UniAct: Unified Motion Generation and Action Streaming for Humanoid Robots

์ €์ž: Nan Jiang, Zimo He, Wanhe Yu, Lexi Pang, Yunhao Li, Hongjie Li, Jieming Cui, Yuhan Li, Yizhou Wang, Yixin Zhu, Siyuan Huang | ๋‚ ์งœ: 2025-12-30 | DOI: 10.48550/arXiv.2512.24321 📄 PDF


Essence

Figure 1

Figure 1. UniAct, a unified framework for multimodal motion generation and action streaming. UniAct enables humanoid rob

UniAct๋Š” MLLM๊ณผ causal streaming pipeline์„ ๊ฒฐํ•ฉํ•œ ๋‘ ๋‹จ๊ณ„ ํ”„๋ ˆ์ž„์›Œํฌ๋กœ, ์ธ๊ฐ„ํ˜• ๋กœ๋ด‡์ด ์–ธ์–ด, ์Œ์•…, ๊ถค์  ๋“ฑ ๋‹ค์–‘ํ•œ multimodal ๋ช…๋ น์„ sub-500ms ์ง€์—ฐ์‹œ๊ฐ„์œผ๋กœ ์‹คํ–‰ํ•  ์ˆ˜ ์žˆ๊ฒŒ ํ•œ๋‹ค.

Motivation

Achievement

Figure 3

Figure 3. UA-Net dataset analysis. (a) Representative text descriptions of human motions from UA-Net. (b) Rendered motio

How

Figure 2

Figure 2. Overview of UniAct and multimodal representations.

Originality

Limitation & Further Study

Evaluation

Novelty: 4/5 Technical Soundness: 3/5 Significance: 4/5 Clarity: 4/5 Overall: 4/5

์ดํ‰: UniAct๋Š” MLLM๊ณผ robust tracking์„ unified framework๋กœ ํ†ตํ•ฉํ•˜์—ฌ ์‹ค์ œ humanoid robot์—์„œ multimodal instruction following์„ low latency๋กœ ๋‹ฌ์„ฑํ•œ ์˜๋ฏธ ์žˆ๋Š” ์—ฐ๊ตฌ์ด๋ฉฐ, UA-Net ๋ฐ์ดํ„ฐ์…‹ ๊ธฐ์—ฌ์™€ ํ•จ๊ป˜ embodied AI ๋ถ„์•ผ์—์„œ ์ค‘์š”ํ•œ ์ง„์ „์„ ๋‚˜ํƒ€๋‚ธ๋‹ค.

← ๋ชฉ๋ก์œผ๋กœ ๋Œ์•„๊ฐ€๊ธฐ

๐ŸŽง Audio Overview

์ด ๋…ผ๋ฌธ ๋ฆฌ๋ทฐ๋ฅผ ํŒŸ์บ์ŠคํŠธํ˜• ์˜ค๋””์˜ค๋กœ ์ƒ์„ฑํ•ฉ๋‹ˆ๋‹ค. (Gemini ยท ํ‚ค๋Š” ๋ธŒ๋ผ์šฐ์ €์—๋งŒ ์ €์žฅ ยท ์™„์„ฑ๋ณธ์€ ์ด๋ฉ”์ผ๋กœ๋„ ์ „์†ก)
โ–ธ ๊ณ ๊ธ‰: ๊ตฌ์„ฑ ๋ฐฉํ–ฅ(๋Œ€๋ณธ ์ž‘์„ฑ ์ง€์นจ) ์ง์ ‘ ์ˆ˜์ •