Perpetual Humanoid Control for Real-time Simulated Avatars

์ €์ž: Zhengyi Luo, Jinkun Cao, Alexander Winkler, Kris Kitani, Weipeng Xu | ๋‚ ์งœ: 2023-05-10 | URL: https://arxiv.org/abs/2305.06456 📄 PDF


Essence

Figure 1

Figure 1: We propose a motion imitator that can naturally recover from falls and walk to far-away reference motion, perp

Physics ๊ธฐ๋ฐ˜ humanoid controller์ธ Perpetual Humanoid Controller (PHC)๋Š” noisy input๊ณผ unexpected falls์— ๊ฐ•๊ฑดํ•˜๋ฉด์„œ 10,000๊ฐœ์˜ motion clips์„ ํ•™์Šตํ•  ์ˆ˜ ์žˆ์œผ๋ฉฐ, ์ƒˆ๋กœ์šด Progressive Multiplicative Control Policy (PMCP)๋ฅผ ํ†ตํ•ด catastrophic forgetting ์—†์ด ๋Œ€๊ทœ๋ชจ motion database์—์„œ ํ•™์Šต ๊ฐ€๋Šฅํ•˜๋‹ค.

Motivation

Achievement

Figure 4

Figure 4: (a) Imitating high-quality MoCap โ€“ spin and kick. (b) Recover from fallen state and go back to reference motio

How

Figure 2

Figure 2: Our progressive training procedure to train primitives P(1), P(2), ยท ยท ยท , P(K) by gradually learning harder a

Originality

Limitation & Further Study

Evaluation

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

์ดํ‰: ์ด ๋…ผ๋ฌธ์€ external force ์ œ๊ฑฐ์™€ PMCP๋ผ๋Š” novel mechanism์œผ๋กœ physics-based motion imitation์˜ scalability ๋ฌธ์ œ๋ฅผ ํšจ๊ณผ์ ์œผ๋กœ ํ•ด๊ฒฐํ•˜๋ฉฐ, natural fail-state recovery์™€ noisy input ๊ฐ•๊ฑด์„ฑ์œผ๋กœ ์‹ค์ œ video ๊ธฐ๋ฐ˜ avatar application์— ์ฒ˜์Œ์œผ๋กœ ์‹ค์šฉ์ ์ธ solution์„ ์ 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˜์ •