H2-COMPACT: Human-Humanoid Co-Manipulation via Adaptive Contact Trajectory Policies

์ €์ž: Geeta Chandra Raju Bethala, Hao Huang, Niraj Pudasaini, Abdullah Mohamed Ali, Shuaihang Yuan, Congcong Wen, Anthony Tzes, Yi Fang | ๋‚ ์งœ: 2025-05-23 | URL: https://arxiv.org/abs/2505.17627 📄 PDF


Essence

Figure 2

Fig. 2: Hยฒ-COMPACTโ€™s pipeline: raw force/torque and RGB inputs are cleaned by SAM2 and WHAM, then passed through

ํž˜๊ฐ ์„ผ์„œ ๊ธฐ๋ฐ˜ haptic intent inference์™€ reinforcement learning ๊ธฐ๋ฐ˜ locomotion policy๋ฅผ ๊ณ„์ธต์ ์œผ๋กœ ๊ฒฐํ•ฉํ•˜์—ฌ ์ธ๊ฐ„-ํœด๋จธ๋…ธ์ด๋“œ ํ˜‘๋ ฅ ๋ฌผ์ฒด ์šด๋ฐ˜์„ ์‹คํ˜„ํ•œ๋‹ค.

Motivation

Achievement

Figure 1

Fig. 1: Real-world humanโ€“humanoid co-manipulation. The human leads the humanoid robotโ€”unaware of the route or

How

Figure 2

Fig. 2: Hยฒ-COMPACTโ€™s pipeline: raw force/torque and RGB inputs are cleaned by SAM2 and WHAM, then passed through

Originality

Limitation & Further Study

Evaluation

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

์ดํ‰: Haptic-based intent inference์™€ force-adaptive legged locomotion์˜ ๊ณ„์ธต์  ๊ฒฐํ•ฉ์œผ๋กœ ์ธ๊ฐ„-ํœด๋จธ๋…ธ์ด๋“œ ํ˜‘๋ ฅ ๋ฌผ์ฒด ์šด๋ฐ˜์˜ ์ƒˆ๋กœ์šด ํŒจ๋Ÿฌ๋‹ค์ž„์„ ์ œ์‹œํ•˜๋ฉฐ, motion-capture free ๋ฐ์ดํ„ฐ ์ˆ˜์ง‘๊ณผ sim-to-real ๊ฒ€์ฆ์„ ํ†ตํ•ด ์‹ค์šฉ์„ฑ ๋†’์€ ์—ฐ๊ตฌ๋กœ ํ‰๊ฐ€๋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˜์ •