The reviewed record of science sign in
Pith

arxiv: 2605.30342 · v1 · pith:ZXUQYKQ4 · submitted 2026-05-28 · cs.CV · cs.RO

Uncertainty-driven 3D Gaussian Splatting Active Mapping via Anisotropic Visibility Field

Shangjie Xue , Jesse Dill , Dhruv Ahuja , Frank Dellaert , Panagiotis Tsiotras , Danfei Xu This is my paper

Reviewed by Pith T0 review T1 audit T2 compute T3 formal T4 kernel 2026-06-29 07:30 UTCgrok-4.3pith:ZXUQYKQ4record.jsonopen to challenge →

classification cs.CV cs.RO
keywords 3D Gaussian Splattinguncertainty quantificationactive mappingvisibility fieldspherical harmonicsinformation gainBayesian networkneural rendering
0
0 comments X

The pith

A visibility field computed from training views lets 3D Gaussian Splatting quantify uncertainty for new images at 200 frames per second.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper claims that regions unseen during training produce unreliable outputs in 3D Gaussian Splatting models. To quantify this, it defines an anisotropic visibility field for each Gaussian particle relative to the training views and stores the field using spherical harmonics. This field is then fed into a Bayesian network rasterizer that computes uncertainty on the fly. The uncertainty drives view selection in an active mapping loop that maximizes information gain. If correct, the approach yields faster and more accurate mapping than earlier uncertainty methods while remaining compatible with existing Gaussian Splatting pipelines.

Core claim

The central claim is that the anisotropic visibility of each particle with respect to the training views, represented using spherical harmonics, can be integrated into a Bayesian Network-based uncertainty-aware 3DGS rasterizer. This integration produces real-time uncertainty estimates for synthesized views and supports active mapping under a maximum information gain objective.

What carries the argument

The anisotropic visibility field of each Gaussian particle, represented using spherical harmonics and integrated into a Bayesian network rasterizer.

If this is right

  • Uncertainty quantification runs at 200 FPS inside the rasterizer.
  • Active mapping selects next views by maximizing information gain derived from the visibility-based uncertainty.
  • The method outperforms prior uncertainty approaches in both accuracy and speed across tested environments.
  • The visibility field can be applied after training to improve the performance of existing 3DGS models without retraining.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The real-time speed opens the possibility of running active mapping on mobile robots with limited compute.
  • Because the field is stored with spherical harmonics, visibility can be queried from any direction without additional storage.
  • Post-hoc compatibility suggests the technique could serve as a lightweight add-on module for other neural rendering systems.

Load-bearing premise

The assumption that the anisotropic visibility field computed from training views accurately captures the unreliability of predictions in regions unseen from those views.

What would settle it

Compare the uncertainty values predicted by the visibility field against measured reconstruction error on a held-out set of test views; if low-visibility regions do not show systematically higher error, the mapping from visibility to uncertainty does not hold.

Figures

Figures reproduced from arXiv: 2605.30342 by Danfei Xu, Dhruv Ahuja, Frank Dellaert, Jesse Dill, Panagiotis Tsiotras, Shangjie Xue.

Figure 1
Figure 1. Figure 1: GAVIS overview. Gaussian Splatting Anisotropic Visi￾bility Field (GAVIS) quantifies uncertainty in 3DGS by modeling visibility, i.e., whether a region is observed by the training views. Observed regions have low uncertainty (left room), whereas unob￾served regions have high uncertainty (right room). number of trainable parameters in 3DGS, accurately and efficiently quantifying its uncertainty remains chall… view at source ↗
Figure 2
Figure 2. Figure 2: GAVIS framework. (Left) Given a trained 3DGS, GAVIS constructs a visibility field (VF CONST) to represent regions invisible to the training views. It then quantifies uncertainty over sampled candidate views using an uncertainty-aware 3DGS rasterizer (UA 3DGS Rasterizer) that queries the visibility field (VF QUERY). Finally, the maximum-uncertainty view is selected as the next observation. (Right) GAVIS ach… view at source ↗
Figure 3
Figure 3. Figure 3: Schematic of directional visibility. (Top) A 2D illus￾tration of how visibility varies with viewing direction. Anisotropic visibility is plotted in polar coordinates for (1) a single training view, (2) a single training view with occlusion, and (3) multiple training views. (Bottom) A 3D illustration of spherical harmonics expansion of V˜ (d). See Secs. 4.1 and 4.2 for details. visible from a single view). … view at source ↗
Figure 4
Figure 4. Figure 4 [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Qualitative active mapping. Reconstruction results and camera-view distributions (green frustums) from different meth￾ods’ active-mapping trajectories on HST scene (top) and Lego scene (bottom). Full results are provided in Sec. 12 [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Qualitative uncertainty estimation. All methods are trained on the same set of views that only partially cover the scene, leaving some regions underexplored. GT Vis. indicates rasterized ground-truth mesh visibility (binary face labels), where brighter denotes higher uncertainty (invisible faces) and darker denotes lower uncertainty (visible faces). Our method accurately assigns high uncertainty to invisib… view at source ↗
Figure 7
Figure 7. Figure 7: Illustration of the two major components of GAVIS: (1) anisotropic visibility (effects highlighted with gray boxes) and (2) [PITH_FULL_IMAGE:figures/full_fig_p015_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Qualitative results for uncertainty quantification in indoor scenes (Gibson and HM3D). From left to right: (1) Top-down view of the scene with training views (blue frustums) that cover only part of the room, and the queried view for uncertainty evaluation (green frustum); (2) Ground-truth RGB image rendered from the scene mesh; (3) Synthesized RGB image from a 3DGS model trained on the partial-view dataset… view at source ↗
Figure 9
Figure 9. Figure 9: Qualitative results for uncertainty quantification in HST (from space dataset) and Lego (from NeRF Synthetic dataset) scenes. From left to right: (1) Isometric view of the object with training views (blue frustums) that cover only part of the object geometry, and the queried view for uncertainty evaluation (green frustum); (2) Ground-truth RGB image rendered from the scene mesh; (3) Synthesized RGB image f… view at source ↗
Figure 10
Figure 10. Figure 10: Qualitative results for active mapping in all HM3D scenes From left to right: (1) Ground-truth top-down view; Reconstruction results from (2) FisherRF, (3) VIMC, (4) NVF, (5) GAVIS, (6) FisherRF+GAVIS, (7) VIMC+GAVIS. Planned camera poses are shown as green frustums [PITH_FULL_IMAGE:figures/full_fig_p028_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Qualitative results for active mapping in all Gibson scenes From left to right: (1) Ground-truth top-down view; Reconstruction results from (2) FisherRF, (3) VIMC, (4) NVF, (5) GAVIS, (6) FisherRF+GAVIS, (7) VIMC+GAVIS. Planned camera poses are shown as green frustums [PITH_FULL_IMAGE:figures/full_fig_p029_11.png] view at source ↗
read the original abstract

