Pith. sign in

REVIEW 3 major objections 4 minor 1 cited by

Self-supervised cost of transport estimation for multimodal path planning

T0 review · 3 major / 4 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read This paper claims that a self-supervised RGB-D pipeline can estimate terrain cost of transport well enough to plan energy-efficient paths on the M4 robot.

desk verdict Central claim unverified due to circular validation; the system is plausible and worth revising with independent energy measurements. read the letter →

arxiv 2412.06101 v1 pith:OU57CAL5 submitted 2024-12-08 cs.RO cs.CVcs.LG

Vincent Gherold , Ioannis Mandralis , Eric Sihite , Adarsh Salagame , Alireza Ramezani , Morteza Gharib This is my paper · ORCID
classification cs.ROcs.CVcs.LG
keywords costoftransportself-supervisedlearningtraversabilityestimationRGB-Dperceptionbird'seyeviewmappingpathplanningmultimodalrobotenergyefficiency
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

The paper tries to establish that a robot can learn, from its own past driving alone, how much energy different terrains will cost to cross, with no human-labeled examples. It presents a self-supervised pipeline that turns RGB-D images into pixel-wise estimates of the cost of transport, fuses them into a global bird's-eye-view map, and feeds that map to an A* planner on the M4, a robot that can drive, fly, walk, or crawl. The authors report that the system assigns different costs to grass and smooth road and can pick a longer route that is nevertheless cheaper in total energy. If correct, this gives multimodal robots a practical way to choose routes, and eventually locomotion modes, from onboard vision and power sensing alone.

What carries the argument

The central object is the cost of transport, $COT=E/(mgd)=P/(mgv)$, a dimensionless ratio of energy consumed to weight times distance traveled. The argument is carried by a label-generation chain: the robot's logged trajectory and power draw are converted into a mesh with COT values, projected back into camera frames; a segmentation model extends the trajectory labels to the full visible terrain patch; and a reconstruction-confidence autoencoder labels the remaining unknown regions as non-traversable. Training uses a mean-absolute-error (MAE) loss, chosen for robustness to label noise, after which the network's pixel-wise COT predictions are projected into a local bird's-eye-view map and merged into a global one by a heuristic that keeps the most recent close-range observations. This global map is the input to an A* planner, which chooses paths by total accumulated COT rather than distance alone.

What would settle it

Take the trained model to a new site and measure the actual electrical power the robot draws while crossing each terrain class at constant speed; if the measured COT ordering across terrains, or the predicted per-pixel values, disagrees with the model's outputs beyond the reported error, the central claim is falsified.

Watch

Extended reading notes

Core claim

The central claim is that a continuous, physically meaningful traversability metric—the dimensionless cost of transport, $COT=E/(mgd)=P/(mgv)$—can be learned end to end from self-generated labels. Labels are produced by projecting the robot's measured power draw along its SLAM trajectory into camera keyframes, extending those labels to whole terrain segments, and using a reconstruction-confidence autoencoder to mark remaining unknowns as non-traversable. A regression network then predicts a pixel-wise COT image from RGB-D input, and a heuristic merger projects these predictions into a global bird's-eye-view map for planning. On the paper's test set, the best architecture reaches a pixel-wise MSE of $0.0252$ and the full pipeline runs at about 4 fps on an embedded GPU, which the authors take as evidence that the approach is both accurate enough and fast enough for real-world energy-aware navigation.

Load-bearing premise

The quantitative check assumes that labeling every pixel of a terrain type with the average cost of transport measured for that terrain over the whole dataset is a valid ground truth, so the reported accuracy would not hold if cost varies within a terrain or the collected routes bias the average.

Editorial extensions

If this is right

  • A robot can build and refresh an energy-cost map of a new environment during normal operation, with no human labeling effort.
  • Route planning can minimize energy rather than distance alone; the demonstrated planner chooses a longer path with lower total COT when the shorter path is more expensive.
  • The same per-terrain cost estimates can support locomotion-mode selection, such as crawling over rock and driving on pavement, which the paper lists as future work but follows directly from having a continuous cost map.
  • Because the training signal comes from the robot's own power and motion sensors, the pipeline should transfer to other robots equipped with RGB-D vision, odometry, and current sensing.
  • The reported embedded deployment at 4 fps makes onboard, real-time COT mapping feasible on current hardware.

