SafeDoorManip50k
收藏资源简介:
SafeDoorManip50k是由中山大学开发的用于机器人门操作任务的大规模模拟数据集。该数据集包含57种不同结构和颜色纹理的门,分为45个已见门和12个未见门,总计47,727条训练数据和9,018条测试数据。数据集通过模拟环境生成,涵盖了门的类型、尺寸、位置、机械属性和光照条件等多种随机化场景。创建过程结合了视觉和触觉反馈,旨在通过多模态数据训练模型,确保在门操作过程中机器人施加的力是安全的。该数据集主要应用于机器人操作任务中的安全状态规划,旨在解决机器人操作过程中因不当力导致物体或机器人损坏的问题。
SafeDoorManip50k is a large-scale simulated dataset developed by Sun Yat-sen University for robotic door manipulation tasks. This dataset contains 57 doors with distinct structures, colors and textures, which are divided into 45 seen doors and 12 unseen doors, with a total of 47,727 training samples and 9,018 test samples. The dataset is generated in a simulated environment, covering various randomized scenarios including door type, dimension, position, mechanical properties and lighting conditions. Its development integrates visual and haptic feedback, aiming to train models using multimodal data to ensure the safety of forces applied by robots during door manipulation. This dataset is mainly applied to safety state planning in robotic manipulation tasks, aiming to solve the problem of damage to objects or robots caused by improper forces during robotic operations.