We present Gaussian Splatting Anisotropic Visibility Field (GAVIS), a novel framework for uncertainty quantification and active mapping in 3DGS. Our key insight is that regions unseen from the training views yield unreliable predictions from the 3DGS. To address this, we introduce a principled and efficient method for quantifying the visibility field in 3DGS, defined as the anisotropic visibility of each particle with respect to the training views, and represented using spherical harmonics. The resulting visibility field is integrated into a Bayesian Network-based uncertainty-aware 3DGS rasterizer, enabling real-time (200 FPS) uncertainty quantification for synthesized views. Active mapping is further performed within a maximum information gain framework building on this formulation. Extensive experiments across diverse environments demonstrate that GAVIS consistently and significantly outperforms prior approaches in both accuracy and efficiency. Moreover, beyond standalone use, our method can be applied post-hoc to improve the performance of existing approaches.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The manuscript presents GAVIS, a framework for uncertainty quantification and active mapping in 3D Gaussian Splatting. It introduces an anisotropic visibility field for each Gaussian particle computed from training views and represented via spherical harmonics. This field is integrated into a Bayesian network-based uncertainty-aware rasterizer to enable real-time (200 FPS) uncertainty quantification for novel views. The formulation is then used within a maximum information gain framework for active mapping, with claims of consistent outperformance over prior methods across environments and post-hoc applicability to existing approaches.

Significance. If validated, the approach could provide an efficient mechanism for uncertainty-aware active mapping in 3DGS, addressing a practical need in real-time reconstruction and exploration tasks. The reported 200 FPS performance and post-hoc usability are notable if they hold. However, the core assumption that anisotropic visibility from training views dominates other sources of 3DGS prediction error requires stronger empirical grounding to support the downstream uncertainty and information-gain claims.

major comments (3)
  1. [Visibility Field and Bayesian Network Integration] The central assumption that the anisotropic visibility field (computed solely from training views via spherical harmonics) accurately captures unreliability for uncertainty quantification is load-bearing but insufficiently justified. Other error sources in 3DGS, such as optimization artifacts, under-constrained covariances, or color fitting failures in seen regions, are not encoded in the visibility representation and could lead to miscalibrated uncertainty maps.
  2. [Uncertainty-aware 3DGS Rasterizer] The integration of the visibility field into the Bayesian network rasterizer for producing calibrated uncertainty at 200 FPS lacks supporting evidence such as calibration plots, error-vs-uncertainty correlations, or comparisons to ground-truth reconstruction errors. Without these, the claim that the visibility field supplies reliable uncertainty for information gain calculations remains unverified.
  3. [Active Mapping Experiments] In the maximum information gain active mapping experiments, ablations are needed to isolate the contribution of the visibility-based uncertainty versus other components of the framework. The outperformance claims would be strengthened by showing that alternative uncertainty measures (capturing additional error sources) do not yield comparable gains.
minor comments (2)
  1. [Abstract] The abstract states that the method 'consistently and significantly outperforms prior approaches' but does not list the specific baselines, metrics, or quantitative improvements; adding these would improve clarity.
  2. [Method] Notation for the spherical harmonics representation of the visibility field should be defined more explicitly (e.g., degree/order, normalization) to facilitate reproducibility.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for their insightful comments on our manuscript. We address each of the major comments below, providing clarifications and indicating revisions to strengthen the paper.

read point-by-point responses

  1. Referee: [Visibility Field and Bayesian Network Integration] The central assumption that the anisotropic visibility field (computed solely from training views via spherical harmonics) accurately captures unreliability for uncertainty quantification is load-bearing but insufficiently justified. Other error sources in 3DGS, such as optimization artifacts, under-constrained covariances, or color fitting failures in seen regions, are not encoded in the visibility representation and could lead to miscalibrated uncertainty maps.

