Cup-Stacking Force Modulation Dataset

该数据集包含了从机器人操作任务中收集的关节位置、速度和扭矩数据，该任务涉及堆叠杯子的情景，其中机器人通过自然语言指令来调节其抓握力。此外，数据集还包括每个机器人的15维关节数据以及以640x360像素分辨率捕获的RGB图像。该数据集采用了DABI方法进行扩展，有效地将高频控制数据降采样以匹配图像数据的频率。规模上，该数据集由60个演示组成（通过DABI方法从6个扩展而来）。其任务是跟随自然语言指令进行模仿学习。

This dataset contains joint position, velocity and torque data collected from robotic manipulation tasks involving cup-stacking scenarios, where the robot adjusts its grasping force via natural language instructions. Additionally, the dataset includes 15-dimensional joint data for each robot and RGB images captured at a resolution of 640×360 pixels. Expanded using the DABI method, the dataset effectively downsampled high-frequency control data to match the frequency of the image data. In terms of scale, this dataset consists of 60 demonstrations, which were expanded from 6 via the DABI method. The task of this dataset is imitation learning following natural language instructions.