MobileH2R: Learning Generalizable Human to Mobile Robot Handover Exclusively from Scalable and Diverse Synthetic Data

์ €์ž: Zifan Wang, Ziqing Chen, Junyu Chen, Jilong Wang, Yuxin Yang, Yunze Liu, Xueyi Liu, He Wang, Li Yi | ๋‚ ์งœ: 2025-01-08 | URL: https://arxiv.org/abs/2501.04595 📄 PDF


Essence

Figure 1

Figure 1. The overview of MobileH2R. We propose a framework for generalizable human-to-mobile-robot handover, including

MobileH2R๋Š” ๋Œ€๊ทœ๋ชจ ๋‹ค์–‘ํ•œ ํ•ฉ์„ฑ ๋ฐ์ดํ„ฐ๋งŒ์„ ์‚ฌ์šฉํ•˜์—ฌ ๋ชจ๋ฐ”์ผ ๋กœ๋ด‡์ด ์ธ๊ฐ„์œผ๋กœ๋ถ€ํ„ฐ ๋ฌผ์ฒด๋ฅผ ๋ฐ›์„ ์ˆ˜ ์žˆ๋„๋ก ํ•™์Šตํ•˜๋Š” ํ”„๋ ˆ์ž„์›Œํฌ๋ฅผ ์ œ์‹œํ•œ๋‹ค. ์ธ๊ฐ„์˜ ์ „์‹  ๋™์ž‘ ์ƒ์„ฑ, ์•ˆ์ „ํ•œ ์‹œ์—ฐ ์ž๋™ ์ƒ์„ฑ, 4D imitation learning์„ ํ†ตํ•ฉํ•˜์—ฌ ๋ฒ ์ด์Šค-์•” ํ˜‘์กฐ ์ œ์–ด๊ฐ€ ๊ฐ€๋Šฅํ•œ ์ผ๋ฐ˜ํ™”๋œ ์ •์ฑ…์„ ํ•™์Šตํ•œ๋‹ค.

Motivation

Achievement

Figure 4

Figure 4. Qualitative results. We compare different methods in detail in the simulated scene and the real-world scene.

How

Figure 2

Figure 2. The overview of our framework. First, we propose an automatic pipeline to scale up synthetic and diverse full-

Originality

Limitation & Further Study

Evaluation

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

์ดํ‰: MobileH2R๋Š” ๋ชจ๋ฐ”์ผ ๋กœ๋ด‡์˜ ์ธ๊ฐ„-๋กœ๋ด‡ handover ๋ฌธ์ œ๋ฅผ ์ฒด๊ณ„์ ์œผ๋กœ ํ•ด๊ฒฐํ•˜๋Š” ํฌ๊ด„์ ์ด๊ณ  ํ™•์žฅ ๊ฐ€๋Šฅํ•œ ํ”„๋ ˆ์ž„์›Œํฌ๋ฅผ ์ œ์‹œํ•œ๋‹ค. ํ•ฉ์„ฑ ๋ฐ์ดํ„ฐ์˜ ์ƒ์„ฑ, ์•ˆ์ „ํ•œ ์‹œ์—ฐ ์ž๋™ ์ƒ์„ฑ, ํ†ตํ•ฉ ํ•™์Šต์ด๋ผ๋Š” ์„ธ ์š”์†Œ๋ฅผ ์ •๊ตํ•˜๊ฒŒ ์„ค๊ณ„ํ•˜์—ฌ +15% ์ด์ƒ์˜ ์„ฑ๋Šฅ ํ–ฅ์ƒ์„ ๋‹ฌ์„ฑํ–ˆ์œผ๋ฉฐ, ๋Œ€๊ทœ๋ชจ ๋ฐ์ดํ„ฐ์˜ ํšจ๊ณผ๋ฅผ ์‹ค์ฆํ•œ ์ ์—์„œ ์‹ค๋ฌด์  ๊ฐ€์น˜๊ฐ€ ๋†’๋‹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˜์ •