stanford_hydra_dataset_converted_externally_to_rlds

  • 説明

長期的な課題を解決するフランカ

スプリット
'train' 570
  • 機能の構造:
FeaturesDict({
    'episode_metadata': FeaturesDict({
        'file_path': Text(shape=(), dtype=string),
    }),
    'steps': Dataset({
        'action': Tensor(shape=(7,), dtype=float32),
        'discount': Scalar(shape=(), dtype=float32),
        'is_dense': Scalar(shape=(), dtype=bool),
        'is_first': bool,
        'is_last': bool,
        'is_terminal': bool,
        'language_embedding': Tensor(shape=(512,), dtype=float32),
        'language_instruction': Text(shape=(), dtype=string),
        'observation': FeaturesDict({
            'image': Image(shape=(240, 320, 3), dtype=uint8),
            'state': Tensor(shape=(27,), dtype=float32),
            'wrist_image': Image(shape=(240, 320, 3), dtype=uint8),
        }),
        'reward': Scalar(shape=(), dtype=float32),
    }),
})
  • 機能ドキュメント:
特徴クラスDタイプ説明
特徴辞書
エピソード_メタデータ特徴辞書
エピソードメタデータ/ファイルパス文章元のデータ ファイルへのパス。
ステップデータセット
ステップ/アクションテンソル(7、) float32ロボットのアクションは、[3x EEF 位置デルタ、3x オイラー角の EEF 方向デルタ、1x クローズ グリッパー] で構成されます。
歩数/割引スカラーfloat32割引が指定されている場合、デフォルトは 1 です。
ステップ/is_denseスカラーブール状態がウェイポイント (010) またはデンス モード (x111) の場合は True。
ステップ/is_firstテンソルブール
ステップ/is_lastテンソルブール
ステップ/is_terminalテンソルブール
ステップ/言語_埋め込みテンソル(512,) float32コナ言語の埋め込み。 https://tfhub.dev/google/universal-sentence-encoder-large/5を参照してください。
ステップ/言語説明文章言語指導。
ステップ/観察特徴辞書
手順・観察・イメージ画像(240、320、3) uint8メインカメラRGB観察。
ステップ/観察/状態テンソル(27,) float32ロボットの状態は、[3x EEF 位置、4x クォータニオンでの EEF 方向、3x オイラー角での EEF 方向、7x ロボット関節角度、7x ロボット関節速度、3x グリッパー状態で構成されます。
歩数/観察/手首画像画像(240、320、3) uint8リストカメラRGB観察。
歩数/報酬スカラーfloat32提供されている場合は報酬、デモの最終ステップで 1。
@article{belkhale2023hydra,
 title={HYDRA: Hybrid Robot Actions for Imitation Learning},
 author={Belkhale, Suneel and Cui, Yuchen and Sadigh, Dorsa},
 journal={arxiv},
 year={2023}
}