WorldToken: ロボット模倣学習のための時間優先シーケンスモデリング
WorldToken: Time-First Sequence Modeling for Robotic Imitation Learning
ロボットポリシーにおいて、各タイムステップの多視点画像やプロプリオセプションなどの観測を1つのワールドトークンに融合し、時間方向にTransformerでモデル化する新しい手法を提案。RoboCasaタスクで高い成功率を達成し、データ量やモデルサイズの影響を分析した。
詳しい要約
1. どんなもの？
2. 先行研究と比べてどこがすごい？
3. 技術・手法の肝は？
4. どうやって有効だと検証した？
5. 議論はある？
6. 次に読むべき論文は？
※ AIが要旨から生成した要約です。正確性は原文をご確認ください。
著者: Chunkai Yang, Andong Yang, Chao Gao
分類: cs.RO
原文アブストラクト
Robot policies receive heterogeneous observations at each decision step, yet sequence models differ in how they organize these inputs over time. We introduce WorldToken, a time-first policy instantiation that fuses multiview images, proprioception, and task conditioning within each policy timestep into one world token. A causal temporal Transformer models the resulting world-token sequence, and a diffusion action head generates action chunks. On 23 RoboCasa tasks, an 85.3M-parameter policy trained from scratch apart from a frozen pretrained CLIP text encoder achieves 59.45% mean closed-loop success using 2,900 generated demonstrations per task. A complete factorial sweep over five dataset sizes, five model sizes, and two training seeds shows consistent gains from additional target-domain data and diminishing returns beyond moderate model size. Under same-checkpoint history truncation, reducing visible history to one or two policy timesteps lowers closed-loop success for all 50 RoboCasa policies. On RMBench Blocks Ranking, reducing visible history from 146 to 8 seconds lowers evaluator success from 95% to 28%, while an exploratory extended rollout sustains the reference swap sequence for over 850 seconds. These results establish the empirical feasibility of the complete WorldToken instantiation and characterize its data-scaling and temporal-context behavior under the tested recipes. They do not establish superiority over alternative sequence organizations or isolate which components of the complete implementation drive the observed performance.