A Survey of Robotic Navigation and Manipulation with Physics Simulators in the Era of Embodied AI

์ €์ž: Lik Hang Kenny Wong, Xueyang Kang, Kaixin Bai, Jianwei Zhang | ๋‚ ์งœ: 2025-05-01 | URL: https://arxiv.org/abs/2505.01458 📄 PDF


Essence

Figure 2

Fig. 2. A taxonomy of this survey, focusing on two main tasks of Embodied AI: Navigation and Manipulation. We discuss th

๋ณธ ๋…ผ๋ฌธ์€ Embodied AI ์‹œ๋Œ€์— ๋กœ๋ด‡์˜ ๋„ค๋น„๊ฒŒ์ด์…˜๊ณผ ์กฐ์ž‘ ์ž‘์—…์„ ์œ„ํ•œ Physics Simulator์˜ ์—ญํ• ์„ ์ข…ํ•ฉ์ ์œผ๋กœ ๋ถ„์„ํ•˜๊ณ , sim-to-real ์ „์ด์˜ ๊ฐ„๊ทน์„ ์ขํžˆ๊ธฐ ์œ„ํ•œ ์‹œ๋ฎฌ๋ ˆ์ดํ„ฐ ์†์„ฑ, ๋ฒค์น˜๋งˆํฌ, ํ‰๊ฐ€ ์ง€ํ‘œ ๋ฐ ์ตœ์‹  ๋ฐฉ๋ฒ•๋ก ์„ ์ œ์‹œํ•œ๋‹ค.

Motivation

Achievement

Figure 1

Fig. 1. Timeline illustrating the evolution of navigation (top) and manipulation (bottom) research in Embodied AI from

How

Figure 3

Fig. 3. This diagram outlines the four key steps in navigation tasksโ€”Perception, Memory Building, Decision-Making, and

Originality

Limitation & Further Study

Evaluation

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

์ดํ‰: ๋ณธ ๋…ผ๋ฌธ์€ Embodied AI ์‹œ๋Œ€์˜ navigation๊ณผ manipulation ์—ฐ๊ตฌ๋ฅผ ํฌ๊ด„์ ์œผ๋กœ ์ •๋ฆฌํ•œ ์‹œ์˜์ ์ ˆํ•œ ์„ค๋ฌธ์กฐ์‚ฌ๋กœ, ํ˜„๋Œ€์  simulator ๊ธฐ์ˆ ๊ณผ ์ตœ์‹  ๋ฐฉ๋ฒ•๋ก (world model, geometric equivariance, VLA)์„ ์ฒด๊ณ„์ ์œผ๋กœ ๋ถ„์„ํ•˜์—ฌ ์—ฐ๊ตฌ์ž๋“ค์˜ ๋„๊ตฌ ์„ ํƒ๊ณผ ๋ฐฉ๋ฒ•๋ก  ์„ค๊ณ„์— ์‹ค์งˆ์  ๊ฐ€์ด๋“œ๋ฅผ ์ œ๊ณตํ•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˜์ •