Pith. sign in

REVIEW 3 major objections 5 minor 28 references

DOA: A Degeneracy Optimization Agent with Adaptive Pose Compensation Capability based on Deep Reinforcement Learning

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read A PPO-trained agent detects SLAM degeneracy online and shifts the particle distribution toward the motion model, cutting GMapping's ATE by up to 95.2% in simulations.

desk verdict A useful, well-engineered RL-based degeneracy compensator for particle-filter SLAM, with a post hoc threshold and unvalidated reward proxies that keep it from being a clean accept. read the letter →

arxiv 2507.19742 v1 pith:5BCXQB7D submitted 2025-07-26 cs.RO cs.LGcs.SYeess.SY

Yanbin Li , Canran Xiao , Hongyang He , Shenghai Yuan , Zong Ke , Jiajie Yu , Zixiong Qin , Zhiguo Zhang
show 2 more authors
Wenzheng Chi Wei Zhang
This is my paper
classification cs.ROcs.LGcs.SYeess.SY
keywords ReinforcementLearningDegeneracyOptimizationDetectionTransferParticleFilterSLAMLidarLocalizationSensorFusionPPO
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper aims to solve the degeneracy problem in particle-filter 2D SLAM, where long featureless corridors make lidar constraints rank-deficient and localization drifts. Instead of fixed sensor-fusion rules or supervised learning with ambiguous labels, the authors train a Proximal Policy Optimization agent to output a degeneracy factor $a \in (0,1)$ that controls how much the observation particle distribution is pulled toward the motion-model distribution. The training reward encourages smaller particle covariance, higher scan-match likelihood, and higher effective particle count, with a penalty on abrupt factor changes. In simulated corridor scenes, the resulting DOA system reduces GMapping's absolute trajectory error by up to 95%, and it detects degeneracy with roughly 90% success across simulation and real-world scenes. If these results hold, the approach offers a parameter-free, label-free way to make particle-filter SLAM robust in geometry-poor indoor environments.

What carries the argument

The load-bearing component is the fusion model of Eqs. (12)–(14): a linear interpolation between the centroids and inverse covariances of the observation distribution $p_z$ and the motion-model distribution $p_u$, weighted by the degeneracy factor $a \in (0,1)$ learned by the PPO policy. In a degenerate direction the observation inverse covariance $\Sigma_z^{-1}$ vanishes, so the fused distribution collapses onto $p_u$, and the optimal fused mean becomes $g_c = (1-a)g_z + a g_u$. The agent's policy network is a shared backbone with multi-head attention over a 30-particle state vector, trained with a reward that encourages covariance contraction, higher scan-match likelihood, higher effective particle count $N_{\text{eff}}$, and smooth factor changes; transfer learning pre-trains in a feature-rich scene and fine-tunes in a corridor with the backbone frozen.

What would settle it

Measure, on a held-out simulated corridor with Gazebo ground truth, the per-step values of the three reward terms alongside the ATE of the compensated trajectory; if episodes where the reward increases show higher ATE than episodes where it decreases, the reward-proxy assumption is false. A second check: run the agent in a corridor where wheel odometry drifts (for example, tire slip) and see whether the compensation factor keeps shifting trust to the motion model while ATE grows.

Watch

Extended reading notes

Core claim

The central claim is that a single learned scalar — the degeneracy factor $a$ output by a PPO policy — is enough to characterise and compensate for SLAM degeneracy in real time. When the observation distribution $p_z$ loses constraints along a degenerate direction, the agent shifts the particle swarm toward the motion distribution $p_u$ using the fusion formula $\Sigma_c^{-1} = (1-a)\Sigma_z^{-1} + a\Sigma_u^{-1}$ and $g_c = (1-a)g_z + a g_u$, then selects the better of the original and fused distributions by likelihood score. The paper argues this is the first PPO-based dynamic sensor fusion for particle-filter SLAM in degenerate environments, and supports it with ablation studies, comparison against Switch-SLAM and a supervised learning baseline, and qualitative real-world maps.

Load-bearing premise

The agent is trained on a reward that rewards smaller particle covariance, higher scan-match likelihood, and higher effective particle count, but the paper never checks against ground truth that these proxy improvements actually mean better localization; if the proxies misalign (for example, if shrinking covariance mainly means trusting odometry too much), the policy could score well while the true trajectory gets worse.

