cliport

该数据集名为Cliport基准，包含了在PyBullet仿真环境中进行的10项语言条件控制操作任务。在这些任务中，由具有吸盘式夹具的Universal Robot UR5e机器人执行操作。每个任务中的每个情节都是通过采样一组物体和属性构建的，并通过语言目标描述的所需构型来操作这些物体的一个子集。该数据集的规模涵盖了10项语言条件控制操作任务，主要任务是进行语言条件控制操作。This dataset is named the Cliport Benchmark, which includes 10 language-conditioned robotic manipulation tasks conducted in the PyBullet simulation environment. All tasks are performed by a Universal Robots UR5e robotic arm equipped with a suction cup gripper. Each episode for every task is constructed by sampling a set of objects and their attributes, and involves manipulating a subset of these objects to reach the desired configuration specified by a natural language goal description. The core task of this dataset is language-conditioned robotic manipulation.