    Authors: We acknowledge that 3DGS can have multiple sources of error. Our approach specifically targets visibility from training views as the dominant factor for uncertainty in novel views during active mapping, as the Gaussians are fitted to observed data. The anisotropic representation via spherical harmonics allows efficient computation of this. To address the concern, we will revise the manuscript to include a more detailed discussion of potential other error sources and their relation to our visibility field. revision: yes

  2. Referee: [Uncertainty-aware 3DGS Rasterizer] The integration of the visibility field into the Bayesian network rasterizer for producing calibrated uncertainty at 200 FPS lacks supporting evidence such as calibration plots, error-vs-uncertainty correlations, or comparisons to ground-truth reconstruction errors. Without these, the claim that the visibility field supplies reliable uncertainty for information gain calculations remains unverified.

    Authors: The real-time performance at 200 FPS is achieved through the efficient integration into the rasterizer. While the current manuscript emphasizes the active mapping results, we agree that additional validation is beneficial. We will add calibration plots and error-uncertainty correlation analyses in the revised version to provide stronger empirical support for the uncertainty estimates. revision: yes

  3. Referee: [Active Mapping Experiments] In the maximum information gain active mapping experiments, ablations are needed to isolate the contribution of the visibility-based uncertainty versus other components of the framework. The outperformance claims would be strengthened by showing that alternative uncertainty measures (capturing additional error sources) do not yield comparable gains.

    Authors: We will incorporate ablations in the experiments section that compare our visibility-based uncertainty against alternative uncertainty measures, such as those derived from Gaussian covariances or reconstruction residuals. This will help isolate the contribution of the anisotropic visibility field to the information gain framework. revision: yes

Circularity Check

0 steps flagged

No circularity; derivation self-contained with no reductions to inputs

full rationale

The abstract and context present GAVIS as introducing a visibility field quantified from training views and integrated into a Bayesian Network rasterizer for uncertainty and active mapping. No equations are shown, and no derivation steps reduce by construction to fitted parameters, self-citations, or renamed inputs. The central insight (unseen regions yield unreliable predictions) is stated as motivation rather than derived from prior results within the paper. Claims of 200 FPS performance and experimental outperformance are presented as empirical outcomes, not forced by internal definitions. This matches the reader's abstract-only assessment that no circularity can be diagnosed. The method is self-contained against external benchmarks such as real-time rasterization and information-gain frameworks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Only the abstract is available; the ledger is therefore populated with the minimal assumptions extractable from the text. The central claim rests on the unverified premise that unseen regions produce unreliable 3DGS predictions and that spherical-harmonics visibility quantifies this reliably.

axioms (1)
  • domain assumption Regions unseen from the training views yield unreliable predictions from the 3DGS.
    Stated as the key insight in the abstract; no supporting derivation or measurement is provided.

pith-pipeline@v0.9.1-grok · 5709 in / 1317 out tokens · 18212 ms · 2026-06-29T07:30:38.184855+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

122 extracted references · 5 canonical work pages · 1 internal anchor

  1. [1]

    US Government printing office, 1948

    Milton Abramowitz and Irene A Stegun.Handbook of mathematical functions with formulas, graphs, and math- ematical tables. US Government printing office, 1948. 14

    1948

  2. [2]

    3dgs-to-pc: 3d gaussian splatting to dense point clouds

    Lewis AG Stuart, Andrew Morton, Ian Stavness, and Michael P Pound. 3dgs-to-pc: 3d gaussian splatting to dense point clouds. InProceedings of the IEEE/CVF In- ternational Conference on Computer Vision, pages 3730– 3739, 2025. 22

    2025

  3. [3]

    Hubble space telescope 3d model.https://science.nasa.gov/resource/ hubble-space-telescope-3d-model-2/, 2019

    NASA Visualization Technology Applications and De- velopment (VTAD). Hubble space telescope 3d model.https://science.nasa.gov/resource/ hubble-space-telescope-3d-model-2/, 2019. 21

    2019

  4. [4]

    International space station 3d model.https://science.nasa.gov/resource/ international - space - station - 3d - model/,

    NASA Visualization Technology Applications and De- velopment (VTAD). International space station 3d model.https://science.nasa.gov/resource/ international - space - station - 3d - model/,

  5. [5]

    Decentralized active information acquisi- tion: Theory and application to multi-robot slam

    Nikolay Atanasov, Jerome Le Ny, Kostas Daniilidis, and George J Pappas. Decentralized active information acquisi- tion: Theory and application to multi-robot slam. In2015 IEEE International Conference on Robotics and Automa- tion (ICRA), pages 4775–4782. IEEE, 2015. 2

    2015

  6. [6]

    A shadowcasting-based next-best-view planner for autonomous 3d exploration.IEEE Robotics and Automation Letters, 7(2):2969–2976, 2022

    Ana Batinovic, Antun Ivanovic, Tamara Petrovic, and Stjepan Bogdan. A shadowcasting-based next-best-view planner for autonomous 3d exploration.IEEE Robotics and Automation Letters, 7(2):2969–2976, 2022. 2

    2022

  7. [7]

    Receding horizon” next- best-view” planner for 3d exploration

    Andreas Bircher, Mina Kamel, Kostas Alexis, Helen Oleynikova, and Roland Siegwart. Receding horizon” next- best-view” planner for 3d exploration. In2016 IEEE in- ternational conference on robotics and automation (ICRA), pages 1462–1468. IEEE, 2016. 1, 2

    2016

  8. [8]