Reading between the lines

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

  • A natural validation extension is to compare predicted per-pixel COT against direct power measurements on held-out terrain patches that were never visited during training; the paper's dataset-average ground truth does not exercise that comparison.
  • The same label-generation chain could be applied to other continuous cost signals, such as traction, vibration, or risk of slip, by replacing the physical quantity that is projected onto the trajectory mesh.
  • Because commanded speed was held constant during data collection, the method does not yet cover speed-dependent energy effects such as motor inefficiency; testing on varying commanded speeds is a natural extension that could reveal whether the learned costs are true terrain properties or artifacts of the fixed speed.
  • If the COT map is accurate, a multimodal planner could jointly optimize route and gait, choosing to fly only when ground cost exceeds flight cost; the paper states this as future work rather than a demonstrated result.
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 / 4 minor

Summary. The paper proposes a self-supervised pipeline for pixel-wise cost-of-transport (COT) estimation from RGBD images on the M4 multimodal robot. Labels are generated automatically by projecting the robot's trajectory, with power and speed measurements, into camera images; these labels are then extended using SAM-based mask propagation and an autoencoder-based confidence method. Several segmentation-style architectures are compared, with Asymformer selected for deployment on a Jetson Orin Nano. Predicted COT images are projected into a bird's-eye-view map and used with A* to plan energy-aware paths. The authors report MSE values on a test set and a field demonstration showing path selection based on aggregated COT.

Significance. The self-supervised labeling scheme and low-latency deployment are practically valuable: the method avoids manual labeling, runs at 4 fps on an embedded GPU, and integrates into a complete mapping and planning stack. If the COT estimates were independently verified, the work would make a useful contribution to energy-efficient navigation for multimodal robots. The architecture comparison and ablation study are also well structured. However, the quantitative evidence for the central claim of accurate COT assignment is compromised by the circular construction of the test ground truth, and no independent energy measurements are provided to validate the path-planning demonstration. As presented, the paper does not establish that the predicted COT values reflect true terrain energy costs.

major comments (3)
  1. [§IV-C and §IV-D, Table I] The test ground truth used for the quantitative evaluation is constructed in Section IV-C by assigning each terrain the average COT for that terrain computed over the whole dataset. These per-terrain averages are derived from the same self-supervised label-generation pipeline (trajectory-projected power measurements, SAM extension, and autoencoder confidence labeling) used to create the training labels. The MSE reported in Table I therefore measures agreement between the model and a smoothed, class-conditional summary of its own training targets, not agreement with independently measured energy costs. This cannot rule out that the model is merely predicting a per-terrain constant, and it provides no evidence about within-terrain COT variation (e.g., wet vs. dry grass, slope, or soil condition). To support the abstract's claim of accurate COT assignment, the authors must compare against held-out ground-truth COT obtained from power and speed measurements on terrain patches and routes not used for training.
  2. [§IV-F, Fig. 2] The A* path-planning demonstration reports the aggregated COT of the red and yellow paths as computed from the same model-generated COT map. No energy consumption is measured while the robot actually follows either path. Consequently, the claim that the red path is more energy-efficient is not independently confirmed; it only shows that the planner optimizes the model's own cost estimates. A direct comparison of measured energy consumption along candidate paths, or a comparison against measured per-trajectory COT on held-out routes, is necessary to validate the utility of the proposed cost map for energy-aware navigation.
  3. [§III-C and §III-D] The label-generation procedure introduces several free parameters that directly shape the training targets but receive no sensitivity analysis: the 5-meter spatial averaging horizon, the arbitrary high COT value assigned to non-traversable regions, the geometry of the rectangular non-traversable region above the robot, and the manually tuned decision boundary for the autoencoder confidence labeling. Because the evaluation is circular, these choices are not checked against any external measurement. The reported numerical COT values are therefore not anchored, making it difficult to interpret the MSE numbers or the path-planning aggregates as physically meaningful energy costs. A sensitivity study or independent calibration would be needed to demonstrate that the pipeline's outputs are robust to these choices.
minor comments (4)
  1. [References [25] and [26]] References [25] and [26] appear to cite the same paper (Schmid et al., Self-Supervised Traversability Prediction by Learning to Reconstruct Safe Terrain); this duplicate citation should be corrected.
  2. [§V, Discussion and Conclusion] The conclusion states that the model was demonstrated 'both in simulation and in real-world field tests,' but no simulation experiments are reported in the manuscript; either remove the simulation claim or add the corresponding experiments.
  3. [Eq. (2)] The text accompanying Eq. (2) says '|Si| is the number of mask in the batch,' but the summation over (x, y) in Si indicates that |Si| should be the number of pixels in mask Si; please clarify and fix the wording.
  4. [Fig. 6 caption] The caption states that 'a fixed threshold has been set to label highly reconstructed elements as non-traversable,' which is confusing because the text explains that high reconstruction error indicates non-traversability; rephrase the caption to say 'elements with reconstruction error beyond a decision boundary.'

