Learning Adaptive Neural Teleoperation for Humanoid Robots: From Inverse Kinematics to End-to-End Control

์ €์ž: Sanjar Atamuradov | ๋‚ ์งœ: 2025-11-15 | URL: https://arxiv.org/abs/2511.12390 📄 PDF


Essence

Figure 1

Figure 1: Neural teleoperation policy architecture. The network takes VR controller poses (14-dim), joint states (28-

VR ํ…”๋ ˆ์˜คํผ๋ ˆ์ด์…˜์—์„œ ์ „ํ†ต์ ์ธ IK+PD ํŒŒ์ดํ”„๋ผ์ธ์„ RL ๊ธฐ๋ฐ˜ ์‹ ๊ฒฝ๋ง ์ •์ฑ…์œผ๋กœ ๋Œ€์ฒดํ•˜์—ฌ ํž˜ ์ ์‘, ๊ถค์  ๋ถ€๋“œ๋Ÿฌ์›€, ์‚ฌ์šฉ์ž ์ ์‘์„ ๋™์‹œ์— ๋‹ฌ์„ฑํ•˜๋Š” ํ•™์Šต ๊ธฐ๋ฐ˜ ํ”„๋ ˆ์ž„์›Œํฌ๋ฅผ ์ œ์•ˆํ•œ๋‹ค.

Motivation

Achievement

Figure 3

Figure 3 provides a detailed breakdown of performance

How

Figure 1

Figure 1: Neural teleoperation policy architecture. The network takes VR controller poses (14-dim), joint states (28-

Originality

Limitation & Further Study

Evaluation

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

์ดํ‰: ํ•™์Šต ๊ธฐ๋ฐ˜ ์‹ ๊ฒฝ๋ง ์ •์ฑ…์œผ๋กœ VR ํ…”๋ ˆ์˜คํผ๋ ˆ์ด์…˜์˜ ๊ทผ๋ณธ์  ํ•œ๊ณ„๋ฅผ ํ•ด๊ฒฐํ•˜๊ณ  ๋ช…ํ™•ํ•œ ์„ฑ๋Šฅ ํ–ฅ์ƒ์„ ๋ณด์—ฌ์ฃผ๋Š” ์‹ค์งˆ์ ์œผ๋กœ ๊ฐ€์น˜ ์žˆ๋Š” ์—ฐ๊ตฌ์ด๋ฉฐ, ๋ชจ๋ฐฉ ํ•™์Šต๊ณผ ๊ต๊ณผ ํ•™์Šต์˜ ์กฐํ•ฉ ์„ค๊ณ„๊ฐ€ ์šฐ์ˆ˜ํ•˜๋‹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˜์ •