Plan-Seq-Learn: Language Model Guided RL for Solving Long Horizon Robotics Tasks

์ €์ž: Murtaza Dalal, Tarun Chiruvolu, Devendra Chaplot, Ruslan Salakhutdinov | ๋‚ ์งœ: 2024-05-02 | URL: https://arxiv.org/abs/2405.01534 📄 PDF


Essence

Figure 2

Figure 2: Method overview. PSL decomposes tasks into a list of regions and stage termination conditions

Plan-Seq-Learn (PSL)์€ LLM์˜ ๊ณ ์ˆ˜์ค€ ๊ณ„ํš, motion planning์˜ ์‹œํ€€์‹ฑ, RL์˜ ์ €์ˆ˜์ค€ ์ œ์–ด ํ•™์Šต์„ ํ†ตํ•ฉํ•˜์—ฌ ์‚ฌ์ „ ์ •์˜๋œ ์Šคํ‚ฌ ๋ผ์ด๋ธŒ๋Ÿฌ๋ฆฌ ์—†์ด ์žฅ์‹œ๊ฐ„ ๋กœ๋ด‡ ์ž‘์—…์„ ํ•ด๊ฒฐํ•œ๋‹ค.

Motivation

Achievement

Figure 3

Figure 3: Sample Efficiency Results. We plot task success rate as a function of the number of trials. PSL

How

Figure 2

Figure 2: Method overview. PSL decomposes tasks into a list of regions and stage termination conditions

Originality

Limitation & Further Study

Evaluation

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

์ดํ‰: PSL์€ LLM, motion planning, RL์˜ ์ƒํ˜ธ ๋ณด์™„์  ๊ฐ•์ ์„ ์ฐฝ์˜์ ์œผ๋กœ ํ†ตํ•ฉํ•˜์—ฌ ์‚ฌ์ „ ์ •์˜๋œ ์Šคํ‚ฌ ์—†์ด ์žฅ์‹œ๊ฐ„ ๋กœ๋ด‡ ์ž‘์—…์„ ํšจ์œจ์ ์œผ๋กœ ํ•ด๊ฒฐํ•˜๋Š” ์‹ค์งˆ์ ์ด๊ณ  ๊ฐ•๋ ฅํ•œ ๋ฐฉ๋ฒ•์„ ์ œ์‹œํ•œ๋‹ค. ๊ด‘๋ฒ”์œ„ํ•œ ์‹คํ—˜๊ณผ ๋ช…ํ™•ํ•œ ์„ค๋ช…์œผ๋กœ ๋†’์€ ๊ฐ€์น˜์˜ ๊ธฐ์—ฌ๋ฅผ ์ž…์ฆํ•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˜์ •