It Takes Two: Learning Interactive Whole-Body Control Between Humanoid Robots

์ €์ž: Zuhong Liu, Junhao Ge, Minhao Xiong, Jiahao Gu, Bowei Tang, Wei Jing, Siheng Chen | ๋‚ ์งœ: 2025-10-11 | DOI: 10.48550/arXiv.2510.10206 📄 PDF


Essence

Figure 2

Figure 2: Overview of the proposed Harmanoid framework. It contains two key components: (i) contact-aware motion retarge

Harmanoid๋Š” ๋‘ ๊ฐœ์˜ ํœด๋จธ๋…ธ์ด๋“œ ๋กœ๋ด‡ ๊ฐ„ ์ƒํ˜ธ์ž‘์šฉ ๋™์ž‘์„ ๋ชจ๋ฐฉํ•˜๋Š” ํ”„๋ ˆ์ž„์›Œํฌ๋กœ, ์ ‘์ด‰ ์ธ์‹ motion retargeting๊ณผ ์ƒํ˜ธ์ž‘์šฉ ๊ธฐ๋ฐ˜ motion controller๋ฅผ ํ†ตํ•ด ํ‚ค๋„ค๋งˆํ‹ฑ ์ถฉ์‹ค๋„์™€ ๋ฌผ๋ฆฌ์  ํ˜„์‹ค์„ฑ์„ ๋™์‹œ์— ๋ณด์กดํ•œ๋‹ค.

Motivation

Achievement

Figure 5

Figure 5: Qualitative comparison between Inter-X mocap data, single-PHC retargeting, and our contact-aware retargeting

How

Figure 3

Figure 3: Overview of our motion retargeting pipeline that

Originality

Limitation & Further Study

Evaluation

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

์ดํ‰: Harmanoid๋Š” ๋‹ค์ค‘ ํœด๋จธ๋…ธ์ด๋“œ ์ƒํ˜ธ์ž‘์šฉ ๋™์ž‘ ๋ชจ๋ฐฉ์˜ ๋ช…ํ™•ํ•œ ๋ฌธ์ œ๋ฅผ ์ฒด๊ณ„์ ์œผ๋กœ ํ•ด๊ฒฐํ•˜๋ฉฐ, contact-aware retargeting๊ณผ interaction-aware control์˜ ๊ฒฐํ•ฉ์œผ๋กœ ๊ณ ๋ฆฝ ๋ฌธ์ œ๋ฅผ ํšจ๊ณผ์ ์œผ๋กœ ๊ทน๋ณตํ•˜๋Š” ์ฒซ ํ”„๋ ˆ์ž„์›Œํฌ์ด๋‹ค. ์ข…ํ•ฉ์ ์ธ ์‹คํ—˜๊ณผ ์šฐ์ˆ˜ํ•œ ์„ฑ๋Šฅ์œผ๋กœ humanoid robotics ๋ถ„์•ผ์— ์ค‘์š”ํ•œ ๊ธฐ์—ฌ๋ฅผ ํ•˜๋‚˜, sim-to-real ๊ฒ€์ฆ ๋ถ€์žฌ์™€ 2-agent ์ œํ•œ์ด ์‹ค์ œ ์ ์šฉ์˜ ์™„์ „์„ฑ์„ ์ œ์•ฝํ•œ๋‹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˜์ •