Circularity Check

2 steps flagged · score 6.0 of 10

Quantitative evaluation is circular: §IV-C builds test ground truth from the same self-supervised labels used for training, so Table I's MSE does not independently validate COT accuracy.

  1. fitted input called prediction [Section IV-C (Qualitative Analysis) and Section IV-D (Quantitative Analysis, Table I)]
    ""We manually labeled the test dataset by assigning each terrain, the average COT for that specific terrain computed over the whole dataset.""

    The test target is a per-terrain average 'computed over the whole dataset' of COT labels generated by the same self-supervised pipeline used to produce the training labels (Section III-C/D: trajectory power data projected into images, then extended via SAM masks and confidence-based autoencoding). Section IV-A splits these labels into training, validation, and test sets. Thus the MSE in Table I measures agreement between the model and a smoothed, class-conditional summary of its own training targets, not agreement with independently measured energy costs. A model predicting a per-terrain constant could achieve low MSE by construction, and within-terrain COT variation (slope, wetness, soil condition) is not evaluated.

  2. other [Figure 2 caption and Section IV-F (Practical Deployment and Real-World Testing)]
    ""The red path represents the most efficient route, with a total aggregate COT of 817 over a distance of 59 meters. In contrast, the yellow path is a suboptimal solution with a COT of 839 and a distance of 56 meters. Despite being longer, the red path is more energy-efficient because it predominantly follows the road.""

    The aggregate COT values along the red and yellow paths are computed by summing the model's predicted COT values on the reconstructed BEV map, which is also the cost field minimized by A*. The demonstration therefore only shows that A* optimizes the model's own cost map. Calling the red path 'more energy-efficient' is a restatement of the planner's objective using the model's outputs, not a measurement of energy consumption along either path. It provides no independent confirmation that the preferred route actually consumes less energy.

full rationale

The underlying COT values originate from physical power, current, voltage, and speed measurements (Section III-C), so the label-generation pipeline is not definitionally circular with respect to energy data. However, the reported validation is circular: the test ground truth is a per-terrain average of the same self-supervised labels used to train the model, so the MSE in Table I is an internal-consistency check, not an independent test of COT accuracy. The A* demonstration similarly sums the model's own outputs. The paper does not provide held-out energy measurements on routes or terrain patches not used for training, and the Discussion acknowledges reliance on SAM mask quality. No load-bearing self-citation chain was found; previous M4 references are platform background. Score 6 reflects partial, not total, circularity: the predicted quantities are not defined solely in terms of the evaluation target, but the central accuracy claim is supported only by a metric that reduces to agreement with the training-label distribution.

Assumptions & free parameters 4 free parameters · 6 assumptions · 0 invented entities

The central claim rests on three labeling assumptions, the quality of SAM segmentation, the autoencoder reconstruction assumption, and the constant-speed condition during data collection. The free parameters are mostly manual thresholds and geometries that affect label quality.

free parameters (4)
  • Autoencoder decision boundary = Manual threshold on reconstruction error
    Manually tuned decision boundary on autoencoder reconstruction error (Section III-D, Fig. 6) used to label unlabeled pixels as non-traversable; affects 35.95% of labels.
  • COT averaging horizon = 5 m
    Length of trajectory used for spatial moving average of power when assigning COT labels (Section III-C).
  • Non-traversable COT value = High arbitrary
    Assigned to walls, vegetation, and rocks via assumption (2); value not specified (Section III-C).
  • Rectangular non-traversable region geometry = Unspecified
    Dimensions of predetermined region above the robot used to label non-traversable obstacles are not given (Section III-C).
assumptions (6)
  • domain assumption Traversed areas are traversable (Assumption 1)
    Section III-C: every point on the robot's path is labeled as traversable, which can be wrong on slippery or adverse terrain.
  • ad hoc to paper Rectangular region above robot is non-traversable (Assumption 2)
    Section III-C: used to label walls/vegetation/rocks; geometry is arbitrary and platform-specific.
  • domain assumption Similar terrain features imply similar COT (Assumption 3)
    Section III-D: basis for SAM-based label extension; can fail when visually similar terrain has different traversability.
  • domain assumption Autoencoder reconstruction error separates traversable from non-traversable
    Section III-D: adapted from [6]; assumes non-traversable regions reconstruct poorly.
  • domain assumption SAM segmentation quality is sufficient for mask extension
    Section V: 'The success of the entire work heavily relies on the quality of the SAM mask segmentation.'
  • domain assumption Commanded speed is constant during data collection
    Section V: 'During data collection, the commanded speed... was kept constant.' Makes COT proportional to power and avoids motor efficiency confounds.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Self-supervised cost of transport estimation for multimodal path planning." pith.science (2026). https://pith.science/paper/OU57CAL5

