VIGOR: Visual Goal-In-Context Inference for Unified Humanoid Fall Safety

์ €์ž: Osher Azulay, Zhengjie Xu, Andrew Scheffer, Stella X. Yu | ๋‚ ์งœ: 2026-03-03 | DOI: 10.48550/arXiv.2602.16511 📄 PDF


Essence

Figure 1

Fig. 1. Vision-enabled unified fall safety for humanoids. A single learned policy integrates fall mitigation and stand-u

ํœด๋จธ๋…ธ์ด๋“œ ๋กœ๋ด‡์˜ ๋„˜์–ด์ง ์•ˆ์ „์„ฑ์„ ์œ„ํ•ด teacher-student ์ฆ๋ฅ˜ ๋ฐฉ์‹์œผ๋กœ egocentric depth์™€ proprioception๋งŒ ์‚ฌ์šฉํ•˜์—ฌ ์‹œ๊ฐ์  goal-in-context ํ‘œํ˜„์„ ํ•™์Šตํ•˜๋Š” ํ†ตํ•ฉ ์ ‘๊ทผ๋ฒ•์„ ์ œ์‹œํ•œ๋‹ค.

Motivation

Achievement

Figure 1

Fig. 1. Vision-enabled unified fall safety for humanoids. A single learned policy integrates fall mitigation and stand-u

How

Figure 2

Fig. 2: Factorized data generation yields sample-efficient imitation and scalable adaptation for humanoid fall safety

Originality

Limitation & Further Study

Evaluation

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

์ดํ‰: ํœด๋จธ๋…ธ์ด๋“œ์˜ ํ†ตํ•ฉ์  fall safety๋ฅผ ์‹œ๊ฐ ๊ธฐ๋ฐ˜์œผ๋กœ ํ•ด๊ฒฐํ•˜๋Š” ์ฐฝ์˜์  ์ ‘๊ทผ์œผ๋กœ, factorized data generation๊ณผ goal-in-context representation์˜ ๊ฐœ๋…์ด ์šฐ์ˆ˜ํ•˜๋ฉฐ zero-shot transfer ๊ฒฐ๊ณผ๊ฐ€ ์ธ์ƒ์ ์ด๋‹ค. ๋‹ค๋งŒ ์‹ค์ œ ํ™˜๊ฒฝ ์ ์šฉ์„ฑ์„ ๋” ๊ด‘๋ฒ”์œ„ํ•˜๊ฒŒ ๊ฒ€์ฆํ•  ํ•„์š”๊ฐ€ ์žˆ๋‹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˜์ •