Drone Reacher and Manipulator Reacher Datasets

该数据集包含了10万个离线16步序列，这些序列是在模拟环境中为两个任务——无人机抓取（DR）和机械臂抓取（MR）——从随机动作中生成的。数据集还包括了RGB颜色值或形状类别ID，以指定代理的目标，并且每个剧集的固定时长为100步。规模上，该数据集拥有100,000个序列，其任务旨在为多自由度控制任务提供模仿学习。This dataset contains 100,000 offline 16-step sequences, which were generated through random actions in a simulated environment for two tasks: drone grasping (DR) and robotic arm grasping (MR). The dataset also includes RGB color values or shape category IDs to specify the agent's target, and each episode has a fixed duration of 100 steps. In terms of scale, this dataset comprises 100,000 sequences, and its purpose is to provide imitation learning support for multi-degree-of-freedom control tasks.