Learning World Models from Egocentric Video for Home Robots
日本語版は未提供のため、英語原文で表示しています。
30秒で読む
A video world model trained on robot egocentric footage predicts outcomes of actions for evaluation and planning.
研究課題
Can world models replace physical evaluation?
問題
Evaluating home robots is slow and unsafe.
従来の手法
Physical rollouts or hand-built simulators.
新しい手法
Action-conditioned video prediction.
結果
Correlates with real success rates on household tasks.
限界
Hallucinated physics; single robot.
産業への影響
Positions video generation vendors as robotics infrastructure.