RoBridge: A Hierarchical Architecture Bridging Cognition and Execution for General Robotic Manipulation

์ €์ž: Kaidong Zhang, Rongtao Xu, Pengzhen Ren, Junfan Lin, Hefeng Wu, Liang Lin, Xiaodan Liang | ๋‚ ์งœ: 2025-05-03 | URL: https://arxiv.org/abs/2505.01709 📄 PDF


Essence

Figure 1

Figure 1. Comparison of RoBridge and previous methods. Declarative skill methods (left) directly generate specific contr

RoBridge๋Š” Vision-Language Model์˜ ์„ ์–ธ์  ๋Šฅ๋ ฅ๊ณผ ๊ฐ•ํ™”ํ•™์Šต์˜ ์ ˆ์ฐจ์  ๋Šฅ๋ ฅ์„ ํ†ตํ•ฉํ•˜๋Š” ๊ณ„์ธต์  ์•„ํ‚คํ…์ฒ˜๋กœ, Invariant Operable Representation(IOR)์„ ์ƒ์ง•์  ๋ธŒ๋ฆฟ์ง€๋กœ ํ™œ์šฉํ•˜์—ฌ ๋กœ๋ด‡์˜ ์ธ์ง€์™€ ์‹คํ–‰ ๊ฐ„ ๊ฒฉ์ฐจ๋ฅผ ํ•ด์†Œํ•œ๋‹ค.

Motivation

Achievement

Figure 1

Figure 1. Comparison of RoBridge and previous methods. Declarative skill methods (left) directly generate specific contr

How

Figure 2

Figure 2. RoBridge overview. RoBridge adopts a three-layer architecture, consisting of a high-level cognitive planner (H

Originality

Limitation & Further Study

Evaluation

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

์ดํ‰: RoBridge๋Š” ์ธ์ง€์™€ ์‹คํ–‰์˜ ๊ทผ๋ณธ์  ๋ถ„๋ฆฌ ๋ฌธ์ œ๋ฅผ IOR์ด๋ผ๋Š” ์ƒˆ๋กœ์šด ์ƒ์ง•์  ํ‘œํ˜„์œผ๋กœ ์šฐ์•„ํ•˜๊ฒŒ ํ•ด๊ฒฐํ•œ ํ˜์‹ ์  ์•„ํ‚คํ…์ฒ˜์ด๋ฉฐ, ๋†’์€ ์„ฑ๊ณต๋ฅ ๊ณผ Sim-to-Real ์„ฑ๋Šฅ์œผ๋กœ ๋กœ๋ด‡ ์กฐ์ž‘ ๋ถ„์•ผ์˜ ์ค‘์š”ํ•œ ์ง„์ „์„ ์ œ์‹œํ•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˜์ •