@misc{pith2026241206101,
  author       = {Pith},
  title        = {Pith review of: Self-supervised cost of transport estimation for multimodal path planning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OU57CAL5}},
  note         = {Machine review of arXiv:2412.06101}
}
read the original abstract

Autonomous robots operating in real environments are often faced with decisions on how best to navigate their surroundings. In this work, we address a particular instance of this problem: how can a robot autonomously decide on the energetically optimal path to follow given a high-level objective and information about the surroundings? To tackle this problem we developed a self-supervised learning method that allows the robot to estimate the cost of transport of its surroundings using only vision inputs. We apply our method to the multi-modal mobility morphobot (M4), a robot that can drive, fly, segway, and crawl through its environment. By deploying our system in the real world, we show that our method accurately assigns different cost of transports to various types of environments e.g. grass vs smooth road. We also highlight the low computational cost of our method, which is deployed on an Nvidia Jetson Orin Nano robotic compute unit. We believe that this work will allow multi-modal robotic platforms to unlock their full potential for navigation and exploration tasks.

Figures

Figures reproduced from arXiv: 2412.06101 by the authors.

Figure 1
Figure 1. Close-up view of our robot platform, the M4 robot, capable of multiple modes of locomotion, including driving, flying, and walking. The robot is equipped with an RGB-D camera along with an embedded companion computer. [33], [16]. However, these methods require manual feature engineering and are usually designed for specific terrain classes and may not be easily generalized [1]. Supervised learning based methods have… view at source ↗
Figure 2
Figure 2. Results of our self-supervised COT. The left image shows the terrain that the robot will encounter, composed of road, grass, and vegetation. The middle image displays the trajectory taken by the robot, while the right image presents the resulting COT map, colored using a colormap ranging from 0.5 to 2. On this COT map, the A* algorithm has been applied to illustrate how COT influences the optimal path to minimize en… view at source ↗
Figure 3
Figure 3. Our RGBD COT model takes as inputs an RGB image and depth image and outputs a pixel-wise COT image. RGB and Depth are of size 3 × H × W and 1 × H × W respectively. Then the pixel-wise COT image is projected in a local BEV map using the robot position and the depth image. A heuristic map merger combines all the local BEV maps into a global map. at inference time are RGBD images and their positions in the 3D world, th… view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: The left image displays the camera’s view, while the right one shows the rendered world. The elements have been colored for visualization purposes. The triangle mesh COT is represented in black, the point cloud in red, and the non-traversable elements from assumption (…
Figure 6
Figure 6. Figure 6: shows the distribution of mask loss SE(Si) for a batch of images after the model have been trained on the dataset. As expected the labeled samples are correctly reconstructed. For the unlabeled samples, we observe a bimodal distribution, indicating that the traversable…
Figure 7
Figure 7. Figure 7: Inference on four images of the test dataset. The first two columns are the input given to the model. The third column represents the ground truth manually labeled. The last four columns show the COT images from the models considered. These COT images have been colored…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Analysis of Harpy's Constrained Trotting and Jumping Maneuver

    cs.RO 2025-08 unverdicted novelty 4.0 of 10

    The provided manuscript text does not contain the claimed analysis of the Harpy robot, making the abstract's conclusions unverifiable from this document.

