ManiSkill-HAB: A Benchmark for Low-Level Manipulation in Home Rearrangement Tasks

์ €์ž: Arth Shukla, Stone Tao, Hao Su | ๋‚ ์งœ: 2024-12-09 | URL: https://arxiv.org/abs/2412.13211 📄 PDF


Essence

Figure 2

Figure 2: Interact benchmark comparing MS-HAB (ours) with Habitat. Each data point is annotated

MS-HAB๋Š” GPU ๊ฐ€์†ํ™”๋œ Home Assistant Benchmark์˜ ๊ตฌํ˜„์œผ๋กœ, ํ˜„์‹ค์ ์ธ ์ €์ˆ˜์ค€ ์กฐ์ž‘๊ณผ ๋น ๋ฅธ ์‹œ๋ฎฌ๋ ˆ์ด์…˜ ์†๋„(4300 SPS)๋ฅผ ์ง€์›ํ•˜๋ฉฐ ๋Œ€๊ทœ๋ชจ ๋ฐ์ดํ„ฐ์…‹ ์ƒ์„ฑ์„ ์œ„ํ•œ ์ž๋™ํ™”๋œ ๊ถค์  ํ•„ํ„ฐ๋ง ์‹œ์Šคํ…œ์„ ์ œ๊ณตํ•œ๋‹ค.

Motivation

Achievement

Figure 4

Fig. 4 shows the RL and IL policiesโ€™ progressive completion rate. We provide an optimistic upper

How

Figure 3

Figure 3: Renders of low-level, whole-body control policies solving Pick, Place, Open, and Close

Originality

Limitation & Further Study

Evaluation

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

์ดํ‰: MS-HAB๋Š” ํ˜„์‹ค์ ์ธ ์ €์ˆ˜์ค€ ์กฐ์ž‘ ์ œ์–ด, ๊ณ ์† GPU ์‹œ๋ฎฌ๋ ˆ์ด์…˜, ๊ทธ๋ฆฌ๊ณ  ์ž๋™ํ™”๋œ ๋ฐ์ดํ„ฐ ์ƒ์„ฑ์„ ํ†ตํ•ฉํ•˜์—ฌ ๊ฐ€์ •์šฉ ๋กœ๋ด‡ ์กฐ์ž‘ ์—ฐ๊ตฌ์˜ ์ค‘์š”ํ•œ ๋ฒค์น˜๋งˆํฌ๋ฅผ ์ œ๊ณตํ•˜๋ฉฐ, ๊ด‘๋ฒ”์œ„ํ•œ ๊ธฐ๋ฐ˜์„ ๊ณผ ํˆฌ๋ช…ํ•œ ํ‰๊ฐ€ ์ง€ํ‘œ๋Š” ํ›„์† ์—ฐ๊ตฌ์— ํฐ ๊ฐ€์น˜๋ฅผ ์ 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˜์ •