Pith. sign in

REVIEW 4 major objections 6 minor 40 references

A mapless navigation system reaches 90% success on outdoor routes by generating hundreds of trajectory hypotheses with a CVAE and selecting the best one with a lightweight vision-language model.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

A lightweight CLIPSeg semantic scorer selects among 200 CVAE-generated trajectories, giving 90% success on 120-240 m mapless outdoor routes.

T0 review reviewed 2026-08-03 challenge →

load-bearing objection The architecture is a real step forward for onboard open-vocabulary trajectory selection, but the headline numbers don't all check out: the map-based upper bound claim is missing, and 'unseen environments' is just a different sector of the same campus. the 4 major comments →

arxiv 2602.01429 v2 pith:DSUMG7OB submitted 2026-02-01 cs.RO

Sem-NaVAE: Semantically-Guided Outdoor Mapless Navigation via Generative Trajectory Priors

Gonzalo Olgu\'in , Javier Ruiz-del-Solar This is my paper
classification cs.RO
keywords mapless navigationconditional variational autoencodertrajectory generationopen-vocabulary segmentationvision-language modelsoutdoor robot navigationgenerate-and-selectCLIPSeg
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

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 argues that outdoor robot navigation without a global map can be solved as a two-stage 'generate and select' problem: a conditional variational autoencoder (CVAE) proposes a large, diverse set of kinematically feasible trajectories from LiDAR, goal, and state history, and a lightweight vision-language model (CLIPSeg) scores each candidate by how well it matches user-specified terrain classes (e.g., pavement vs. grass) and geometric progress toward the goal. The paper reports that on unseen test routes of 120–240 m, the resulting system, Sem-NaVAE, reaches the goal 90% of the time, outperforming the strongest baseline by 10% and coming within 7% of a map-based planner. The significance would be showing that semantic commonsense (open-vocabulary terrain preferences) can be injected at decision time without requiring a metric map or a large language model API, while retaining real-time operation. The paper also presents a training objective that preserves multimodality in the generator—avoiding the trajectory-averaging failure of deterministic planners—by using a log-mean-exp reconstruction loss that acts as a soft winner-takes-all.

Core claim

The paper's central claim is that a mapless outdoor navigation system built from a CVAE trajectory generator and a CLIPSeg-based trajectory selector achieves a 90% success rate on 120–240 m routes in an unseen sector of the training campus, beating the MTG baseline and a retrained variant (MTG') and staying within 7 percentage points of a map-based upper bound. The generator produces more than 200 diverse, kinematically feasible waypoint sequences using a learned prior, a log-mean-exp (importance-weighted) reconstruction loss on the prior that keeps distinct trajectory modes alive, and a collision loss against an offline semantic map. The selector projects candidate trajectories into the cam

What carries the argument

The core mechanism is the 'generate-and-select' loop. On the generation side, a NaVAE—a CVAE with a PointNet front end, LSTM/MLP encoders, and an autoregressive GRU decoder outputting velocity distributions—is trained with a log-mean-exp reconstruction loss on the prior (to preserve multimodality) plus a collision loss. On the selection side, CLIPSeg zero-shot segmentation builds a semantic cost map from natural-language terrain prompts, and candidate waypoints are projected into the image with an occlusion mask and scored by discounted semantic cost plus a geometric goal-progress term. An asynchronous update rule re-costs the currently executed trajectory at the perception frequency and swi

Load-bearing premise

The whole pipeline depends on the 1,706-sample training set from 15 runs on a single campus—with traversability labels generated offline from a static global semantic map—being representative enough that the PointNet heatmap and the CVAE prior transfer to unseen sectors of the same campus; if that campus map is stale, misaligned, or the environment type changes, the reported success rate will not transfer.

