Pushing Task

该数据集是基于开源版本的OpenAI健身房FetchPush-v1任务，专为机器人操作实验设计。它包含了依赖于目标的稀疏奖励函数以及供代理执行的行动计划。该数据集的状态空间为10维，包括末端执行器的位置和箱子的位置，任务内容为机器人操作。This dataset is based on the open-source version of the OpenAI Gym FetchPush-v1 task, which is specifically designed for robotic manipulation experiments. It contains a goal-dependent sparse reward function and action plans for the agent to execute. The state space of this dataset is 10-dimensional, including the positions of the end-effector and the box, and the task focuses on robotic manipulation.