    Receding horizon path planning for 3d exploration and surface inspection.Au- tonomous Robots, 42(2):291–306, 2018

    Andreas Bircher, Mina Kamel, Kostas Alexis, Helen Oleynikova, and Roland Siegwart. Receding horizon path planning for 3d exploration and surface inspection.Au- tonomous Robots, 42(2):291–306, 2018. 1, 2

    2018

  9. [9]

    Active slam and exploration with particle filters using kullback-leibler divergence.Journal of Intelligent & Robotic Systems, 75(2):291–311, 2014

    Luca Carlone, Jingjing Du, Miguel Kaouk Ng, Basilio Bona, and Marina Indri. Active slam and exploration with particle filters using kullback-leibler divergence.Journal of Intelligent & Robotic Systems, 75(2):291–311, 2014. 2

    2014

  10. [10]

    On the comparison of uncertainty criteria for active slam

    Henry Carrillo, Ian Reid, and Jos ´e A Castellanos. On the comparison of uncertainty criteria for active slam. In2012 IEEE International Conference on Robotics and Automa- tion, pages 2080–2087. IEEE, 2012. 2

    2080

  11. [11]

    Autonomous exploration under uncer- tainty via deep reinforcement learning on graphs

    Fanfei Chen, John D Martin, Yewei Huang, Jinkun Wang, and Brendan Englot. Autonomous exploration under uncer- tainty via deep reinforcement learning on graphs. In2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 6140–6147. IEEE, 2020. 2

    2020

  12. [12]

    Activegamer: Active gaussian mapping through efficient rendering.arXiv preprint arXiv:2501.06897, 2025

    Liyan Chen, Huangying Zhan, Kevin Chen, Xiangyu Xu, Qingan Yan, Changjiang Cai, and Yi Xu. Activegamer: Active gaussian mapping through efficient rendering.arXiv preprint arXiv:2501.06897, 2025. 2

    work page arXiv 2025

  13. [13]

    Hallucinated neural radiance fields in the wild

    Xingyu Chen, Qi Zhang, Xiaoyu Li, Yue Chen, Ying Feng, Xuan Wang, and Jue Wang. Hallucinated neural radiance fields in the wild. InProceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition, pages 12943–12952, 2022. 2

    2022

  14. [14]

    The determination of next best views

    Cl Connolly. The determination of next best views. InPro- ceedings. 1985 IEEE international conference on robotics and automation, pages 432–435. IEEE, 1985. 2

    1985

  15. [15]

    Fast frontier-based information-driven autonomous exploration with an mav

    Anna Dai, Sotiris Papatheodorou, Nils Funk, Dimos Tzoumanikas, and Stefan Leutenegger. Fast frontier-based information-driven autonomous exploration with an mav. In2020 IEEE international conference on robotics and au- tomation (ICRA), pages 9570–9576. IEEE, 2020. 2

    2020

  16. [16]

    Graph-based path plan- ning for autonomous robotic exploration in subterranean environments

    Tung Dang, Frank Mascarich, Shehryar Khattak, Christos Papachristos, and Kostas Alexis. Graph-based path plan- ning for autonomous robotic exploration in subterranean environments. In2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 3105–

  17. [17]

    Depth-supervised nerf: Fewer views and faster training for free

    Kangle Deng, Andrew Liu, Jun-Yan Zhu, and Deva Ra- manan. Depth-supervised nerf: Fewer views and faster training for free. InProceedings of the IEEE/CVF con- ference on computer vision and pattern recognition, pages 12882–12891, 2022. 21

    2022

  18. [18]

    A frontier- void-based approach for autonomous exploration in 3d.Ad- vanced Robotics, 27(6):459–468, 2013

    Christian Dornhege and Alexander Kleiner. A frontier- void-based approach for autonomous exploration in 3d.Ad- vanced Robotics, 27(6):459–468, 2013. 2

    2013

  19. [19]

    Naruto: Neural active reconstruction from uncertain target observations

    Ziyue Feng, Huangying Zhan, Zheng Chen, Qingan Yan, Xiangyu Xu, Changjiang Cai, Bing Li, Qilun Zhu, and Yi Xu. Naruto: Neural active reconstruction from uncertain target observations. InProceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, pages 21572–21583, 2024. 2

    2024

  20. [20]

    Plenoxels: Radiance fields without neural networks

    Sara Fridovich-Keil, Alex Yu, Matthew Tancik, Qinhong Chen, Benjamin Recht, and Angjoo Kanazawa. Plenoxels: Radiance fields without neural networks. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 5501–5510, 2022. 2, 5

    2022

  21. [21]

    Dropout as a bayesian approximation: Representing model uncertainty in deep learning

    Yarin Gal and Zoubin Ghahramani. Dropout as a bayesian approximation: Representing model uncertainty in deep learning. Ininternational conference on machine learning, pages 1050–1059. PMLR, 2016. 2

    2016

  22. [22]

    A survey of uncertainty in deep neural net- works.Artificial Intelligence Review, 56(Suppl 1):1513– 1589, 2023

    Jakob Gawlikowski, Cedrique Rovile Njieutcheu Tassi, Mohsin Ali, Jongseok Lee, Matthias Humt, Jianxiang Feng, Anna Kruspe, Rudolph Triebel, Peter Jung, Ribana Roscher, et al. A survey of uncertainty in deep neural net- works.Artificial Intelligence Review, 56(Suppl 1):1513– 1589, 2023. 1

