BROWSE
Detail
Conference Paper
Isaac Lab 기반 깊이 영상을 통한 사족로봇 장애물 회피 네비게이션 강화학습 환경
Cited - time in 
Share 
Share
- Issue Date
- 2026-07
- Citation
- 제어·로봇·시스템학회 학술 대회 (ICROS) 2026, pp.415-416
- Publisher
- 제어·로봇·시스템학회
- Language
- Korean
- Type
- Conference Paper
- Abstract
- This paper presents a reinforcement learning environment for depth-image-based obstacle-avoiding navigation of quadruped robots in Isaac Lab. A pretrained blind locomotion policy is reused as a frozen low-level controller, and a high-level policy learns velocity commands from a depth image and a goal vector. Experiments with ANYmal-C show stable obstacle avoidance and a final mean reward of 84.3, demonstrating the efficiency of the proposed hierarchical structure.
- Keyword
- Quadruped Navigation, Vision-Based RL, Isaac Lab
- KSP Keywords
- Depth image, Hierarchical structure, Image-based, Learning Environment, Obstacle Avoidance, Quadruped Robot, Vision-based, obstacle-avoiding, reinforcement learning(RL)