用神经控制屏障函数提升机器人在复杂地形下的安全导航能力
T-CBF: Traversability-based Control Barrier Function to Navigate Vertically Challenging Terrain
- 基于可通行性设计新型神经控制屏障函数，突破传统避障局限
- 在真实垂直挑战地形上使机器人安全移动率提升30%
- 适合需在非结构化野外环境作业的自主移动机器人研究者
安全性在运动规划与控制中至关重要，近年来备受关注。现有研究多聚焦于碰撞避免，但在非结构化、垂直挑战性的越野地形中，车辆侧翻和陷入同样构成重大安全隐患。本文提出一种基于可通行性的控制屏障函数（T-CBF），利用神经控制屏障函数，通过分析可通行性相关安全因素，在复杂地形中实现超越传统避碰的安全保障。该方法在特定于可通行性安全的正负样本上训练神经T-CBF，并用于生成安全轨迹。实验结果表明，无论在仿真还是物理平台Verti-4 Wheeler（V4W）上，T-CBF均能有效保障可通行性安全并成功抵达目标。相较于以往规划器，其在真实垂直挑战地形上的安全与机动性能提升30%。
原文摘要 · Abstract (English)
Safety has been of paramount importance in motion planning and control techniques and is an active area of research in the past few years. Most safety research for mobile robots target at maintaining safety with the notion of collision avoidance. However, safety goes beyond just avoiding collisions, especially when robots have to navigate unstructured, vertically challenging, off-road terrain, where vehicle rollover and immobilization is as critical as collisions. In this work, we introduce a novel Traversability-based Control Barrier Function (T-CBF), in which we use neural Control Barrier Functions (CBFs) to achieve safety beyond collision avoidance on unstructured vertically challenging terrain by reasoning about new safety aspects in terms of traversability. The neural T-CBF trained on safe and unsafe observations specific to traversability safety is then used to generate safe trajectories. Furthermore, we present experimental results in simulation and on a physical Verti-4 Wheeler (V4W) platform, demonstrating that T-CBF can provide traversability safety while reaching the goal position. T-CBF planner outperforms previously developed planners by 30\% in terms of keeping the robot safe and mobile when navigating on real world vertically challenging terrain.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。