    2023

  23. [23]

    Bayes’ rays: Uncertainty quantifi- cation for neural radiance fields

    Lily Goli, Cody Reading, Silvia Sell ´an, Alec Jacobson, and Andrea Tagliasacchi. Bayes’ rays: Uncertainty quantifi- cation for neural radiance fields. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 20061–20070, 2024. 2, 7, 20, 22

    2024

  24. [24]

    Topological frontier-based exploration and map-building using semantic information.Sensors, 19(20):4595, 2019

    Clara Gomez, Alejandra C Hernandez, and Ramon Barber. Topological frontier-based exploration and map-building using semantic information.Sensors, 19(20):4595, 2019. 2

    2019

  25. [25]

    On calibration of modern neural networks

    Chuan Guo, Geoff Pleiss, Yu Sun, and Kilian Q Weinberger. On calibration of modern neural networks. InInternational conference on machine learning, pages 1321–1330. PMLR,

  26. [26]

    Evaluating the efficiency of frontier-based explo- ration strategies

    Dirk Holz, Nicola Basilico, Francesco Amigoni, and Sven Behnke. Evaluating the efficiency of frontier-based explo- ration strategies. InISR 2010 (41st International Sympo- sium on Robotics) and ROBOTIK 2010 (6th German Con- ference on Robotics), pages 1–8. VDE, 2010. 2

    2010

  27. [27]

    3d visibility-aware generalizable neural ra- diance fields for interacting hands

    Xuan Huang, Hanhui Li, Zejun Yang, Zhisheng Wang, and Xiaodan Liang. 3d visibility-aware generalizable neural ra- diance fields for interacting hands. InProceedings of the AAAI Conference on Artificial Intelligence, pages 2400– 2408, 2024. 2

    2024

  28. [28]

    On entropy approximation for gaussian mixture random vectors

    Marco F Huber, Tim Bailey, Hugh Durrant-Whyte, and Uwe D Hanebeck. On entropy approximation for gaussian mixture random vectors. In2008 IEEE International Con- ference on Multisensor Fusion and Integration for Intelli- gent Systems, pages 181–188. IEEE, 2008. 18

    2008

  29. [29]

    Uncer- tainty estimates and multi-hypotheses networks for optical flow

    Eddy Ilg, Ozgun Cicek, Silvio Galesso, Aaron Klein, Osama Makansi, Frank Hutter, and Thomas Brox. Uncer- tainty estimates and multi-hypotheses networks for optical flow. InProceedings of the European Conference on Com- puter Vision (ECCV), pages 652–667, 2018. 7, 22

    2018

  30. [30]

    John Wi- ley & Sons, Inc., Hoboken, NJ, 3rd edition, 1999

    John David Jackson.Classical Electrodynamics. John Wi- ley & Sons, Inc., Hoboken, NJ, 3rd edition, 1999. 14

    1999

  31. [31]

    Ag-slam: Active gaussian splatting slam

    Wen Jiang, Boshu Lei, Katrina Ashton, and Kostas Dani- ilidis. Ag-slam: Active gaussian splatting slam. 2024. 19

    2024

  32. [32]

    Fisherrf: Ac- tive view selection and uncertainty quantification for radi- ance fields using fisher information.ECCV, 2024

    Wen Jiang, Boshu Lei, and Kostas Daniilidis. Fisherrf: Ac- tive view selection and uncertainty quantification for radi- ance fields using fisher information.ECCV, 2024. 1, 2, 6, 7, 16, 18, 20, 22

    2024

  33. [33]

    Neu-nbv: Next best view planning using uncer- tainty estimation in image-based neural rendering

    Liren Jin, Xieyuanli Chen, Julius R ¨uckin, and Marija Popovi´c. Neu-nbv: Next best view planning using uncer- tainty estimation in image-based neural rendering. In2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 11305–11312. IEEE, 2023. 2, 22

    2023

  34. [34]

    Activegs: Active scene re- construction using gaussian splatting.IEEE Robotics and Automation Letters, 2025

    Liren Jin, Xingguang Zhong, Yue Pan, Jens Behley, Cyrill Stachniss, and Marija Popovi ´c. Activegs: Active scene re- construction using gaussian splatting.IEEE Robotics and Automation Letters, 2025. 2

    2025

  35. [35]

    Gs-planner: A gaussian-splatting- based planning framework for active high-fidelity recon- struction

    Rui Jin, Yuman Gao, Yingjian Wang, Yuze Wu, Haojian Lu, Chao Xu, and Fei Gao. Gs-planner: A gaussian-splatting- based planning framework for active high-fidelity recon- struction. In2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 11202– 11209. IEEE, 2024. 2

    2024

  36. [36]

    Splatam: Splat track & map 3d gaus- sians for dense rgb-d slam

    Nikhil Keetha, Jay Karhade, Krishna Murthy Jatavallab- hula, Gengshan Yang, Sebastian Scherer, Deva Ramanan, and Jonathon Luiten. Splatam: Splat track & map 3d gaus- sians for dense rgb-d slam. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 21357–21366, 2024. 2

    2024

  37. [37]

    Robot exploration with fast frontier detection: Theory and experiments

    Matan Keidar and Gal A Kaminka. Robot exploration with fast frontier detection: Theory and experiments. In Proceedings of the 11th International Conference on Au- tonomous Agents and Multiagent Systems-Volume 1, pages 113–120, 2012. 2