What would settle it

Run Sem-NaVAE on a different outdoor campus or park not used in training, keeping the same class prompts and costs; if success rate drops below roughly 0.63 (the MTG' baseline) or EPT falls well below 95%, the central claim about navigating 'unseen environments' is falsified. More locally, deliberately perturb the global semantic map used to produce training labels (e.g., shift it by 1 m) and retrain: if the 90% success rate collapses, the offline-label premise is load-bearing.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • If the reported numbers hold, mapless navigation can close most of the gap to map-based planning using an inexpensive open-vocabulary segmentation model rather than a full-scale VLM API, removing the need for prior map construction and an internet connection.
  • The 'generate many, select semantically' architecture is a concrete recipe for avoiding the unimodality problem in learned planners: a log-mean-exp objective can keep multiple valid modes alive while training with only one observed future per context.
  • The system is agnostic to the local planner: the paper shows nearly identical success rates with both a reinforcement-learning-based planner and the DWA planner, implying the framework can be layered onto existing controllers.
  • User constraints are modular and changeable at runtime: varying the cost of a class like 'grass' or adding a new class like 'sand' changes the robot's route selection in the reported experiments.
  • The asynchronous cost-comparison update is a concrete mechanism to balance stability and reactivity under noisy, occluded perception, and could be adopted by other receding-horizon planners.

Where Pith is reading between the lines

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

  • The evaluation covers only held-out sectors of the same campus where training data were collected; the natural next test is deployment on a different outdoor environment (different appearance, terrain, and layout) to see whether the 90% success rate is a property of the method or of the campus distribution.
  • The occlusion mask is a proxy for depth; replacing it with an actual depth source could reduce misclassifications of trajectories that pass behind obstacles and likely tighten the gap to map-based upper bounds.
  • The traversability labels come from an offline static global semantic map that is not available at inference; if that map were stale or misaligned, the PointNet heatmap and prior would be biased, so an online self-supervised label source (e.g., proprioceptive feedback) is a plausible substitute for true mapless training.
  • The soft winner-takes-all formulation of the reconstruction loss is a general idea for CVAE-based forecasting beyond navigation (e.g., pedestrian or vehicle trajectory prediction), where multiple socially acceptable futures exist and averaging is known to fail.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper proposes Sem-NaVAE, a mapless outdoor navigation system that combines a CVAE-based trajectory generator with an open-vocabulary semantic selector built on CLIPSeg. The generator produces a large set of kinematically feasible trajectory hypotheses from LiDAR, goal, and past-state context, while the selector projects candidates into an FPV semantic costmap, scores them via a weighted combination of semantic and goal costs, and switches trajectories asynchronously. The system is trained on 1,706 samples collected on the University of Chile engineering campus and evaluated in an unseen sector of that campus across five routes of 120–240 m, each repeated six times. Reported results show Sem-NaVAE achieving SR 0.90, SPL 0.76, EPT 95.0%, NTR 23.4%, and T_ratio 1.33, against MTG' (SR 0.63) and MTG (SR 0.17). Ablations study the loss function, collision loss, selection VLM, local planner, and update policy.

Significance. If the claims hold, Sem-NaVAE is a meaningful step in mapless outdoor navigation: it demonstrates a practical generate-and-select architecture where a lightweight open-vocabulary model provides semantic trajectory selection, with real-robot validation against published baselines. The asynchronous update and recovery mechanisms are plausible contributions. However, the paper's headline claims are broader than the evidence: the experiments are limited to one campus, no map-based upper bound is reported, and the statistical basis of the central comparison is thin. The core idea is sound and worth pursuing, but the current evidence base is incomplete.

