austin_sirius_dataset_converted_externally_to_rlds

  • 説明

Franka の卓上操作タスク

スプリット
'train' 559
  • 機能の構造:
FeaturesDict({
    'episode_metadata': FeaturesDict({
        'file_path': Text(shape=(), dtype=string),
    }),
    'steps': Dataset({
        'action': Tensor(shape=(7,), dtype=float32),
        'action_mode': Tensor(shape=(1,), dtype=float32),
        'discount': Scalar(shape=(), dtype=float32),
        'intv_label': Tensor(shape=(1,), dtype=float32),
        '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=(84, 84, 3), dtype=uint8),
            'state': Tensor(shape=(8,), dtype=float32),
            'state_ee': Tensor(shape=(16,), dtype=float32),
            'state_gripper': Tensor(shape=(1,), dtype=float32),
            'state_joint': Tensor(shape=(7,), dtype=float32),
            'wrist_image': Image(shape=(84, 84, 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 ee 相対位置、3x ee 相対回転、1x グリッパー アクション] で構成されます。
ステップ/アクションモードテンソル(1,) float32インタラクションのタイプ。 -1: 最初の人間によるデモンストレーション。 1: 介入。 0: 自律ロボット実行 (介入前のクラスを含む)
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/intv_labelテンソル(1,) float32 action_modes と同じですが、介入前の 15 タイムステップが -10 としてラベル付けされる点が異なります。
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(84、84、3) uint8メインカメラRGB観察。
ステップ/観察/状態テンソル(8,) float32デフォルトのロボット状態は、[7x ロボット ジョイント状態、1x グリッパー状態] で構成されます。
ステップ/観察/state_eeテンソル(16,) float32エンドエフェクタの状態。ee ポーズの 4x4 同次変換行列として表されます。
ステップ/観察/state_gripperテンソル(1,) float32ロボットグリッパーの開口幅。 ~0 (クローズ) ~ ~0.077 (オープン) の範囲
ステップ/観察/state_jointテンソル(7、) float32ロボットの 7-dof ジョイント情報。
歩数/観察/手首画像画像(84、84、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。
  • 引用
@inproceedings{liu2022robot,
    title = {Robot Learning on the Job: Human-in-the-Loop Autonomy and Learning During Deployment},
    author = {Huihan Liu and Soroush Nasiriany and Lance Zhang and Zhiyao Bao and Yuke Zhu},
    booktitle = {Robotics: Science and Systems (RSS)},
    year = {2023}
}