Editorial extensions

If this is right

  • GMapping plus DOA needs no changes to the SLAM core: it only consumes a scalar per step, so it can be dropped onto existing particle-filter pipelines.
  • Detection and compensation run at about 15 ms forward-pass time, well under the 170 ms SLAM update period, supporting real-time use.
  • Training requires no manual degeneracy annotations, replacing supervised labeling with a reward signal, which is what makes the method applicable to new environments via transfer learning.
  • In the four simulation scenes, ATE falls from 0.692 m to 0.126 m in S2, 4.32 m to 0.266 m in S3, and 2.561 m to 0.122 m in S4, with the maximum x/y error also shrinking.
  • The same factor can be used for degeneracy detection alone, reaching about 90% success with the optimization module removed.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A natural extension would be to output a per-axis or per-sensor compensation vector instead of a single scalar, since degenerate directions vary.
  • The fixed detection threshold of 0.75 is a manual choice; a learned confidence could replace it and make detection comparisons fairer.
  • Because the reward never uses ground-truth pose, a validation study comparing each reward term against ATE on held-out corridors would show whether the proxy alignment transfers; the paper does not report that correlation.
  • If the method generalizes to different particle counts — which the authors list as future work — the state representation would need distribution moments or embeddings rather than raw particle coordinates.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper proposes DOA, a Proximal Policy Optimization (PPO) agent that detects degeneracy in particle-filter-based 2D SLAM and outputs a scalar compensation factor a in (0,1). The agent observes the particle positions of the observation distribution p_z and the motion-model distribution p_u, and the SLAM system shifts p_z toward p_u according to the linear interpolation g_c = (1-a)g_z + a g_u. Training uses a reward (Eq. 18) based on particle-covariance shrinkage, scan-matching likelihood, effective particle number, and action smoothness; a transfer-learning stage pre-trains in a non-degenerate scene and fine-tunes a frozen backbone in degenerate scenes. Experiments report detection success ratios around 90% and ATE reductions relative to GMapping of 35.8% (S1) and 81.8%, 93.8%, 95.2% (S2-S4), as well as reduced resampling frequency in real corridors.

Significance. The manuscript addresses a real operational problem, localization drift of particle-filter SLAM in geometry-poor corridors, and proposes a DRL-based adaptive fusion mechanism rather than another fixed-rule heuristic. If the results hold, the method would avoid supervised degeneracy annotations and hand-tuned thresholds, and the transfer-learning scheme is a practical addition. The paper also includes ablation studies, comparisons against SAC and DDPG, runtime measurements, and real-robot validation. However, the significance is conditional on fixing three load-bearing issues: the fusion formula in Eq. (14) is not derived from Eq. (12), the detection threshold in Section VII-B is fitted to the evaluation data, and the reward in Eq. (18) is not shown to correlate with ground-truth pose error.