major comments (4)
  1. [Abstract and Section IV] The abstract claims 90% success 'in unseen environments' and 'within 7% of a map-based upper bound'. The results section only describes tests in an 'unseen sector' of the same campus used for training (Section IV-A), and no map-based upper bound is defined or reported anywhere in the paper. These are headline claims; they must either be supported with a cross-site evaluation and an explicit map-based comparison, or be removed/qualified.
  2. [Tables I and II] Sem-NaVAE success rate is reported as 0.90 in Table I but 6/6 in Table II, with NTR 23.4 vs 23.8 and T_ratio 1.33 vs 1.34. The tables correspond to different experimental protocols (5 routes × 6 repetitions vs 2 routes × 3 repetitions), but this is not stated in the text. Because these numbers are the evidence for the central claim, the discrepancy must be explained and a consistent aggregate reported.
  3. [Section IV-B, Tables I and II] No error bars, confidence intervals, or per-route success counts are provided. With only six trials per route and five routes, the 27-point SR gap between Sem-NaVAE and MTG' could be driven by a single route or by high variance. Report per-route results and variance measures so the reader can assess whether the improvement over the nearest baseline is statistically meaningful.
  4. [Section IV-A] The PointNet traversability labels are generated offline from a static global semantic map of the training campus, and the test sector is on the same campus. The abstract's phrase 'unseen environments' therefore overstates the generality of the results. The authors should either provide an evaluation in a geographically distinct environment (e.g., a different campus or park) or explicitly restrict the claim to same-campus unseen sectors. This also bears on the missing 'map-based upper bound' claim.
minor comments (6)
  1. [Title/Abstract] The system name is inconsistently spaced: 'Sem-NaV AE' in the title and abstract, 'Sem-NaVAE' in tables and text. Please unify.
  2. [Eq. (4)] The KL term is written as β DKL(pθ||qϕ). In standard CVAE training this direction is usually reversed (KL(qϕ||pθ)). The authors should clarify whether the reverse direction is intentional and how the gradients behave.
  3. [Section IV-A] The collision loss weight is said to be 'set to 10' and denoted α, but Eq. (7) uses λ. Please use consistent notation.
  4. [Section IV-B] Typo: 'por trajectory selection' should be 'for trajectory selection'.
  5. [Section IV-C / Fig. 5] The qualitative comparison of class-cost variation is only described briefly. It would be helpful to state how many runs were performed and whether the observed behavior changes were consistent across runs.
  6. [Section IV-B / Eq. (11)] The occlusion mask depends on the same semantic costmap used for scoring; a thresholded cost value determines both the mask and the penalty. This circularity is a known limitation of 2D projection; a short discussion of its effect and of alternative depth-aware approaches would improve the paper.

Circularity Check

0 steps flagged

No significant circularity: central CVAE/CLIPSeg derivation is self-contained, and the only same-author citation is not load-bearing.

full rationale

Walking the claimed derivation chain: (1) trajectory generation optimizes a standard CVAE ELBO with KL, log-mean-exp reconstruction, and collision penalty (Eqs. 4-7); (2) trajectory selection builds a dense cost from CLIPSeg class probabilities and a geometric goal term (Eqs. 9-13); (3) switching uses hysteresis (Eq. 14). None of these equations reuses its output as an input or defines the predicted quantity as the fitted quantity. Training uses a static semantic map for labels and collision supervision, but the paper states the map is unavailable at inference; the headline SR/SPL results are measured by physical goal reaching against external baselines (MTG, MTG') and ablation variants (DWA, GPT, fixed-frequency updates), not against the model's own scoring functions. The only same-author citation is the RL local planner [40]; Table II shows DWA achieves similar SR/SPL, so that citation is not load-bearing. The abstract's 'map-based upper bound' is never defined or reported in the results, a completeness/evidence gap, but without a definition no circular reduction can be exhibited. No specific equation or fitted parameter is renamed as a prediction; therefore no circular steps are established.

Axiom & Free-Parameter Ledger

6 free parameters · 5 axioms · 0 invented entities

The system is empirical; no new physical entities or mathematical axioms are introduced. The defaults are many hand-set costs/thresholds that directly shape the selection cost and thus the reported success rate. The largest domain assumption is that a static global semantic map of one campus yields training labels that transfer to unseen areas of the same campus.

