BROWSE
Detail
Conference Paper
협소 환경에서 차륜형 로봇의 고속 주행 성능 향상을 위한 단일 정책 PPO 기반 안전 내비게이션
Cited - time in 
Share 
Share
- Issue Date
- 2026-07
- Citation
- 제어로봇시스템학회 학술 대회 (ICROS) 2026, pp.1019-1020
- Publisher
- 제어로봇시스템학회
- Language
- Korean
- Type
- Conference Paper
- Abstract
- This paper proposes a single-policy PPO-based navigation framework for safe wheeled robot navigation in narrow environments. To this end, the Agile But Safe concept is adapted to the Clearpath Jackal platform, and its effects are examined through a systematic analysis of design factors that influence navigation performance and recovery in cluttered spaces. The proposed method replaces the original joint-level action structure with a velocity-command-based policy suitable for wheeled robots, and uses 31-dimensional range observations and goal-relative states as input. We analyze the effects of key design factors, including observation configuration, action range, and episode length. Simulation results show that the proposed method maintains a low collision rate while achieving reliable goal-reaching performance in narrow environments. These findings suggest that the proposed approach provides a practical basis for improving high-speed navigation performance in cluttered environments.
- Keyword
- Wheeled Robot, Navigation, Deep Reinforcement Learning, PPO, Narrow Environment
- KSP Keywords
- Cluttered environment, Deep reinforcement learning, Design factors, High-speed, Key design, Navigation performance, Robot Navigation, Systematic analysis, Wheeled Robot, action structure, collision rate