    2012

  38. [38]

    Efficient frontier de- tection for robot exploration.The International Journal of Robotics Research, 33(2):215–236, 2014

    Matan Keidar and Gal A Kaminka. Efficient frontier de- tection for robot exploration.The International Journal of Robotics Research, 33(2):215–236, 2014. 2

    2014

  39. [39]

    3d gaussian splatting for real-time radiance field rendering.ACM Trans

    Bernhard Kerbl, Georgios Kopanas, Thomas Leimk ¨uhler, and George Drettakis. 3d gaussian splatting for real-time radiance field rendering.ACM Trans. Graph., 42(4):139–1,

  40. [40]

    1, 2, 3, 5, 15, 17, 21

  41. [41]

    arXiv preprint arXiv:2208.00549 (2022) Title Suppressed Due to Excessive Length 17

    Andreas Kirsch and Yarin Gal. Unifying approaches in active learning and active sampling via fisher informa- tion and information-theoretic quantities.arXiv preprint arXiv:2208.00549, 2022. 19, 20

    work page arXiv 2022

  42. [42]

    Sources of uncertainty in 3d scene reconstruction

    Marcus Klasson, Riccardo Mereu, Juho Kannala, and Arno Solin. Sources of uncertainty in 3d scene reconstruction. In European Conference on Computer Vision, pages 271–289. Springer, 2025. 2

    2025

  43. [43]

    Trajectory optimization using reinforcement learning for map exploration.The In- ternational Journal of Robotics Research, 27(2):175–196,

    Thomas Kollar and Nicholas Roy. Trajectory optimization using reinforcement learning for map exploration.The In- ternational Journal of Robotics Research, 27(2):175–196,

  44. [44]

    Informed sampling explo- ration path planner for 3d reconstruction of large scenes

    Yves Kompis, Luca Bartolomei, Ruben Mascaro, Lucas Teixeira, and Margarita Chli. Informed sampling explo- ration path planner for 3d reconstruction of large scenes. IEEE Robotics and Automation Letters, 6(4):7893–7900,

  45. [45]

    Improving nerf quality by progressive camera placement for free-viewpoint navigation

    Georgios Kopanas and George Drettakis. Improving nerf quality by progressive camera placement for free-viewpoint navigation. 2023. 2

    2023

  46. [46]

    Active neural mapping at scale

    Zijia Kuang, Zike Yan, Hao Zhao, Guyue Zhou, and Hong- bin Zha. Active neural mapping at scale. In2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 7152–7159. IEEE, 2024. 2

    2024

  47. [47]

    Uncertainty guided pol- icy for active robotic 3d reconstruction using neural radi- ance fields.IEEE Robotics and Automation Letters, 7(4): 12070–12077, 2022

    Soomin Lee, Le Chen, Jiahao Wang, Alexander Liniger, Suryansh Kumar, and Fisher Yu. Uncertainty guided pol- icy for active robotic 3d reconstruction using neural radi- ance fields.IEEE Robotics and Automation Letters, 7(4): 12070–12077, 2022. 2, 3, 19

    2022

  48. [48]

    Bayesian nerf: Quantifying uncertainty with volume density for neural implicit fields.arXiv preprint arXiv:2404.06727, 2024

    Sibeak Lee, Kyeongsu Kang, Seongbo Ha, and Hyeon- woo Yu. Bayesian nerf: Quantifying uncertainty with volume density for neural implicit fields.arXiv preprint arXiv:2404.06727, 2024. 2

    work page arXiv 2024

  49. [49]

    Variational multi-scale rep- resentation for estimating uncertainty in 3d gaussian splat- ting.Advances in Neural Information Processing Systems, 37:87934–87958, 2024

    Ruiqi Li and Yiu-ming Cheung. Variational multi-scale rep- resentation for estimating uncertainty in 3d gaussian splat- ting.Advances in Neural Information Processing Systems, 37:87934–87958, 2024. 2

    2024

  50. [50]

    Bavs- nerf: Batch active view selection for neural radiance field using scene uncertainty

    You Li, Rui Li, Shengjun Tang, and Yanjie Gu. Bavs- nerf: Batch active view selection for neural radiance field using scene uncertainty. InThe First Workshop on Popu- lating Empty Cities–Virtual Humans for Robotics and Au- tonomous Driving at CVPR 2024. 2

    2024

  51. [51]

    Activesplat: High-fidelity scene reconstruction through active gaussian splatting.IEEE Robotics and Automation Letters, 2025

    Yuetao Li, Zijia Kuang, Ting Li, Qun Hao, Zike Yan, Guyue Zhou, and Shaohui Zhang. Activesplat: High-fidelity scene reconstruction through active gaussian splatting.IEEE Robotics and Automation Letters, 2025. 2, 22

    2025

  52. [52]

    Active mapping and robot exploration: A survey.Sensors, 21(7): 2445, 2021

    Iker Lluvia, Elena Lazkano, and Ander Ansuategi. Active mapping and robot exploration: A survey.Sensors, 21(7): 2445, 2021. 1

    2021

  53. [53]

    Optimal frontier-based autonomous exploration in unconstructed environment using rgb-d sensor.Sensors, 20(22):6507,

    Liang Lu, Carlos Redondo, and Pascual Campoy. Optimal frontier-based autonomous exploration in unconstructed environment using rgb-d sensor.Sensors, 20(22):6507,

  54. [54]

