Large VLM-based Vision-Language-Action Models for Robotic Manipulation: A Survey

์ €์ž: Rui Shao, Wei Li, Lingsen Zhang, Renshan Zhang, Zhiyang Liu, Ran Chen, Liqiang Nie | ๋‚ ์งœ: 2025-08-18 | URL: https://arxiv.org/abs/2508.13073 📄 PDF


Essence

Figure 2

Fig. 2: Outline of the organization of our comprehensive survey (top) and a chronological timeline of notable developmen

๋Œ€๊ทœ๋ชจ Vision-Language Model(VLM)์„ ๊ธฐ๋ฐ˜์œผ๋กœ ํ•œ Vision-Language-Action(VLA) ๋ชจ๋ธ๋“ค์„ ๋กœ๋ด‡ ๋งค๋‹ˆํ“ฐ๋ ˆ์ด์…˜์— ์ ์šฉํ•˜๋Š” ์—ฐ๊ตฌ์˜ ์ฒซ ๋ฒˆ์งธ ์ฒด๊ณ„์  ์„ค๋ฌธ์กฐ์‚ฌ๋กœ, Monolithic ๋ชจ๋ธ๊ณผ Hierarchical ๋ชจ๋ธ์ด๋ผ๋Š” ๋‘ ๊ฐ€์ง€ ์ฃผ์š” ์•„ํ‚คํ…์ฒ˜ ํŒจ๋Ÿฌ๋‹ค์ž„์„ ์ œ์‹œํ•œ๋‹ค.

Motivation

Achievement

Figure 3

Fig. 3: Comparison of the two principal categories of large VLM-based VLA models. Monolithic models (Sec. 3) integrate

How

Figure 2

Fig. 2: Outline of the organization of our comprehensive survey (top) and a chronological timeline of notable developmen

Originality

Limitation & Further Study

Evaluation

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

์ดํ‰: ๋ณธ ์„ค๋ฌธ์กฐ์‚ฌ๋Š” ๋น ๋ฅด๊ฒŒ ์„ฑ์žฅํ•˜๋Š” VLM ๊ธฐ๋ฐ˜ VLA ๋ถ„์•ผ์˜ ์ฒซ ๋ฒˆ์งธ ์ฒด๊ณ„์  ์ข…ํ•ฉ์œผ๋กœ, ๋ช…ํ™•ํ•œ ์ •์˜, ์ผ๊ด€๋œ ๋ถ„๋ฅ˜์ฒด๊ณ„, ๊ทธ๋ฆฌ๊ณ  ํฌ๊ด„์  ๋ถ„์„์„ ํ†ตํ•ด ํ•™๊ณ„์˜ ์—ฐ๊ตฌ ๋‹จํŽธํ™”๋ฅผ ํ•ด์†Œํ•˜๊ณ  ํ–ฅํ›„ ๋ฐœ์ „ ๋ฐฉํ–ฅ์„ ์ œ์‹œํ•˜๋Š” ์˜์˜๊ฐ€ ํฌ๋‹ค. ์ •๊ธฐ์  ์—…๋ฐ์ดํŠธ ๊ณ„ํš๋„ ๋ถ„์•ผ์˜ ๋น ๋ฅธ ์ง„์ „์„ ๋ฐ˜์˜ํ•˜๋Š” ๊ฐ•์ ์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˜์ •