free parameters (6)
  • semantic class costs C = 'pavement':0, 'tree':3, 'grass':2, 'wall':3, 'stairs':3, 'person':3, 'hole':3, 'sky':4
    Manually chosen by the authors; directly determine J(y') via Eq. (9). Changing them changes the selected trajectory; no systematic tuning procedure reported.
  • discount factor γ and goal weights α1, α2 = 0.8, 2, 0.2
    Set for all experiments; define J_goal and future discounting in Eqs. (11)-(12).
  • occlusion penalty C_u and threshold T_occ = 2 and 2
    Hand-set; C_u adds a fixed penalty when projective mask u_j=1 in Eq. (11).
  • collision loss weight λ = 10 (referred to as α in §IV-A)
    Weight of L_col in Eq. (7); affects generated trajectory feasibility.
  • KL weight β = annealed, not numerically reported
    KL annealing factor in Eq. (7); chosen heuristically during training.
  • maximum slope θ_max = not reported
    Safety threshold in Eq. (8) for collision filtering; exact value absent from the text.
axioms (5)
  • domain assumption PointNet training labels from the global semantic map are correct enough to learn traversability
    Section IV-A: labels are generated by projecting points onto a static global semantic map; if the map is stale or misaligned, generator and collision filter inherit those errors.
  • domain assumption CLIPSeg open-vocabulary probabilities transfer to unseen outdoor scenes without fine-tuning
    Section III-B: the selector uses zero-shot CLIPSeg maps to score trajectories; tests are on an unseen sector but same campus, so distribution shift is mild.
  • domain assumption The elevation map from LiDAR and the slope filter θ_i > θ_max are sufficient to reject kinematically infeasible trajectories
    Section III-B.1: uses an 18m x 18m forward-facing elevation map at 0.1m/px.
  • standard math Variational lower bound / IWAE log-mean-exp reconstruction is a valid objective to maximize multimodal coverage
    Eqs. (4)-(5); standard ELBO/IWAE machinery from cited works.
  • domain assumption The RL local planner [40] can safely execute the selected waypoints and avoid unseen obstacles
    Used for all reported runs; authors claim system is planner-agnostic based on the DWA ablation.

reviewed 2026-08-03 · how reviews work

0 comments
Cite this review

Pith. "Pith review of Sem-NaVAE: Semantically-Guided Outdoor Mapless Navigation via Generative Trajectory Priors." pith.science (2026). https://pith.science/paper/DSUMG7OB

@misc{pith2026260201429,
  author       = {Pith},
  title        = {Pith review of: Sem-NaVAE: Semantically-Guided Outdoor Mapless Navigation via Generative Trajectory Priors},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DSUMG7OB}},
  note         = {Machine review of arXiv:2602.01429}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

This work presents a mapless navigation approach for outdoor applications. It combines the exploratory capacity of conditional variational autoencoders (CVAEs) to generate trajectories and the semantic segmentation capabilities of a lightweight visual language model (VLM) to select the trajectory to execute. Open-vocabulary segmentation is used to score and select the generated trajectories based on natural language, and a state-of-the-art local planner executes velocity commands. One of the key features of the proposed approach is its ability to generate a large variability of trajectories and select them to navigate in real-time. In real-world outdoor experiments, Sem-NaVAE achieves a 90% success rate across routes of 120-240m in unseen environments, outperforming the nearest baseline by 10% while remaining within 7% of a map-based upper bound. A video showing an experimental run of the system can be found in https://youtu.be/i3R5ey5O2yk.

Figures

Figures reproduced from arXiv: 2602.01429 by Gonzalo Olgu\'in, Javier Ruiz-del-Solar.

Figure 1
Figure 1. Figure 1: Examples of autonomous mapless navigation. Six timesteps are shown [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Sem-NaVAE overview: A generator module proposes a number of motion hypotheses based on sensory information. These hypotheses are then filtered by collisions and projected onto and FPV semantic map constructed with a lightweight VLM. Both semantic and goal-distance costs are used to then select the optimal trajectory. Instead of perdiodically updating to a new trajectory at every inference step, the new bes… view at source ↗
Figure 3
Figure 3. Figure 3: NaVAE architecture. The model takes as inputs consecutive pointclouds, past trajectory and navigation goal in polar coordinates. A pre-trained PointNet [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Comparison of generation baselines MTG [23], its PointNet modifi [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Comparison of followed trajectories when varying selection classes [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗

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

40 extracted references · 5 linked inside Pith

  1. [1]

    A Survey of Deep Learn- ing Techniques for Autonomous Driving,

    S. Grigorescu, B. Trasnea, T. Cocias, et al., “A Survey of Deep Learn- ing Techniques for Autonomous Driving,”Journal of Field Robotics, vol. 37, no. 3, pp. 362–386, 2020

    2020

  2. [2]

    End-to-End Deep Learning for Autonomous Navigation of Mobile Robot,

    Y .-H. Kim, J.-I. Jang, and S. Yun, “End-to-End Deep Learning for Autonomous Navigation of Mobile Robot,” inIEEE International Conference on Consumer Electronics (ICCE), 2018, pp. 1–6