    Enabling efficient equivariant operations in the fourier basis via gaunt tensor products.arXiv preprint arXiv:2401.10216,

    Shengjie Luo, Tianlang Chen, and Aditi S Krish- napriyan. Enabling efficient equivariant operations in the fourier basis via gaunt tensor products.arXiv preprint arXiv:2401.10216, 2024. 5, 15

    work page arXiv 2024

  55. [55]

    Manifold sampling for differentiable uncer- tainty in radiance fields

    Linjie Lyu, Ayush Tewari, Marc Habermann, Shunsuke Saito, Michael Zollh ¨ofer, Thomas Leimk ¨uhler, and Chris- tian Theobalt. Manifold sampling for differentiable uncer- tainty in radiance fields. InSIGGRAPH Asia 2024 Confer- ence Papers, pages 1–11, 2024. 1, 2, 3, 6, 7, 16, 18, 19, 20, 22, 23

    2024

  56. [56]

    Gaussian splatting slam

    Hidenobu Matsuki, Riku Murai, Paul HJ Kelly, and An- drew J Davison. Gaussian splatting slam. InProceedings of the IEEE/CVF Conference on Computer Vision and Pat- tern Recognition, pages 18039–18048, 2024. 2, 21

    2024

  57. [57]

    Zehui Meng, Hailong Qin, Ziyue Chen, Xudong Chen, Hao Sun, Feng Lin, and Marcelo H Ang. A two-stage opti- mized next-view planning framework for 3-d unknown en- vironment exploration, and structural reconstruction.IEEE Robotics and Automation Letters, 2(3):1680–1687, 2017. 2

    2017

  58. [58]

    Nerf: Representing scenes as neural radiance fields for view syn- thesis.Communications of the ACM, 65(1):99–106, 2021

    Ben Mildenhall, Pratul P Srinivasan, Matthew Tancik, Jonathan T Barron, Ravi Ramamoorthi, and Ren Ng. Nerf: Representing scenes as neural radiance fields for view syn- thesis.Communications of the ACM, 65(1):99–106, 2021. 1, 2, 3, 6, 17, 21

    2021

  59. [59]

    Information-based active slam via topological feature graphs

    Beipeng Mu, Matthew Giamou, Liam Paull, Ali-akbar Agha-mohammadi, John Leonard, and Jonathan How. Information-based active slam via topological feature graphs. In2016 IEEE 55th Conference on decision and control (Cdc), pages 5583–5590. IEEE, 2016. 2

    2016

  60. [60]

    Instant neural graphics primitives with a multiresolution hash encoding.ACM transactions on graphics (TOG), 41(4):1–15, 2022

    Thomas M ¨uller, Alex Evans, Christoph Schied, and Alexander Keller. Instant neural graphics primitives with a multiresolution hash encoding.ACM transactions on graphics (TOG), 41(4):1–15, 2022. 2, 5

    2022

  61. [61]

    Provnerf: Modeling per point provenance in nerfs as a stochastic field.Advances in Neural Information Processing Systems, 37:99145–99160,

    Kiyohiro Nakayama, Mikaela Angelina Uy, Yang You, Ke Li, and Leonidas J Guibas. Provnerf: Modeling per point provenance in nerfs as a stochastic field.Advances in Neural Information Processing Systems, 37:99145–99160,

  62. [62]

    Deep reinforcement learning robot for search and rescue applications: Exploration in unknown cluttered environments.IEEE Robotics and Automation Letters, 4 (2):610–617, 2019

    Farzad Niroui, Kaicheng Zhang, Zendai Kashino, and Goldie Nejat. Deep reinforcement learning robot for search and rescue applications: Exploration in unknown cluttered environments.IEEE Robotics and Automation Letters, 4 (2):610–617, 2019. 2

    2019

  63. [63]

    Cambridge university press, 2010

    Frank WJ Olver.NIST handbook of mathematical functions hardback and CD-ROM. Cambridge university press, 2010. 5, 14

    2010

  64. [64]

    Can you trust your model’s uncertainty? evaluating predictive uncertainty under dataset shift.Advances in neural information processing systems, 32, 2019

    Yaniv Ovadia, Emily Fertig, Jie Ren, Zachary Nado, David Sculley, Sebastian Nowozin, Joshua Dillon, Balaji Laksh- minarayanan, and Jasper Snoek. Can you trust your model’s uncertainty? evaluating predictive uncertainty under dataset shift.Advances in neural information processing systems, 32, 2019. 1

    2019

  65. [65]

    Ac- tivenerf: Learning where to see with uncertainty estima- tion

    Xuran Pan, Zihang Lai, Shiji Song, and Gao Huang. Ac- tivenerf: Learning where to see with uncertainty estima- tion. InEuropean Conference on Computer Vision, pages 230–246. Springer, 2022. 2

    2022

  66. [66]

    Uncertainty-aware receding horizon exploration and mapping using aerial robots

    Christos Papachristos, Shehryar Khattak, and Kostas Alexis. Uncertainty-aware receding horizon exploration and mapping using aerial robots. In2017 IEEE inter- national conference on robotics and automation (ICRA), pages 4568–4575. IEEE, 2017. 2

    2017

  67. [67]

    A survey on active simultaneous localization and mapping: State of the art and new frontiers.IEEE Transactions on Robotics, 2023

