio-inspired activity context-free grammar for fine-grained job strategy note to train, examination, and confirm picking up from demonstration-based formulas for robot dual-arm control. 15 individuals were hired. The experimenter put reflective pens on their top arm or legs and hips. After that, the individuals rested at a table where 1 or 2 items were put. They executed among the complying with jobs: putting, opening up, and passing, making use of both hands. An RGB cam aiming in the direction of the table videotaped the individuals’ hand activities. Ultimately, an annotator assessed the RGB video clips and created the individuals’ job strategies making use of the bio-inspired activity context-free grammar. The individuals’ upper-limb kinematics were calculated, also, which gives the trajectory-encoded activity series. The resulting dataset, BiCap, has 4,026 job strategies, video clips, and movement information of the 15 individuals.
发布者：David Carmona，转转请注明出处：https://robotalks.cn/bicap-a-novel-bi-modal-dataset-of-daily-living-dual-arm-manipulation-actions/