    2018

  3. [3]

    Mobile Robot Navigation Based on Deep Reinforcement Learning,

    X. Ruan, D. Ren, X. Zhu, et al., “Mobile Robot Navigation Based on Deep Reinforcement Learning,” inChinese Control And Decision Conference (CCDC), 2019, pp. 6174–6178

    2019

  4. [4]

    Fast Traversability Estimation for Wild Visual Navigation,

    J. Frey, M. Mattamala, N. Chebrolu, et al., “Fast Traversability Estimation for Wild Visual Navigation,” inProceedings of Robotics: Science and Systems (RSS), 2023

    2023

  5. [5]

    End-to-end driving via conditional imitation learning,

    F. Codevilla, M. M ¨uller, A. L ´opez, V . Koltun, and A. Dosovitskiy, “End-to-end driving via conditional imitation learning,” in2018 IEEE International Conference on Robotics and Automation (ICRA), 2018, pp. 4693–4700

    2018

  6. [6]

    Implicit behavioral cloning,

    P. Florence, C. Lynch, A. Zeng, O. A. Ramirez, et al., “Implicit behavioral cloning,” inConference on Robot Learning, PMLR, 2022, pp. 158–168

    2022

  7. [7]

    Social GAN: Socially Ac- ceptable Trajectories with Generative Adversarial Networks,

    A. Gupta, J. Johnson, L. Fei-Fei, et al., “Social GAN: Socially Ac- ceptable Trajectories with Generative Adversarial Networks,” inIEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018, pp. 2635–2644

    2018

  8. [8]

    NoMaD: Goal Masked Diffu- sion Policies for Navigation and Exploration,

    A. Sridhar, D. Shah, C. Glossop, et al., “NoMaD: Goal Masked Diffu- sion Policies for Navigation and Exploration,” inIEEE International Conference on Robotics and Automation (ICRA), 2024, pp. 12 224– 12 230

    2024

  9. [9]

    VLM-Social-Nav: Socially Aware Robot Navigation Through Scoring Using Vision-Language Models,

    D. Song, J. Liang, A. Payandeh, et al., “VLM-Social-Nav: Socially Aware Robot Navigation Through Scoring Using Vision-Language Models,”IEEE Robotics and Automation Letters, vol. 10, no. 1, pp. 508–515, 2025

    2025

  10. [10]

    LM-Nav: Robotic Navigation with Large Pre-Trained Models of Language, Vision, and Action,

    D. Shah, B. Osinski, B. Ichter, and S. Levine, “LM-Nav: Robotic Navigation with Large Pre-Trained Models of Language, Vision, and Action,” inProceedings of the 6th Conference on Robot Learning (CoRL), PMLR, 2022, pp. 259–269

    2022

  11. [11]

    Stanley: The robot that won the darpa grand challenge,

    S. Thrun, M. Montemerlo, H. Dahlkamp, D. Stavens, et al., “Stanley: The robot that won the darpa grand challenge,”Journal of Field Robotics, vol. 23, no. 9, pp. 661–692, 2006

    2006

  12. [12]

    Where Should I Walk? Predicting Terrain Properties From Images Via Self-Supervised Learning,

    L. Wellhausen, A. Dosovitskiy, R. Ranftl, et al., “Where Should I Walk? Predicting Terrain Properties From Images Via Self-Supervised Learning,”IEEE Robotics and Automation Letters, vol. 4, no. 2, pp. 1509–1516, 2019

    2019

  13. [13]

    Incorporating Multi-Context Into the Traversability Map for Urban Autonomous Driving Using Deep Inverse Reinforcement Learning,

    C. Jung and D. H. Shim, “Incorporating Multi-Context Into the Traversability Map for Urban Autonomous Driving Using Deep Inverse Reinforcement Learning,”IEEE Robotics and Automation Letters, vol. 6, no. 2, pp. 1662–1669, 2021

    2021

  14. [14]

    Guaman Castro et al.,How does it feel? self-supervised costmap learning for off-road vehicle traversability, IEEE, 2023

    M. Guaman Castro et al.,How does it feel? self-supervised costmap learning for off-road vehicle traversability, IEEE, 2023

    2023

  15. [15]