    Julio A Placed, Jared Strader, Henry Carrillo, Nikolay Atanasov, Vadim Indelman, Luca Carlone, and Jos ´e A Castellanos. A survey on active simultaneous localization and mapping: State of the art and new frontiers.IEEE Transactions on Robotics, 2023. 1, 2, 19

    2023

  68. [68]

    On the uncertainty of self-supervised monocular depth estimation

    Matteo Poggi, Filippo Aleotti, Fabio Tosi, and Stefano Mat- toccia. On the uncertainty of self-supervised monocular depth estimation. InProceedings of the IEEE/CVF con- ference on computer vision and pattern recognition, pages 3227–3237, 2020. 7, 22

    2020

  69. [69]

    Sample- based frontier detection for autonomous robot exploration

    Wenchuan Qiao, Zheng Fang, and Bailu Si. Sample- based frontier detection for autonomous robot exploration. In2018 IEEE International Conference on Robotics and Biomimetics (ROBIO), pages 1165–1170. IEEE, 2018. 2

    2018

  70. [70]

    Habitat-matterport 3d dataset (HM3d): 1000 large-scale 3d environments for embodied AI

    Santhosh Kumar Ramakrishnan, Aaron Gokaslan, Erik Wi- jmans, Oleksandr Maksymets, Alexander Clegg, John M Turner, Eric Undersander, Wojciech Galuba, Andrew West- bury, Angel X Chang, Manolis Savva, Yili Zhao, and Dhruv Batra. Habitat-matterport 3d dataset (HM3d): 1000 large-scale 3d environments for embodied AI. InThirty- fifth Conference on Neural Inform...

    2021

  71. [71]

    Neurar: Neural uncertainty for autonomous 3d reconstruction with implicit neural representations.IEEE Robotics and Automation Let- ters, 8(2):1125–1132, 2023

    Yunlong Ran, Jing Zeng, Shibo He, Jiming Chen, Lincheng Li, Yingfeng Chen, Gimhee Lee, and Qi Ye. Neurar: Neural uncertainty for autonomous 3d reconstruction with implicit neural representations.IEEE Robotics and Automation Let- ters, 8(2):1125–1132, 2023. 2

    2023

  72. [72]

    Fast sampling-based next-best-view exploration algorithm for a mav

    Victor Massagu ´e Respall, Dmitry Devitt, Roman Fe- dorenko, and Alexandr Klimchik. Fast sampling-based next-best-view exploration algorithm for a mav. In2021 IEEE International Conference on Robotics and Automa- tion (ICRA), pages 89–95. IEEE, 2021. 2

    2021

  73. [73]

    Cambridge University Press, 2020

    Jun John Sakurai and Jim Napolitano.Modern quantum mechanics. Cambridge University Press, 2020. 15

    2020

  74. [74]

    Habitat: A platform for embodied ai research

    Manolis Savva, Abhishek Kadian, Oleksandr Maksymets, Yili Zhao, Erik Wijmans, Bhavana Jain, Julian Straub, Jia Liu, Vladlen Koltun, Jitendra Malik, et al. Habitat: A platform for embodied ai research. InProceedings of the IEEE/CVF international conference on computer vision, pages 9339–9347, 2019. 21

    2019

  75. [75]

    Efficient autonomous explo- ration planning of large-scale 3-d environments.IEEE Robotics and Automation Letters, 4(2):1699–1706, 2019

    Magnus Selin, Mattias Tiger, Daniel Duberg, Fredrik Heintz, and Patric Jensfelt. Efficient autonomous explo- ration planning of large-scale 3-d environments.IEEE Robotics and Automation Letters, 4(2):1699–1706, 2019. 2

    2019

  76. [76]

    Towards au- tonomous 3d exploration using surface frontiers

    PGCN Senarathne and Danwei Wang. Towards au- tonomous 3d exploration using surface frontiers. In2016 IEEE International Symposium on Safety, Security, and Rescue Robotics (SSRR), pages 34–41. IEEE, 2016. 2

    2016

  77. [77]

    Stochastic neural radiance fields: Quan- tifying uncertainty in implicit 3d representations

    Jianxiong Shen, Adria Ruiz, Antonio Agudo, and Francesc Moreno-Noguer. Stochastic neural radiance fields: Quan- tifying uncertainty in implicit 3d representations. In2021 International Conference on 3D Vision (3DV), pages 972–

  78. [78]

    Conditional-flow nerf: Accurate 3d mod- elling with reliable uncertainty quantification

    Jianxiong Shen, Antonio Agudo, Francesc Moreno-Noguer, and Adria Ruiz. Conditional-flow nerf: Accurate 3d mod- elling with reliable uncertainty quantification. InEuropean Conference on Computer Vision, pages 540–557. Springer,

  79. [79]

    Estimating 3d uncertainty field: Quan- tifying uncertainty for neural radiance fields

    Jianxiong Shen, Ruijie Ren, Adria Ruiz, and Francesc Moreno-Noguer. Estimating 3d uncertainty field: Quan- tifying uncertainty for neural radiance fields. In2024 IEEE International Conference on Robotics and Automa- tion (ICRA), pages 2375–2381. IEEE, 2024. 2

    2024

  80. [80]

    Shaojie Shen, Nathan Michael, and Vijay Kumar. Stochas- tic differential equation-based exploration algorithm for au- tonomous indoor 3d exploration with a micro-aerial vehi- cle.The International Journal of Robotics Research, 31 (12):1431–1444, 2012. 2

    2012

Showing first 80 references.