LfH-CP 数据集
收藏资源简介:
LfH-CP 数据集是用于动态环境导航的动态障碍物轨迹丰富数据集。该数据集由 Learning from Hallucinating Critical Points (LfH-CP) 框架生成,该框架基于现有的最优运动计划,无需昂贵的专家演示或试错探索。LfH-CP 将幻觉分解为两个阶段:首先识别障碍物必须出现的时间和地点,以便产生最优运动计划,即关键点,然后生成通过这些点的多样化轨迹,同时避免碰撞。LfH-CP 旨在为学习运动规划器提供丰富和多样化的训练数据,以提高动态环境中的导航性能。
The LfH-CP dataset is a rich dataset of dynamic obstacle trajectories for navigation in dynamic environments. It is generated by the Learning from Hallucinating Critical Points (LfH-CP) framework, which leverages existing optimal motion plans and eliminates the need for costly expert demonstrations or trial-and-error exploration. LfH-CP decomposes the hallucination process into two stages: first, identify the exact time and position where obstacles must emerge, which serve as the critical points for generating optimal motion plans; then, generate diverse trajectories that pass through these points while avoiding collisions. The framework aims to provide rich and diverse training data for learned motion planners, thereby enhancing navigation performance in dynamic environments.

- 1Learning from Hallucinating Critical Points for Navigation in Dynamic Environments乔治梅森大学 · 2025年



