A Hybrid Autoencoder for Robust Heightmap Generation from Fused Lidar and Depth Data for Humanoid Robot Locomotion

์ €์ž: Dennis Bank, Joost Cordes, Thomas Seel, Simon F. G. Ehlers | ๋‚ ์งœ: 2026.02 | DOI: N/A 📄 PDF


Essence

Figure 3

Figure 3. The structure is designed to bridge this gap by ex-

์ด ๋…ผ๋ฌธ์€ humanoid robot์˜ unstructured environment ์ด๋™์„ ์œ„ํ•ด LiDAR๊ณผ depth camera ๋ฐ์ดํ„ฐ๋ฅผ fuseํ•˜์—ฌ heightmap์„ ์ƒ์„ฑํ•˜๋Š” hybrid encoder-decoder ์•„ํ‚คํ…์ฒ˜๋ฅผ ์ œ์•ˆํ•œ๋‹ค. CNN ๊ธฐ๋ฐ˜ spatial feature extraction๊ณผ GRU ๊ธฐ๋ฐ˜ temporal consistency๋ฅผ ๊ฒฐํ•ฉํ•œ ์ ‘๊ทผ์œผ๋กœ, multimodal fusion์ด ๋‹จ์ผ ์„ผ์„œ ๋Œ€๋น„ 7.2~9.9% ์žฌ๊ตฌ์„ฑ ์ •ํ™•๋„ ๊ฐœ์„ ์„ ๋‹ฌ์„ฑํ•œ๋‹ค.

Motivation

Achievement

How

Originality

Limitation & Further Study

Evaluation

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

์ดํ‰: ์ด ๋…ผ๋ฌธ์€ multimodal sensor fusion๊ณผ temporal modeling์„ ํ†ตํ•ด humanoid robot์˜ heightmap ์žฌ๊ตฌ์„ฑ ์ •ํ™•๋„๋ฅผ ์ฒด๊ณ„์ ์œผ๋กœ ๊ฐœ์„ ํ•˜๋ฉฐ, spherical projection ๊ธฐ๋ฐ˜ LiDAR ์ฒ˜๋ฆฌ์™€ heightmap ๊ทธ๋ฆฌ๋“œ ํ•ด์ƒ๋„ ์ตœ์ ํ™” ๋“ฑ์˜ ์‹ค์งˆ์  contribution์„ ์ œ๊ณตํ•œ๋‹ค. ๋‹ค๋งŒ ์‹ค์ œ robot platform์—์„œ์˜ locomotion ์„ฑ๋Šฅ ํ–ฅ์ƒ์„ ์ •๋Ÿ‰์ ์œผ๋กœ ์ž…์ฆํ•˜๊ณ , ๋‹ค์–‘ํ•œ ํ™˜๊ฒฝ ๋ฐ ์„ผ์„œ ์กฐํ•ฉ์— ๋Œ€ํ•œ robust์„ฑ์„ ๊ฒ€์ฆํ•ด์•ผ impact๊ฐ€ ๋†’์•„์งˆ ์ˆ˜ ์žˆ๋‹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˜์ •