์ ์: Physical Intelligence, Kevin Black, Noah Brown, James Darpinian, Karan Dhabalia, Danny Driess, Adnan Esmail, Michael Equi, Chelsea Finn, Niccolo Fusai, Manuel Y. Galliker, Dibya Ghosh, Lachy Groom, Karol Hausman, Brian Ichter, Szymon Jakubczak, Tim Jones, Liyiming Ke, Devin LeBlanc, Sergey Levine, Adrian Li-Bell, Mohith Mothukuri, Suraj Nair, Karl Pertsch, Allen Z. Ren, Lucy Xiaoyang Shi, Laura Smith, Jost Tobias Springenberg, Kyle Stachowicz, James Tanner, Quan Vuong, Homer Walke, Anna Walling, Haohuan Wang, Lili Yu, Ury Zhilinsky | ๋ ์ง: 2025-04-22 | URL: https://arxiv.org/abs/2504.16054 📄 PDF
Fig. 1: The ฯ0.5 model transfers knowledge from a heterogeneous range of data sources, including other robots, high-leve
ฯ0.5๋ heterogeneousํ ๋ค์ค ๋ฐ์ดํฐ ์์ค(๋ค์ํ ๋ก๋ด, ์น ๋ฐ์ดํฐ, ์๋ฏธ๋ก ์ ์์ธก)์์ co-trainingํ์ฌ ์ค์ ๊ฐ์ ์์ ์ฅ์๊ฐ์ ๋ณต์กํ ์กฐ์ ์์ ์ ์ํํ ์ ์๋ Vision-Language-Action ๋ชจ๋ธ์ด๋ค.
Fig. 2: ฯ0.5 cleaning a new kitchen. The robot is tasked with cleaning a kitchen in a home that was not in the training
์ดํ: ฯ0.5๋ heterogeneous ๋ฐ์ดํฐ ์์ค์ ์ฒด๊ณ์ ํตํฉ์ ํตํด VLA ๋ชจ๋ธ์ ์ค์ ํ๊ฒฝ ์ผ๋ฐํ ๋ฌธ์ ๋ฅผ ์ฒ์์ผ๋ก ์ค์ง์ ์ผ๋ก ํด๊ฒฐํ ์ฑ๊ณผ์ด๋ฉฐ, ๊ณ์ธต์ ์๋ฏธ๋ก ์ ๊ตฌ์กฐ์ co-training ํ๋ ์์ํฌ๋ ๋ก๋ด ํ์ต์ ์ค์ํ ์ค๊ณ ์์น์ ์ ์ํ๋ค.