major comments (3)
  1. [III, Eqs. (12)-(14)] The jump from the information-form fusion to the scalar interpolation is not justified. Writing Eq. (12) explicitly, the fused mean is g_c = ((1-a)Σ_z^{-1}+aΣ_u^{-1})^{-1}((1-a)Σ_z^{-1}g_z+aΣ_u^{-1}g_u). Eq. (14) instead solves a Euclidean weighted least-squares problem and yields g_c=(1-a)g_z+a g_u. These two expressions coincide only under special conditions on Σ_z and Σ_u (e.g., isotropic or proportional covariance matrices), which do not hold in a degenerate corridor where p_z is stretched anisotropically along the corridor. Please provide a derivation of Eq. (14) from Eq. (12), or replace Eq. (14) with the actual covariance-weighted mean and update the compensation procedure in Fig. 2 accordingly.
  2. [VII-B and Table III] The threshold used to convert the agent output into a binary degeneracy flag is chosen after inspecting the same test curves: the text states 'We set our degeneracy threshold at 0.75, as agent outputs in degenerate scenes are mostly above this value.' This makes the reported success rates partly self-confirming and biases the comparison against Switch-SLAM and Li et al., whose thresholds are fixed constants. Please select the threshold on a held-out calibration set or by a principled criterion that does not use the evaluation scenes, and then report the resulting success ratios.
  3. [V-B, Eq. (18); Tables IV-V] The reward is not validated against the ground-truth pose error that the ATE numbers are meant to reflect. In a degenerate corridor, shifting p_z toward p_u mechanically reduces the covariance of the fused distribution and can increase N_eff because particles become more similar, regardless of whether the fused centroid is closer to the true pose; the scan-matching likelihood s_t is flat along the degenerate direction and provides little corrective signal there. Indeed, Eq. (13) shows that in the degenerate limit the fused information matrix is approximately aΣ_u^{-1}, so the covariance term of the reward effectively encourages trusting the odometry model. No experiment reports a correlation between the reward (or its components) and ATE against Gazebo ground truth. Please add such an analysis or revise the reward to include a ground-truth-based term.
minor comments (5)
  1. [General] There are multiple typographical errors (e.g., 'dateset' in the problem formulation) and awkward phrasings; a thorough proofreading pass is needed.
  2. [Eq. (18)] The term s_t^{(i)} in the reward is not defined precisely; clarify whether it is the likelihood of the best particle, the mean likelihood, or the raw scan-matching score, and state how the four reward terms are normalized before weighting.
  3. [Fig. 4] The caption of Fig. 4 mixes reward curves and the experimental robot; consider splitting the equipment photograph into a separate figure and define the abbreviations (E1-E8) directly in the caption.
  4. [Table III] The ablation labels (P, C, TL, FBN, HN) are undefined in the table caption; define them and report the number of episodes or runs used to compute each success ratio.
  5. [VI-B] The sentence 'transfer learning does not leak the privileged information of the environment into the policy network' is asserted without support; if it is kept, it should be justified by stating exactly what information the frozen backbone receives and why no privileged ground-truth information is encoded.

Circularity Check

1 steps flagged · score 5.0 of 10

Detection success rate is partly threshold-fitted; ATE evaluation remains externally grounded.

  1. fitted input called prediction [Section VII-B 'Degeneracy Detection', paragraph after Fig. 7; results in Table III.]
    "We set our degeneracy threshold at 0.75, as agent outputs in degenerate scenes are mostly above this value."

    The reported detection success ratios in Table III are computed against the 0.75 threshold, but the threshold is chosen from the same agent outputs it is then used to score: the text states the value was selected because 'agent outputs in degenerate scenes are mostly above this value.' No separate validation split or threshold-selection procedure is described, so the ~90% success rate is partly enforced by construction. Placing a threshold just below the bulk of the degenerate-scene outputs will trivially classify those outputs as detections; the metric therefore is a fitted-threshold evaluation rather than an independent prediction of degeneracy.

full rationale

The central optimization claim in Table IV is not circular: ATE is measured against Gazebo ground truth for GMapping, GMapping+DOA, and baseline methods, and those numbers are not generated by the training reward or by the detection threshold. The method comparison is also not carried by a load-bearing self-citation chain: [19] is used as a baseline and is not invoked to justify the DOA design. The one clear circular step is the detection evaluation. The paper fixes the degeneracy threshold at 0.75 because 'agent outputs in degenerate scenes are mostly above this value' and then reports roughly 90% success ratios against that same threshold in Table III, making the detection-success headline partly self-confirming. A separate validation split, or a threshold chosen without inspecting the test outputs, would be needed for the success ratio to be an independent prediction. Separately, the reward in Eq. (18) relies on covariance, scan-matching likelihood, and Neff proxies that the fusion action itself can mechanically influence (Eqs. 12-14), and the paper does not validate those proxies against the Gazebo ATE used in Table IV. This is a reward-misspecification and correctness risk, not a derivation-level circularity, so it does not by itself raise the circularity score further.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The method introduces no new physical entities; its new ingredient is a learned scalar a and the training framework. The central derivation rests on Gaussian assumptions about particle distributions and on a reward design whose proxy validity is unverified. The detection evaluation adds a post hoc threshold, which is a fitted parameter. These are the main ledger items.

