NeurIPS 2026 Spotlight: FAME for Open-Environment Active 3D Mapping
Overview
Tianjin University and Hefei University of Technology propose FAME, using Conditional Flow Matching to generate multiple exploration targets, enhancing robot mapping efficiency in unknown environments. Outperforms existing methods on benchmarks like AiMDoom and Matterport3D. Latest: FAME introduces CFM to create diverse exploration hypotheses, combining path planning and hierarchical selection for better exploration routes in active 3D mapping.
Generated from attributed reports · Updated 30 minutes ago
Event evidence and corrections
1 attributed source owners. Ownership does not establish independent confirmation. Quantities are reported separately and are never added together.
Reported quantity · Reported completion rate: 82.67 percent · Basis not reported · Differing source assertions
“表1：AiMDoom 数据集评估结果，四个难度级别下 FAME 的最终覆盖率与 AUC 均领先于对比方法 在基于真实室内扫描数据构建的 Matterport3D 基准上，FAME 的重建完成率达到 82.67% ，同样优于其他对比方法。”
Exact source · revision 1jiqizhixin
Reported quantity · Reported completion rate: 68.45 percent · Basis not reported · Differing source assertions
“在这一更具挑战性的设置下，FAME 的重建完成率仍达到 68.45% ，相比 NBP 的 61.81% 提高了 6.64 个百分点。”
Exact source · revision 1jiqizhixin
“在这一更具挑战性的设置下，FAME 的重建完成率仍达到 68.45% ，相比 NBP 的 61.81% 提高了 6.64 个百分点。”
Exact source · revision 2Source owner not reported
Artifact availability · paper: available
“论文标题： All Roads Lead to Rome: Flow-driven Multi-Anchor Exploration for Open-Environment Active 3D Mapping 项目主页：https://vollichor.github.io/FAME/ 论文链接： https://arxiv.org/abs/2609.36889 从「只选一条路」，到保留多种探索可能 主动三维建图（Active 3D Mapping） 不仅要求机器人根据观测恢复环境的三维结构，还需要主动决定下一步去哪里采集信息，争取用更少的运动和观测代价，重建尽可能完整的场景。”
Exact source · revision 1jiqizhixin
Reported quantity · Reported completion rate: 82.67 percent · Basis not reported · Differing source assertions
“在基于真实室内扫描数据构建的 Matterport3D 基准上，FAME 的重建完成率达到 82.67% ，同样优于其他对比方法。”
Exact source · revision 2Source owner not reported
Evidence dependency changes (8)
- 2026-10-30T18:54:21.461Z · evidence updated · source revision 1. Evidence extraction was updated; current source attributions are shown above.
- 2026-10-30T18:53:04.013Z · evidence updated · source revision 2. Evidence extraction was updated; current source attributions are shown above.
- 2026-10-30T18:17:57.950Z · evidence updated · source revision 1. Evidence extraction was updated; current source attributions are shown above.
- 2026-10-30T18:15:07.157Z · evidence updated · source revision 2. Evidence extraction was updated; current source attributions are shown above.
- 2026-10-30T17:37:36.747Z · evidence updated · source revision 1. Evidence extraction was updated; current source attributions are shown above.
- 2026-10-30T17:37:36.669Z · evidence updated · source revision 2. Evidence extraction was updated; current source attributions are shown above.
- 2026-10-30T17:14:36.494Z · evidence updated · source revision 1. Evidence extraction was updated; current source attributions are shown above.
- 2026-10-30T17:14:09.696Z · evidence updated · source revision 2. Evidence extraction was updated; current source attributions are shown above.
Report timeline
Follow attributed reports and material updates.
- Chinese robotics — Dataset and collection discoverySignalNeurIPS 2026 Spotlight｜Why Can't Robots Just Follow One Path When Exploring the Unknown? Tianjin University Team Proposes FAME
A team from Tianjin University and Hefei University of Technology proposes FAME (Flow-driven Anchor-based Multimodal Exploration), a new method for active 3D mapping.
- Machine Heart — Robotics on WeChatSignalNeurIPS 2026 Spotlight｜Why Can't Robots Just Follow One Path? Tianjin University Team Proposes FAME
A team from Tianjin University and Hefei University of Technology proposes FAME (Flow-driven Anchor-based Multimodal Exploration), which generates multiple exploration targets through Conditional Flow Matching and combines path planning with hierarchical selection strategies to enhance robot exploration efficiency and map reconstruction quality in unknown environments. The method outperforms existing approaches on AiMDoom and Matterport3D benchmark datasets.
Event coverage history
There is not enough continuous observation data to show a trend.
Timezone · UTC
Article dates follow your selected timezone. Briefing editions use Hong Kong time (UTC+8).