A curated list of awesome spatial exploration policy papers, inspired by awesome-computer-vision and awesome-NeRF.
We classify the primary technical frameworks of existing NBV policies into the following categories: 1) information gain (a.k.a. uncertainty), 2) reinforcement learning, 3) regression/recurrent prediction.
The exploration task, also known as next-best-view planning and active mapping, aims to determine the next best viewpoints to better model the target objects or scenes. The downstream tasks include: 3D reconstruction, neural rendering, SLAM, visual localization, object understanding, and robotic manipulation.
- A Survey on Active Simultaneous Localization and Mapping: State of the Art and New Frontiers], Ahmed et al., T-RO 2023 | Paper
- A Review on Viewpoints and Path Planning for UAV-Based 3-D Reconstruction, Maboudi et al., IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2023 | Paper
- Active SLAM: A Review On Last Decade, Ahmed et al., Sensors, 2023 | Paper
- GLEAM: Learning Generalizable Exploration Policy for Active Mapping in Complex 3D Indoor Scenes, Chen et al., ICCV 2025 | Paper | Project | Code
- Hestia: Voxel-Face-Aware Hierarchical Next-Best-View Acquisition for Efficient 3D Reconstruction, Lu et al., WACV 2026 | Paper | Project | Code
- Learning to Explore using Active Neural SLAM, Chaplot et al., ICLR 2020 | Paper | Code
- Learning Exploration Policies for Navigation, Chen et al., ICLR 2019 | Paper | Code
- NextBestPath: Efficient 3D Mapping of Unseen Environments, Li et al, ICLR 2025 | Paper | Project
- ActiveGS: Active Scene Reconstruction Using Gaussian Splatting, Jin et al., RA-L 2025 | Paper
- Active Neural Mapping at Scale, Kuang et al., IROS 2024 | Paper
- Active Neural Mapping, Yan et al., ICCV 2023 | Paper | Project
- Continual Neural Mapping: Learning an Implicit Scene Representation from Sequential Observations, Yan et al., ICCV 2021 | Paper | Project
- Multi-robot collaborative dense scene reconstruction, Dong et al., TOG 2019 | Paper
- Density-aware NeRF Ensembles: Quantifying Predictive Uncertainty in Neural Radiance Fields, Sünderhauf et al., arXiv 2022 | Paper
- Where to Explore Next? ExHistCNN for History-aware Autonomous 3D Exploration, Wang et al., ECCV 2020 | Paper
- GenNBV: Generalizable Next-Best-View Policy for Active 3D Reconstruction, Chen et al., CVPR 2024 | Paper | Project | Code
- Next-Best View Policy for 3D Reconstruction, Peralta et al., ECCV 2020 Workshop | Paper | Code
- Semantically Informed Next Best View Planning for Autonomous Aerial 3D Reconstruction, Kay et al., IROS 2021 | Paper
- Next-Best-View planning for surface reconstruction of large-scale 3D environments with multiple UAVs, Hardouin et al., IROS 2020 | Paper
- Learn-to-Score: Efficient 3D Scene Exploration by Predicting View Utility, et al., ECCV 2018 | Paper
- RL-NBV: A deep reinforcement learning based next-best-view method for unknown object reconstruction, Wang et al., Pattern Recognition 2024 | Paper
- Bag of Views: An Appearance-based Approach to Next-Best-View Planning for 3D Reconstruction, Gazani et al., arXiv 2023 | Paper
- Learning to Select Camera Views: Efficient Multiview Understanding at Few Glances, Hou et al., CVPR 2024 | Paper | Code
- A Reinforcement Learning Approach to the View Planning Problem, Kaba et al., CVPR 2017 | Paper
- Beyond Uncertainty: Risk-Aware Active View Acquisition for Safe Robot Navigation and 3D Scene Understanding with FisherRF, Liu et al., arXiv 2024 | Paper
- FisherRF: Active View Selection and Uncertainty Quantification for Radiance Fields using Fisher Information, Jiang et al., ECCV 2024 | Paper | Code
- Active Implicit Reconstruction Using One-Shot View Planning, Hu et al., ICRA 2024 Paper
- SO-NeRF: Active View Planning for NeRF using Surrogate Objectives, Lee et al., arXiv 2023 | Paper
- NeU-NBV: Next Best View Planning Using Uncertainty Estimation in Image-Based Neural Rendering, Jin et al., IROS 2023 | Paper | Code
- A Global Generalized Maximum Coverage-based Solution to the Non-model-based View Planning Problem for Object Reconstruction, Pan et al., CVIU 2023 | Paper
- ActiveRMAP: Radiance Field for Active Mapping And Planning, Zhan et al., arXiv 2022 | Paper
- Uncertainty-Driven Active Vision for Implicit Scene Reconstruction, Smith et al., arXiv 2022 | Paper | Code
- ActiveNeRF: Learning where to See with Uncertainty Estimation, Pan et al., ECCV 2022 | Paper
- Uncertainty Guided Policy for Active Robotic 3D Reconstruction using Neural Radiance Fields, Lee et al., R-AL 2022 | Paper | Video
- PC-NBV: A Point Cloud Based Deep Network for Efficient Next Best View Planning, Zeng et al., IROS 2020 | Paper | Code
- Multi-Sensor Next-Best-View Planning as Matroid-Constrained Submodular Maximization, Lauri et al., RA-L 2020 | Paper
- A comparison of volumetric information gain metrics for active 3D object reconstruction, Delmerico et al., Autonomous Robots 2017 | Paper
- An information gain formulation for active volumetric 3D reconstruction, Isler et al., ICRA 2016 | Paper | Project
- Active Object Reconstruction Using a Guided View Planner, Yang et al., IJCAI 2018 | Paper
- Geometry-Aware Recurrent Neural Networks for Active Visual Recognition, Cheng et al., NIPS 2018 | Paper
MIT