free parameters (4)
  • Reward weights β1..β4 = β1=0.3, β2=0.3, β3=0.2, β4=0.2
    Chosen empirically as 'the best combination derived through multiple training iterations' (Section V-B), not derived from theory.
  • Degeneracy threshold = 0.75
    Set after observing that agent outputs in degenerate scenes are mostly above this value (Section VII-B). The reported detection success rate depends directly on this post hoc threshold.
  • Number of particles = 30
    Set by the authors for the agent input dimension (120 = 30 particles × 4 coordinates); not justified by sensitivity analysis.
  • PPO hyperparameters = lr 0.002, discount 0.98, clip 0.2, Tepochs 5, timesteps 200, episodes 1000
    Reported as 'empirically best combinations derived from multiple experiments' (Table II and Section VI-A). Standard RL hyperparameters, but the values are hand chosen.
assumptions (5)
  • standard math Particle filter posterior factorizes as p(xt,M|zt,ut) = p(M|xt,z1:t)p(xt|zt,ut)
    Eq. (1), standard Rao-Blackwellized particle filter assumption for GMapping.
  • domain assumption In degenerate environments the observation information matrix Izm is rank deficient and the motion model distribution pu is more reliable than pz
    Section III, around Eqs. (7)-(9) and Fig. 2; this motivates the entire compensation strategy and is not empirically established with measurements.
  • domain assumption The particle clouds pu and pz can be treated as Gaussian distributions for the fusion equations (12)-(14)
    Section III uses Gaussian fusion formulas, although particle filter distributions are generally non-Gaussian; no normality test is reported.
  • domain assumption Reward terms in Eq. (18) are valid proxies for localization accuracy
    Section V-B; the reward is the only learning signal, but its correlation with ground-truth ATE is never measured.
  • ad hoc to paper Transfer learning with frozen backbone preserves learned general features and does not leak privileged information
    Section VI-B states this, but no analysis or ablation supports the no-leakage claim.

how reviews work

0 comments
Cite this review

Pith. "Pith review of DOA: A Degeneracy Optimization Agent with Adaptive Pose Compensation Capability based on Deep Reinforcement Learning." pith.science (2026). https://pith.science/paper/5BCXQB7D

@misc{pith2026250719742,
  author       = {Pith},
  title        = {Pith review of: DOA: A Degeneracy Optimization Agent with Adaptive Pose Compensation Capability based on Deep Reinforcement Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5BCXQB7D}},
  note         = {Machine review of arXiv:2507.19742}
}
read the original abstract

Particle filter-based 2D-SLAM is widely used in indoor localization tasks due to its efficiency. However, indoor environments such as long straight corridors can cause severe degeneracy problems in SLAM. In this paper, we use Proximal Policy Optimization (PPO) to train an adaptive degeneracy optimization agent (DOA) to address degeneracy problem. We propose a systematic methodology to address three critical challenges in traditional supervised learning frameworks: (1) data acquisition bottlenecks in degenerate dataset, (2) inherent quality deterioration of training samples, and (3) ambiguity in annotation protocol design. We design a specialized reward function to guide the agent in developing perception capabilities for degenerate environments. Using the output degeneracy factor as a reference weight, the agent can dynamically adjust the contribution of different sensors to pose optimization. Specifically, the observation distribution is shifted towards the motion model distribution, with the step size determined by a linear interpolation formula related to the degeneracy factor. In addition, we employ a transfer learning module to endow the agent with generalization capabilities across different environments and address the inefficiency of training in degenerate environments. Finally, we conduct ablation studies to demonstrate the rationality of our model design and the role of transfer learning. We also compare the proposed DOA with SOTA methods to prove its superior degeneracy detection and optimization capabilities across various environments.

Figures

Figures reproduced from arXiv: 2507.19742 by the authors.

Figure 1
Figure 1. The degeneracy problem in Lidar-SLAM is particularly [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. The upper part is overall process. Lack of observation constraint makes red particle point set diverge, reducing localization [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. The architecture diagram of agent. The input part is divided into upper and lower parts, representing the data of two [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: Reward curve of training process and experimental equipments (blue=E1=policy+critic+TL, green=E2=policy+critic, [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Schematic diagram of transfer learning. Nef f = 1 PN i=1 w (i) · w (i) (20) where w (i) represents the normalized weight of ith particle. In addition, ||at −at−1||2 constrains the difference between consecutive degeneracy factors, suppressing policy oscillation and ens…
Figure 6
Figure 6. Figure 6: Diagram of the experimental environment with red motion trajectories. [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 8
Figure 8. Figure 8: Real experimental results of degeneracy detection. [PITH_FULL_IMAGE:figures/full_fig_p007_8.png]
Figure 7
Figure 7. Figure 7: Simulation results of degeneracy detection. [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]
Figure 9
Figure 9. Figure 9: Maps of experiments. The red line representing the true [PITH_FULL_IMAGE:figures/full_fig_p008_9.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

