Human-Humanoid Robots Cross-Embodiment Behavior-Skill Transfer Using Decomposed Adversarial Learning from Demonstration

์ €์ž: Junjia Liu, Zhuo Li, Minghao Yu, Zhipeng Dong, Sylvain Calinon, Darwin Caldwell, Fei Chen | ๋‚ ์งœ: 2024-12-19 | URL: https://arxiv.org/abs/2412.15166 📄 PDF


Essence

Figure 2

Fig. 2: Schematic overview of the cross-embodiment loco-manipulation skill transfer framework. 1) Human embodiment

Unified Digital Human (UDH) ๋ชจ๋ธ์„ ๊ณตํ†ต ํ”„๋กœํ† ํƒ€์ž…์œผ๋กœ ์‚ฌ์šฉํ•˜์—ฌ ์ธ๊ฐ„ ์‹œ์—ฐ์—์„œ ํ–‰๋™ ์›์‹œ ์š”์†Œ๋ฅผ ํ•™์Šตํ•˜๊ณ , ๋ถ„ํ•ด๋œ adversarial imitation learning๊ณผ kinematic motion retargeting์„ ํ†ตํ•ด ๋‹ค์–‘ํ•œ ํœด๋จธ๋…ธ์ด๋“œ ๋กœ๋ด‡ ํ”Œ๋žซํผ์œผ๋กœ ๋กœ์ฝ”-๋งค๋‹ˆํ“ฐ๋ ˆ์ด์…˜ ์Šคํ‚ฌ์„ ํšจ์œจ์ ์œผ๋กœ ์ „์ดํ•œ๋‹ค.

Motivation

Achievement

Figure 1

Fig. 1: Human can serve as the prototype of diverse humanoid robots, efficiently learning generalized loco-manipulation

How

Figure 2

Fig. 2: Schematic overview of the cross-embodiment loco-manipulation skill transfer framework. 1) Human embodiment

Originality

Limitation & Further Study

Evaluation

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

์ดํ‰: ๋ณธ ๋…ผ๋ฌธ์€ UDH๋ฅผ ์ค‘์‹ฌ์œผ๋กœ ํ•œ ์ฐฝ์˜์ ์ธ ๊ต์ฐจ embodiment ํ”„๋ ˆ์ž„์›Œํฌ๋ฅผ ์ œ์‹œํ•˜๋ฉฐ, functional decomposition๊ณผ adversarial imitation learning์˜ ๊ฒฐํ•ฉ, ๊ทธ๋ฆฌ๊ณ  interaction graph ๊ธฐ๋ฐ˜ ๊ณ„ํš์„ ํ†ตํ•ด ํœด๋จธ๋…ธ์ด๋“œ ๋กœ๋ด‡์˜ ๋กœ์ฝ”-๋งค๋‹ˆํ“ฐ๋ ˆ์ด์…˜ ์Šคํ‚ฌ ์ „์ด ๋ฌธ์ œ๋ฅผ ์‹ค์งˆ์ ์œผ๋กœ ํ•ด๊ฒฐํ•˜๋Š” ์ค‘์š”ํ•œ ๊ธฐ์—ฌ๋ฅผ ํ•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˜์ •