RoboTwin 2.0: A Scalable Data Generator and Benchmark with Strong Domain Randomization for Robust Bimanual Robotic Manipulation

์ €์ž: Tianxing Chen, Zanxin Chen, Baijun Chen, Zijian Cai, Yibin Liu, Zixuan Li, Qiwei Liang, Xianliang Lin, Yiheng Ge, Zhenyu Gu, Weiliang Deng, Yubin Guo, Tian Nian, Xuanbing Xie, Qiangyu Chen, Kailun Su, Tianling Xu, Guodong Liu, Mengkang Hu, Huan-ang Gao, Kaixuan Wang, Zhixuan Liang, Yusen Qin, Xiaokang Yang, Ping Luo, Yao Mu | ๋‚ ์งœ: 2025-06-22 | URL: https://arxiv.org/abs/2506.18088 📄 PDF


Essence

Figure 1

Figure 1: Overview of RoboTwin 2.0. RoboTwin 2.0 is a scalable framework for bimanual manipu-

RoboTwin 2.0๋Š” MLLM ๊ธฐ๋ฐ˜ ์ž๋™ ์ฝ”๋“œ ์ƒ์„ฑ๊ณผ ์‹œ๋ฎฌ๋ ˆ์ด์…˜ ์ธ๋ฃจํ”„ ํ”ผ๋“œ๋ฐฑ์„ ํ™œ์šฉํ•˜์—ฌ ๋Œ€๊ทœ๋ชจ ์ด์›ํŒ” ์กฐ์ž‘ ๋ฐ์ดํ„ฐ๋ฅผ ์ƒ์„ฑํ•˜๋Š” ํ™•์žฅ ๊ฐ€๋Šฅํ•œ ํ”„๋ ˆ์ž„์›Œํฌ์ด๋ฉฐ, ๊ตฌ์กฐํ™”๋œ domain randomization์„ ํ†ตํ•ด sim-to-real ์ „์ด๋ฅผ ํฌ๊ฒŒ ํ–ฅ์ƒ์‹œํ‚จ๋‹ค.

Motivation

Achievement

Figure 1

Figure 1: Overview of RoboTwin 2.0. RoboTwin 2.0 is a scalable framework for bimanual manipu-

How

Figure 2

Figure 2: RoboTwin 2.0 Pipeline. Built on RoboTwin-OD and a skill API, the framework uses

Originality

Limitation & Further Study

Evaluation

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

์ดํ‰: RoboTwin 2.0๋Š” MLLM ๊ธฐ๋ฐ˜ ์ž๋™ ์ฝ”๋“œ ์ƒ์„ฑ, ํ๋ฃจํ”„ ํ”ผ๋“œ๋ฐฑ, ๋‹ค์ถ• domain randomization, ์ฒด๊ตฌ ํŠนํ™” ์ ์‘์„ ๊ฒฐํ•ฉํ•˜์—ฌ ์ด์›ํŒ” ์กฐ์ž‘ ์—ฐ๊ตฌ์˜ ์ค‘์š”ํ•œ ๊ธฐ๋ฐ˜์„ ์ œ๊ณตํ•˜๋ฉฐ, 367% sim-to-real ๊ฐœ์„ ๊ณผ ๊ณต๊ฐœ ์ž์‚ฐ/์ฝ”๋“œ๋กœ ๋†’์€ ์‹ค์šฉ์„ฑ์„ ๋ณด์—ฌ์ค€๋‹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˜์ •