28 extracted references · 19 canonical work pages

  1. [19]

    Anti-degeneracy scheme for lidar slam based on particle filter in geometry feature-less environments,

    Y . Li and W. Zhang, “Anti-degeneracy scheme for lidar slam based on particle filter in geometry feature-less environments,” IEEE Robotics and Automation Letters, vol. 10, DOI 10.1109/LRA.2025.3568569, no. 7, pp. 6784–6791, 2025

    work page arXiv 2025

  2. [1]

    Proximal policy optimization algorithms,

    J. Schulman, “Proximal policy optimization algorithms,” 2017. [Online]. Available: https://arxiv.org/abs/1707.06347

    arXiv 2017

  3. [2]

    Improved techniques for grid mapping with rao- blackwellized particle filters,

    G. Grisetti, “Improved techniques for grid mapping with rao- blackwellized particle filters,” IEEE transactions on Robotics , vol. 23, no. 1, pp. 34–46, 2007

    work page 2007

  4. [3]

    Real-time loop closure in 2d lidar slam,

    W. Hess, “Real-time loop closure in 2d lidar slam,” in 2016 IEEE International Conference on Robotics and Automation (ICRA) , DOI 10.1109/ICRA.2016.7487258, pp. 1271–1278, 2016

    arXiv 2016

  5. [4]

    A flexible and scalable slam system with full 3d motion estimation,

    S. Kohlbrecher, “A flexible and scalable slam system with full 3d motion estimation,” in 2011 IEEE international symposium on safety, security, and rescue robotics , pp. 155–160. IEEE, 2011

    work page 2011

  6. [5]

    Orb-slam3: An accurate open-source library for visual, visual–inertial, and multimap slam,

    C. Campos, “Orb-slam3: An accurate open-source library for visual, visual–inertial, and multimap slam,” IEEE Transactions on Robotics , vol. 37, no. 6, pp. 1874–1890, 2021

    work page 2021

  7. [6]

    A fast and accurate visual inertial odometry using hybrid point-line features,

    Z. Chen, “A fast and accurate visual inertial odometry using hybrid point-line features,” IEEE Robotics and Automation Letters , vol. 9, DOI 10.1109/LRA.2024.3490406, no. 12, pp. 11 345–11 352, 2024

    arXiv 2024

  8. [7]

    High-speed stereo visual slam for low-powered comput- ing devices,

    A. Kumar, “High-speed stereo visual slam for low-powered comput- ing devices,” IEEE Robotics and Automation Letters , vol. 9, DOI 10.1109/LRA.2023.3329621, no. 1, pp. 499–506, 2024

    arXiv 2023

Show all 28 references
  1. [8]

    Airslam: An efficient and illumination-robust point-line visual slam system,

    K. Xu and S. Yuan, “Airslam: An efficient and illumination-robust point-line visual slam system,” IEEE Transactions on Robotics , DOI 10.1109/TRO.2025.3539171, pp. 1–20, 2025

    2025

  2. [9]

    Loam: Lidar odometry and mapping in real-time

    J. Zhang, “Loam: Lidar odometry and mapping in real-time.” in Robotics: Science and systems , vol. 2, no. 9, pp. 1–9. Berkeley, CA, 2014

    2014

  3. [10]

    Lego-loam: Lightweight and ground-optimized lidar odometry and mapping on variable terrain,

    T. Shan, “Lego-loam: Lightweight and ground-optimized lidar odometry and mapping on variable terrain,” in 2018 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pp. 4758–4765. IEEE, 2018

    2018

  4. [11]

    Fast-livo2: Fast, direct lidar–inertial–visual odometry,

    C. Zheng, “Fast-livo2: Fast, direct lidar–inertial–visual odometry,” IEEE Transactions on Robotics, vol. 41, DOI 10.1109/TRO.2024.3502198, pp. 326–346, 2025

    2024

  5. [12]

    Liver: A tightly coupled lidar-inertial-visual state estimator with high robustness for underground environments,