Reference graph

Works this paper leans on

39 extracted references · 21 canonical work pages · cited by 1 Pith paper

  1. [1]

    A Survey on Terrain Traversability Analysis for Autonomous Ground Vehicles: Methods, Sensors, and Challenges

    Paulo Borges et al. “A Survey on Terrain Traversability Analysis for Autonomous Ground Vehicles: Methods, Sensors, and Challenges”. In: Field Robotics 2.1 (Mar. 2022), 1567–1627. ISSN : 2771-3989. DOI: 10.55417/fr.2022049 . URL: http://dx.doi.org/10. 55417/fr.2022049

    doi:10.55417/fr.2022049 2022

  2. [2]

    Emerging Properties in Self-Supervised Vision Transformers

    Mathilde Caron et al. “Emerging Properties in Self-Supervised Vision Transformers”. In: Proceedings of the IEEE/CVF International Con- ference on Computer Vision (ICCV) . 2021, pp. 9650–9660

    work page 2021

  3. [3]

    How Does It Feel? Self-Supervised Costmap Learning for Off-Road Vehicle Traversability

    Mateo Guaman Castro et al. “How Does It Feel? Self-Supervised Costmap Learning for Off-Road Vehicle Traversability”. In:2023 IEEE International Conference on Robotics and Automation (ICRA) . IEEE, May 2023. DOI: 10.1109/icra48891.2023.10160856 . URL: http : / / dx . doi . org / 10 . 1109 / ICRA48891 . 2023 . 10160856

    arXiv 2023

  4. [4]

    ImageNet: A large-scale hierarchical image database

    Jia Deng et al. “ImageNet: A large-scale hierarchical image database”. In: 2009 IEEE Conference on Computer Vision and Pattern Recogni- tion. 2009, pp. 248–255. DOI: 10.1109/CVPR.2009.5206848

    arXiv 2009

  5. [5]

    AsymFormer: Asymmetrical Cross-Modal Representation Learning for Mobile Platform Real-Time RGB-D Semantic Segmentation

    Siqi Du et al. AsymF ormer: Asymmetrical Cross-Modal Representation Learning for Mobile Platform Real-Time RGB-D Semantic Segmen- tation. 2024. arXiv: 2309 . 14065 [cs.CV] . URL: https : / / arxiv.org/abs/2309.14065

    work page Pith review arXiv 2024

  6. [6]

    Fast Traversability Estimation for Wild Visual Navi- gation

    Jonas Frey et al. Fast Traversability Estimation for Wild Visual Navi- gation. 2023. arXiv: 2305.08510

    arXiv 2023

  7. [7]

    RoadRunner - Learning Traversability Estimation for Autonomous Off-road Driving

    Jonas Frey et al. RoadRunner - Learning Traversability Estimation for Autonomous Off-road Driving . 2024. arXiv: 2402.19341

    arXiv 2024

  8. [8]

    What price speed ? Specific power requirede for propulsion of vehicles

    G. Gabrielli and Theodore von K ´arm´an. “What price speed ? Specific power requirede for propulsion of vehicles”. In: Mech Eng. 72 (1950), pp. 775–781

    work page 1950

Show all 39 references
  1. [9]

    WayFAST: Navigation With Predictive Traversability in the Field

    Mateus V . Gasparino et al. “WayFAST: Navigation With Predictive Traversability in the Field”. In: IEEE Robotics and Automation Letters 7.4 (Oct. 2022), 10651–10658. ISSN : 2377-3774. DOI: 10 . 1109 / lra.2022.3193464. URL: http://dx.doi.org/10.1109/ LRA.2022.3193464

    2022

  2. [10]

    WayF ASTER: a Self-Supervised Traversability Prediction for Increased Navigation Awareness

    Mateus Valverde Gasparino, Arun Narenthiran Sivakumar, and Girish Chowdhary. WayF ASTER: a Self-Supervised Traversability Prediction for Increased Navigation Awareness. 2024. DOI: 10.48550/ARXIV. 2402.00683. URL: https://arxiv.org/abs/2402.00683

    2024 doi arXiv

  3. [11]

    Simple Copy-Paste is a Strong Data Aug- mentation Method for Instance Segmentation

    Golnaz Ghiasi et al. “Simple Copy-Paste is a Strong Data Aug- mentation Method for Instance Segmentation”. In: 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) . 2021, pp. 2917–2927. DOI: 10.1109/CVPR46437.2021.00294

    2021

  4. [12]

    TNS: Terrain Traversability Mapping and Naviga- tion System for Autonomous Excavators

    Tianrui Guan et al. “TNS: Terrain Traversability Mapping and Naviga- tion System for Autonomous Excavators”. In: Robotics: Science and Systems XVIII . RSS2022. Robotics: Science and Systems Foundation, June 2022. DOI: 10.15607/rss.2022.xviii.049. URL: http: //dx.doi.org/10.156...

