PvP: Data-Efficient Humanoid Robot Learning with Proprioceptive-Privileged Contrastive Representations

์ €์ž: Mingqi Yuan, Tao Yu, Haolin Song, Bo Li, Xin Jin, Hua Chen, Wenjun Zeng | ๋‚ ์งœ: 2026-03-11 | DOI: 10.48550/arXiv.2512.13093 📄 PDF


Essence

Figure 1

Figure 1. (a) PvP employs contrastive learning between proprioceptive and privileged states to learn compact and task-re

PvP๋Š” ๊ณ ์œ  ๊ฐ๊ฐ(proprioceptive)๊ณผ ํŠน๊ถŒ ์ƒํƒœ(privileged state) ์‚ฌ์ด์˜ ๋Œ€์กฐ ํ•™์Šต์„ ํ™œ์šฉํ•˜์—ฌ ํœด๋จธ๋…ธ์ด๋“œ ๋กœ๋ด‡์˜ ์ „์‹  ์ œ์–ด(WBC) ํ•™์Šต์˜ ์ƒ˜ํ”Œ ํšจ์œจ์„ฑ์„ ํฌ๊ฒŒ ํ–ฅ์ƒ์‹œํ‚จ๋‹ค.

Motivation

Achievement

Figure 1

Figure 1. (a) PvP employs contrastive learning between proprioceptive and privileged states to learn compact and task-re

How

Figure 2

Figure 2. An overview of the PvP approach. (a) The components

Originality

Limitation & Further Study

Evaluation

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

์ดํ‰: PvP๋Š” proprioceptive-privileged ๋Œ€์กฐ ํ•™์Šต์ด๋ผ๋Š” ์ง๊ด€์ ์ด๋ฉด์„œ๋„ ํšจ๊ณผ์ ์ธ ๋ฐฉ๋ฒ•์œผ๋กœ ํœด๋จธ๋…ธ์ด๋“œ ๋กœ๋ด‡ ํ•™์Šต์˜ ์ƒ˜ํ”Œ ํšจ์œจ์„ฑ์„ ํฌ๊ฒŒ ํ–ฅ์ƒ์‹œํ‚ค๋ฉฐ, SRL4Humanoid ํ”„๋ ˆ์ž„์›Œํฌ๋Š” ํ•ด๋‹น ๋ถ„์•ผ์˜ ํ‘œ์ค€ ๋„๊ตฌ๋กœ์„œ ์ƒ๋‹นํ•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˜์ •