로보_아이_유_르5에

  • 설명 :

Universal Robots UR5e 데모 데이터셋은 VLA 모델 학습에 사용됩니다. 데이터는 사전 프로그래밍된 로봇 루틴을 사용하여 실제 로봇 작업 공간에서 수집되었습니다.

나뉘다 예시
  • 특징 구조 :
FeaturesDict({
    'episode_metadata': FeaturesDict({
        'episode_index': int32,
        'file_name': string,
    }),
    'steps': Dataset({
        'absolute_action_mask': Tensor(shape=(7,), dtype=bool),
        'action': Tensor(shape=(7,), dtype=float32),
        'action_normalization_mask': Tensor(shape=(7,), dtype=bool),
        'is_first': bool,
        'is_last': bool,
        'is_terminal': bool,
        'language_instruction': string,
        'observation': FeaturesDict({
            'image': Image(shape=(480, 640, 3), dtype=uint8),
            'robot_state': Tensor(shape=(15,), dtype=float32),
        }),
    }),
})
  • 기능 문서화 :
특징 수업 모양 디타입 설명
특징 사전
에피소드 메타데이터 특징 사전
에피소드 메타데이터/에피소드 인덱스 텐서 int32
에피소드 메타데이터/파일 이름 텐서
단계 데이터셋
단계/절대_액션_마스크 텐서 (7,) 부울
단계/조치 텐서 (7,) float32
단계/액션_정규화_마스크 텐서 (7,) 부울
단계/첫 번째입니다 텐서 부울
단계/is_last 텐서 부울
단계/is_terminal 텐서 부울
단계/언어_지침 텐서
단계/관찰 특징 사전
단계/관찰/이미지 영상 (480, 640, 3) uint8
단계/관찰/로봇 상태 텐서 (15,) float32
@article{RoboAIUR5e2025,
  title        = {RoboAI UR5e Training Dataset},
  author       = {Joonas Rouhiainen},
  institution  = {RoboAI Research Center, Satakunta University of Applied Sciences},
  year         = {2025},
  howpublished = {\url{https://sites.google.com/view/roboai-ur5e/home} },
}