    2022 doi

  5. [13]

    V-STRONG: Visual Self-Supervised Traversability Learning for Off-road Navigation

    Sanghun Jung et al. V-STRONG: Visual Self-Supervised Traversability Learning for Off-road Navigation . 2023. DOI: 10.48550/ARXIV. 2312.16016. URL: https://arxiv.org/abs/2312.16016

    2023 doi arXiv

  6. [14]

    RViz: a toolkit for real domain data visu- alization

    Hyeong Ryeol Kam et al. “RViz: a toolkit for real domain data visu- alization”. In: Telecommunication Systems 60.2 (Apr. 2015), 337–345. ISSN : 1572-9451. DOI: 10.1007/s11235- 015- 0034- 5 . URL: http://dx.doi.org/10.1007/s11235-015-0034-5

    2015 doi

  7. [15]

    Supervised Contrastive Learning

    Prannay Khosla et al. “Supervised Contrastive Learning”. In: Advances in Neural Information Processing Systems . Ed. by H. Larochelle et al. V ol. 33. Curran Associates, Inc., 2020, pp. 18661–18673. URL: https : / / proceedings . neurips . cc / paper _ files / paper / 2020 / f...

    2020

  8. [16]

    Traversability classification for UGV navigation: a comparison of patch and super- pixel representations

    Dongshin Kim, Sang Min Oh, and James M. Rehg. “Traversability classification for UGV navigation: a comparison of patch and super- pixel representations”. In: 2007 IEEE/RSJ International Conference on Intelligent Robots and Systems . IEEE, Oct. 2007. DOI: 10.1109/ iros.2007.439...

    2007

  9. [17]

    Segment Anything

    Alexander Kirillov et al. “Segment Anything”. In: Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) . 2023, pp. 4015–4026

    2023

  10. [18]

    RTAB-Map as an open- source lidar and visual simultaneous localization and mapping library for large-scale and long-term online operation

    Mathieu Labb ´e and Franc ¸ois Michaud. “RTAB-Map as an open- source lidar and visual simultaneous localization and mapping library for large-scale and long-term online operation”. In: Journal of Field Robotics 36.2 (Oct. 2018), 416–446. ISSN : 1556-4967. DOI: 10 . 1002/rob.21...

    2018

  11. [19]

    Learning robust autonomous navigation and locomo- tion for wheeled-legged robots

    Joonho Lee et al. “Learning robust autonomous navigation and locomo- tion for wheeled-legged robots”. In: Science Robotics 9.89 (Apr. 2024). ISSN : 2470-9476. DOI: 10.1126/scirobotics.adi9641. URL: http://dx.doi.org/10.1126/scirobotics.adi9641

    2024 doi

  12. [20]

    Minimum Time Trajectory Generation for Bounding Flight: Combining Posture Control and Thrust Vectoring

    Ioannis Mandralis et al. “Minimum Time Trajectory Generation for Bounding Flight: Combining Posture Control and Thrust Vectoring”. In: 2023 European Control Conference (ECC) . IEEE, June 2023. DOI: 10.23919/ecc57647.2023.10178360 . URL: http://dx. doi.org/10.23919/ECC57647.202...

    2023

  13. [21]

    Real-time Semantic Mapping for Autonomous Off-Road Navigation

    Daniel Maturana et al. “Real-time Semantic Mapping for Autonomous Off-Road Navigation”. In: Proceedings of 11th International Confer- ence on Field and Service Robotics (FSR ’17) . 2017, pp. 335 –350

    2017

  14. [22]

    PyTorch: An Imperative Style, High-Performance Deep Learning Library

    Adam Paszke et al. “PyTorch: An Imperative Style, High-Performance Deep Learning Library”. In: Advances in Neural Information Pro- cessing Systems . Ed. by H. Wallach et al. V ol. 32. Curran Associates, Inc., 2019. URL: https : / / proceedings . neurips . cc / paper _ files / ...

    2019

  15. [23]

    ROS: an open-source Robot Operating System

    Morgan Quigley et al. “ROS: an open-source Robot Operating System”. en. In: (), p. 6

  16. [24]

    U-Net: Con- volutional Networks for Biomedical Image Segmentation

    Olaf Ronneberger, Philipp Fischer, and Thomas Brox. “U-Net: Con- volutional Networks for Biomedical Image Segmentation”. In: Med- ical Image Computing and Computer-Assisted Intervention – MIC- CAI 2015 . Springer International Publishing, 2015, 234–241. ISBN : 9783319245744. D...

