VB-Com: Learning Vision-Blind Composite Humanoid Locomotion Against Deficient Perception

์ €์ž: Junli Ren, Tao Huang, Huayi Wang, Zirui Wang, Qingwei Ben, Junfeng Long, Yanchao Yang, Jiangmiao Pang, Ping Luo | ๋‚ ์งœ: 2025-02-20 | URL: https://arxiv.org/abs/2502.14814 📄 PDF


Essence

Figure 1

Fig. 1: Overview. VB-Com enables humanoid robots (move direction in orange arrorw) to traverse dynamic terrains and obst

VB-Com์€ ํœด๋จธ๋…ธ์ด๋“œ ๋กœ๋ด‡์ด ์‹œ๊ฐ ์ •๋ณด์˜ ๊ฒฐ์†์— ๋Œ€์‘ํ•˜๊ธฐ ์œ„ํ•ด ์‹œ๊ฐ ๊ธฐ๋ฐ˜ ์ •์ฑ…๊ณผ ๊ณ ์œ ๊ฐ๊ฐ ๊ธฐ๋ฐ˜์˜ ๋งน๋ชฉ ์ •์ฑ…์„ ๋™์ ์œผ๋กœ ์ „ํ™˜ํ•˜๋Š” ๋ณตํ•ฉ ์ œ์–ด ํ”„๋ ˆ์ž„์›Œํฌ๋ฅผ ์ œ์•ˆํ•œ๋‹ค.

Motivation

Achievement

Figure 1

Fig. 1: Overview. VB-Com enables humanoid robots (move direction in orange arrorw) to traverse dynamic terrains and obst

How

Figure 2

Fig. 2: Overview of our framework: In VB-Com, we develop two locomotion policiesโ€”one perceptive and one non-perceptiveโ€”t

Originality

Limitation & Further Study

Evaluation

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

์ดํ‰: VB-Com์€ ํœด๋จธ๋…ธ์ด๋“œ ๋กœ๋ด‡์˜ ์ง€๊ฐ ๊ฒฌ๊ณ ์„ฑ ๋ฌธ์ œ๋ฅผ ์ •์ฑ… ํ•ฉ์„ฑ์œผ๋กœ ์šฐ์•„ํ•˜๊ฒŒ ํ•ด๊ฒฐํ•˜๋ฉฐ, return estimator ๊ธฐ๋ฐ˜ ๋™์  ์„ ํƒ ๋ฉ”์ปค๋‹ˆ์ฆ˜์€ ์ฐฝ์˜์ ์ด๊ณ  ์‹ค์šฉ์ ์ด๋‹ค. ๋™์  ์ง€ํ˜• ๋ฐ ์ง€๊ฐ ๋…ธ์ด์ฆˆ ์‹œ๋‚˜๋ฆฌ์˜ค์˜ ์ฒด๊ณ„์  ๊ตฌ์„ฑ๊ณผ ๋‘ ํœด๋จธ๋…ธ์ด๋“œ ํ”Œ๋žซํผ์—์„œ์˜ ๊ฒ€์ฆ์ด ๊ฐ•์ ์ด๋‚˜, ์‹ค์ œ ๋ฐฐํฌ ๊ฒฐ๊ณผ ํ™•์žฅ๊ณผ ์ผ๋ฐ˜ํ™” ๋Šฅ๋ ฅ ๋ถ„์„์ด ๋ณด๊ฐ•๋˜๋ฉด ๋”์šฑ ์„ค๋“๋ ฅ ์žˆ์„ ๊ฒƒ์ด๋‹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˜์ •