    T. Wen, “Liver: A tightly coupled lidar-inertial-visual state estimator with high robustness for underground environments,” IEEE Robotics and Automation Letters, vol. 9, DOI 10.1109/LRA.2024.3355778, no. 3, pp. 2399–2406, 2024

    2024

  6. [13]

    Intermittent vio-assisted lidar slam against degeneracy: Recog- nition and mitigation,

    J. Xu, “Intermittent vio-assisted lidar slam against degeneracy: Recog- nition and mitigation,” IEEE Transactions on Instrumentation and Mea- surement, vol. 74, DOI 10.1109/TIM.2024.3507053, pp. 1–13, 2025

    2024

  7. [14]

    Imu augment tightly coupled lidar-visual- inertial odometry for agricultural environments,

    Q. H. Hoang and G.-W. Kim, “Imu augment tightly coupled lidar-visual- inertial odometry for agricultural environments,” IEEE Robotics and Automation Letters , vol. 9, DOI 10.1109/LRA.2024.3440728, no. 10, pp. 8483–8490, 2024

    2024

  8. [15]

    Switch-slam: Switching-based lidar-inertial-visual slam for degenerate environments,

    J. Lee, “Switch-slam: Switching-based lidar-inertial-visual slam for degenerate environments,” IEEE Robotics and Automation Letters , 2024

    2024

  9. [16]

    P2d-do: Degeneracy optimization for lidar slam with point-to- distribution detection factors,

    W. Chen, “P2d-do: Degeneracy optimization for lidar slam with point-to- distribution detection factors,” IEEE Robotics and Automation Letters , DOI 10.1109/LRA.2024.3522839, pp. 1–8, 2024

    2024

  10. [17]

    Lidar/uwb fusion based slam with anti-degeneration capa- bility,

    H. Zhou, “Lidar/uwb fusion based slam with anti-degeneration capa- bility,” IEEE Transactions on V ehicular Technology , vol. 70, no. 1, pp. 820–830, 2020

    2020

  11. [18]

    Learning-based localizability estimation for robust lidar localization,

    J. Nubert, “Learning-based localizability estimation for robust lidar localization,” in 2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pp. 17–24. IEEE, 2022

    2022

  12. [20]

    X-icp: Localizability-aware lidar registration for robust local- ization in extreme environments,

    T. Tuna, “X-icp: Localizability-aware lidar registration for robust local- ization in extreme environments,” IEEE Transactions on Robotics , 2023

    2023

  13. [21]

    Pointnetkl: Deep inference for gicp covariance estimation in bathymetric slam,

    I. Torroba, “Pointnetkl: Deep inference for gicp covariance estimation in bathymetric slam,” IEEE Robotics and Automation Letters , vol. 5, no. 3, pp. 4078–4085, 2020

    2020

  14. [22]

    Cello-3d: Estimating the covariance of icp in the real world,

    D. Landry, “Cello-3d: Estimating the covariance of icp in the real world,” in 2019 International Conference on Robotics and Automation (ICRA) , pp. 8190–8196. IEEE, 2019

    2019

  15. [23]

    Reinforcement learning helps slam: Learning to build maps,

    N. Botteghi, “Reinforcement learning helps slam: Learning to build maps,” The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences , vol. 43, pp. 329–335, 2020

    2020

  16. [24]

    Reinforcement learning meets visual odometry,

    N. Messikommer, “Reinforcement learning meets visual odometry,” in European Conference on Computer Vision , pp. 76–92. Springer, 2024

    2024

  17. [25]

    Continuous control with deep reinforcement learning,

    T. P. Lillicrap, “Continuous control with deep reinforcement learning,”

  18. [26]

    Soft actor-critic for discrete action settings,

    P. Christodoulou, “Soft actor-critic for discrete action settings,” 2019. [Online]. Available: https://arxiv.org/abs/1910.07207

    2019 arXiv

  19. [27]

    evo: Python package for the evaluation of odometry and slam

    M. Grupp, “evo: Python package for the evaluation of odometry and slam.” https://github.com/MichaelGrupp/evo, 2017

    2017

  20. [2019]

    Available: https://arxiv.org/abs/1509.02971

    [Online]. Available: https://arxiv.org/abs/1509.02971

    arXiv

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.