arXiv · Robotics· Yang Li, Aming Wu, Zihao Zhang, Ziju Han, Sijia Zhang, Yahong Han·· 7 小时前AI 评分15
All Roads Lead to Rome: Flow-driven Multi-Anchor Exploration for Open-Environment Active 3D Mapping
All Roads Lead to Rome: Flow-driven Multi-Anchor Exploration for Open-Environment Active 3D Mapping
AI 导读
提出基于 Conditional Flow 的多锚点生成方法，将长程目标预测改为条件多模态锚点生成，再经障碍感知规划转为候选路径。引入探索模式聚类压缩几何相似轨迹，并通过层次选择模块重排序最终轨迹。在开放环境中提升泛化性与重建效率，被 NeurIPS 2026 接收。
来源：arXiv · Robotics · arxiv.org