    IN-Sight: Interactive Navigation through Sight,

    P. Schoch, F. Yang, Y . Ma, et al., “IN-Sight: Interactive Navigation through Sight,” inIEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2024, pp. 7794–7800

    2024

  16. [16]

    Emerging Properties in Self- Supervised Vision Transformers,

    M. Caron, H. Touvron, I. Misra, et al., “Emerging Properties in Self- Supervised Vision Transformers,” inIEEE/CVF International Confer- ence on Computer Vision (ICCV), 2021, pp. 9630–9640

    2021

  17. [17]

    Velociraptor: Leverag- ing Visual Foundation Models for Label-Free, Risk-Aware Off-Road Navigation,

    S. Triest, M. Sivaprakasam, S. Aich, et al., “Velociraptor: Leverag- ing Visual Foundation Models for Label-Free, Risk-Aware Off-Road Navigation,” inProceedings of the 8th Conference on Robot Learning (CoRL), PMLR, 2024

    2024

  18. [18]

    ViPlanner: Visual Semantic Imperative Learning for Local Navigation,

    P. Roth, J. Nubert, F. Yang, et al., “ViPlanner: Visual Semantic Imperative Learning for Local Navigation,” inIEEE International Conference on Robotics and Automation (ICRA), 2024, pp. 5243–5249

    2024

  19. [19]

    Trajec- tron++: Multi-Agent Generative Trajectory Forecasting With Heteroge- neous Data for Control,

    T. Salzmann, B. Ivanovic, P. Chakravarty, and M. Pavone, “Trajec- tron++: Multi-Agent Generative Trajectory Forecasting With Heteroge- neous Data for Control,” inEuropean Conference on Computer Vision (ECCV), 2020, pp. 683–700

    2020

  20. [20]

    Multi-Scale V AE for Environment-Aware Long Term Trajectory Prediction,

    M. Lee, S. S. Sohn, S. Moon, et al., “Multi-Scale V AE for Environment-Aware Long Term Trajectory Prediction,” inIEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022, pp. 8110–8119

    2022

  21. [21]

    MotionDiffuser: Control- lable Multi-Agent Motion Prediction Using Diffusion,

    C. M. Jiang, A. Cornman, C. E. Park, et al., “MotionDiffuser: Control- lable Multi-Agent Motion Prediction Using Diffusion,” inIEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2023, pp. 9644–9653

    2023

  22. [22]

    Stochastic Trajectory Prediction via Mo- tion Indeterminacy Diffusion,

    T. Gu, G. Chen, J. Li, et al., “Stochastic Trajectory Prediction via Mo- tion Indeterminacy Diffusion,” inIEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022, pp. 17 092–17 101

    2022

  23. [23]

    MTG: Mapless Trajectory Gener- ator with Traversability Coverage for Outdoor Navigation,

    J. Liang, P. Gao, X. Xiao, et al., “MTG: Mapless Trajectory Gener- ator with Traversability Coverage for Outdoor Navigation,” inIEEE International Conference on Robotics and Automation (ICRA), 2024, pp. 2396–2402

    2024

  24. [24]

    Dtg : Diffusion-based trajectory generation for mapless global navigation,

    J. Liang, A. Payandeh, D. Song, X. Xiao, and D. Manocha, “Dtg : Diffusion-based trajectory generation for mapless global navigation,” in IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2024, pp. 5340–5347

    2024

  25. [25]

    team,Gpt-4 technical report, 2024

    O. team,Gpt-4 technical report, 2024. arXiv: 2303.08774[cs.CL]

    Pith/arXiv arXiv 2024

  26. [26]

    Team,Gemini: A family of highly capable multimodal models, 2025

    G. Team,Gemini: A family of highly capable multimodal models, 2025. arXiv: 2312.11805[cs.CL]

    Pith/arXiv arXiv 2025

  27. [27]

    Hierarchical Open- V ocabulary 3D Scene Graphs for Language-Grounded Robot Nav- igation,

    A. Werby, C. Huang, M. B ¨uchner, et al., “Hierarchical Open- V ocabulary 3D Scene Graphs for Language-Grounded Robot Nav- igation,” inProceedings of Robotics: Science and Systems, Delft, Netherlands, 2024

    2024

  28. [28]

