HumanoidVLM: Vision-Language-Guided Impedance Control for Contact-Rich Humanoid Manipulation

์ €์ž: Yara Mahmoud, Yasheerah Yaqoot, Miguel Altamirano Cabrera, Dzmitry Tsetserukou | ๋‚ ์งœ: 2026-01-21 | URL: https://arxiv.org/abs/2601.14874 📄 PDF


Essence

Figure 1

Figure 1: Overall architecture of the proposed HumanoidVLM framework.

HumanoidVLM์€ vision-language model๊ณผ retrieval-augmented generation์„ ๊ฒฐํ•ฉํ•˜์—ฌ ํœด๋จธ๋…ธ์ด๋“œ ๋กœ๋ด‡์ด egocentric ์ด๋ฏธ์ง€๋กœ๋ถ€ํ„ฐ task-specific impedance parameters์™€ gripper configuration์„ ์ž๋™์œผ๋กœ ์„ ํƒํ•˜๋Š” ์ ์‘ํ˜• ์กฐ์ž‘ ํ”„๋ ˆ์ž„์›Œํฌ์ด๋‹ค.

Motivation

Achievement

Figure 3

Figure 3: Retrieval accuracy of the VLMโ€“RAG system across

How

Figure 1

Figure 1: Overall architecture of the proposed HumanoidVLM framework.

Originality

Limitation & Further Study

Evaluation

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

์ดํ‰: ๋ณธ ๋…ผ๋ฌธ์€ VLM๊ณผ RAG๋ฅผ humanoid manipulation์— ํšจ๊ณผ์ ์œผ๋กœ ์ ์šฉํ•˜์—ฌ semantic perception๊ณผ compliant control์„ ์ฒ˜์Œ ์ฒด๊ณ„์ ์œผ๋กœ ์—ฐ๊ฒฐํ–ˆ์œผ๋ฉฐ, ๋†’์€ retrieval ์ •ํ™•๋„์™€ ์‹ค์ œ ๋กœ๋ด‡ ์‹คํ—˜์„ ํ†ตํ•ด ํƒ€๋‹น์„ฑ์„ ์ž…์ฆํ–ˆ๋‹ค. ๋‹ค๋งŒ ๊ณ ์ •๋œ database ๊ทœ๋ชจ์™€ sensor ์ œ์•ฝ์ด ํ–ฅํ›„ ํ™•์žฅ์„ฑ์„ ์ œํ•œํ•˜๋Š” ์ ์ด ๊ฐœ์„  ๋Œ€์ƒ์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˜์ •