    GOAT: Go to Any Thing,

    M. Chang, T. Gervet, M. Khanna, et al., “GOAT: Go to Any Thing,” inRobotics: Science and Systems (RSS), 2024

    2024

  29. [29]

    VLM-GroNav: Robot Navigation Using Physically Grounded Vision-Language Mod- els in Outdoor Environments,

    M. Elnoor, K. Weerakoon, G. Seneviratne, et al., “VLM-GroNav: Robot Navigation Using Physically Grounded Vision-Language Mod- els in Outdoor Environments,” inIEEE International Conference on Robotics and Automation (ICRA), 2025, pp. 2391–2398

    2025

  30. [30]

    Nasiriany, F

    S. Nasiriany, F. Xia, W. Yu, et al.,PIVOT: Iterative Visual Prompting Elicits Actionable Knowledge for VLMs, arXiv preprint arXiv:2402.07872, 2024. arXiv: 2402.07872[cs.RO]

    Pith/arXiv arXiv 2024

  31. [31]

    CoNVOI: Context-Aware Navigation Using Vision Language Models in Outdoor and Indoor Environments,

    A. J. Sathyamoorthy, K. Weerakoon, M. Elnoor, et al., “CoNVOI: Context-Aware Navigation Using Vision Language Models in Outdoor and Indoor Environments,” inProceedings of the 8th Conference on Robot Learning (CoRL), PMLR, 2024

    2024

  32. [32]

    Behav: Behavioral rule guided autonomy us- ing vlms for robot navigation in outdoor scenes,

    K. Weerakoon et al., “Behav: Behavioral rule guided autonomy us- ing vlms for robot navigation in outdoor scenes,” in2025 IEEE International Conference on Robotics and Automation (ICRA), 2025, pp. 7044–7051

    2025

  33. [33]

    Learning Transferable Visual Models From Natural Language Supervision,

    A. Radford, J. W. Kim, C. Hallacy, et al., “Learning Transferable Visual Models From Natural Language Supervision,” inInternational Confer- ence on Machine Learning (ICML), PMLR, 2021, pp. 8748–8763

    2021

  34. [34]

    VL-TGS: Trajectory Generation and Selection Using Vision Language Models in Mapless Outdoor Environments,

    D. Song, J. Liang, X. Xiao, and D. Manocha, “VL-TGS: Trajectory Generation and Selection Using Vision Language Models in Mapless Outdoor Environments,”IEEE Robotics and Automation Letters, 2025

    2025

  35. [35]

    Liang, K

    J. Liang, K. Weerakoon, D. Song, et al.,MOSU: Autonomous Long- Range Robot Navigation with Multi-Modal Scene Understanding, arXiv preprint arXiv:2507.04686, 2025. arXiv: 2507.04686[cs.RO]

    Pith/arXiv arXiv 2025

  36. [36]

    PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation,

    C. R. Qi, H. Su, K. Mo, et al., “PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation,” inIEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2017, pp. 77–85

    2017

  37. [37]

    Importance weighted autoencoders,

    Y . Burda, R. B. Grosse, and R. Salakhutdinov, “Importance weighted autoencoders,” inConference on Learning Representations (ICLR), Y . Bengio and Y . LeCun, Eds., 2016

    2016

  38. [38]

    Robot-Centric Elevation Mapping with Uncertainty Estimates,

    P. Fankhauser, M. Bloesch, C. Gehring, et al., “Robot-Centric Elevation Mapping with Uncertainty Estimates,” inInternational Conference on Climbing and Walking Robots (CLA WAR), 2014, pp. 433–440

    2014

  39. [39]

    Image segmentation using text and image prompts,

    T. L ¨uddecke and A. Ecker, “Image segmentation using text and image prompts,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022, pp. 7086–7096

    2022

  40. [40]

    Leiva and J

    F. Leiva and J. Ruiz-del-Solar,Combining RL and IL Using a Dy- namic, Performance-Based Modulation over Learning Signals and Its Application to Local Planning, arXiv preprint arXiv:2405.09760, 2024. arXiv: 2405.09760[cs.RO]

    Pith/arXiv arXiv 2024

This paper was first reviewed by deepseek-v4-flash on August 3, 2026.