    2015 doi

  17. [26]

    Self-Supervised Traversability Prediction by Learning to Reconstruct Safe Terrain

    Robin Schmid et al. “Self-Supervised Traversability Prediction by Learning to Reconstruct Safe Terrain”. In: 2022 IEEE/RSJ Interna- tional Conference on Intelligent Robots and Systems (IROS). IEEE, Oct

    2022

  18. [27]

    Learning Off-Road Terrain Traversability With Self-Supervisions Only

    Junwon Seo, Sungdae Sim, and Inwook Shim. “Learning Off-Road Terrain Traversability With Self-Supervisions Only”. In:IEEE Robotics and Automation Letters 8.8 (Aug. 2023), 4617–4624. ISSN : 2377-3774. DOI: 10.1109/lra.2023.3284356 . URL: http://dx.doi. org/10.1109/LRA.2023.3284356

    2023

  19. [28]

    URL: http: //dx.doi.org/10.1109/IROS47612.2022.9981368

    DOI: 10.1109/iros47612.2022.9981368. URL: http: //dx.doi.org/10.1109/IROS47612.2022.9981368

    2022

  20. [29]

    ScaTE: A Scalable Framework for Self- Supervised Traversability Estimation in Unstructured Environments

    Junwon Seo et al. “ScaTE: A Scalable Framework for Self- Supervised Traversability Estimation in Unstructured Environments”. In: IEEE Robotics and Automation Letters 8.2 (Feb. 2023), 888–895. ISSN : 2377-

    2023

  21. [30]

    META V erse: Meta-Learning Traversability Cost Map for Off-Road Navigation

    Junwon Seo et al. META V erse: Meta-Learning Traversability Cost Map for Off-Road Navigation . 2023. DOI: 10 . 48550 / ARXIV . 2307 . 13991. URL: https://arxiv.org/abs/2307.13991

    2023 doi arXiv

  22. [31]

    Demonstrating Autonomous 3D Path Planning on a Novel Scalable UGV-UA V Morphing Robot

    Eric Sihite et al. “Demonstrating Autonomous 3D Path Planning on a Novel Scalable UGV-UA V Morphing Robot”. In: 2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) . IEEE, Oct. 2023. DOI: 10.1109/iros55552.2023.10342189 . URL: http : / / dx . doi . ...

    2023

  23. [32]

    Multi-Modal Mobility Morphobot (M4) with ap- pendage repurposing for locomotion plasticity enhancement

    Eric Sihite et al. “Multi-Modal Mobility Morphobot (M4) with ap- pendage repurposing for locomotion plasticity enhancement”. In: Na- ture Communications 14.1 (June 2023). ISSN : 2041-1723. DOI: 10. 1038/s41467- 023- 39018- y. URL: http://dx.doi.org/ 10.1038/s41467-023-39018-y

    2023 doi

  24. [33]

    A Survey of Traversability Estimation for Mobile Robots

    Christos Sevastopoulos and Stasinos Konstantopoulos. “A Survey of Traversability Estimation for Mobile Robots”. In: IEEE Access 10 (2022), pp. 96331–96347. DOI: 10 . 1109 / ACCESS . 2022 . 3202545

    2022

  25. [34]

    D-former: a U-shaped Dilated Transformer for 3D medical image segmentation

    Yixuan Wu et al. “D-former: a U-shaped Dilated Transformer for 3D medical image segmentation”. In: Neural Computing and Applications 35.2 (Oct. 2022), 1931–1944. ISSN : 1433-3058. DOI: 10 . 1007 / s00521 - 022 - 07859 - 1. URL: http : / / dx . doi . org / 10 . 1007/s00521-022-07859-1

    2022

  26. [35]

    SegFormer: Simple and Efficient Design for Se- mantic Segmentation with Transformers

    Enze Xie et al. “SegFormer: Simple and Efficient Design for Se- mantic Segmentation with Transformers”. In: Advances in Neural Information Processing Systems . Ed. by M. Ranzato et al. V ol. 34. Curran Associates, Inc., 2021, pp. 12077–12090. URL: https : / / proceedings.neuri...

    2021

  27. [36]

    Probabilistic traversability map generation using 3D- LIDAR and camera

    Juil Sock et al. “Probabilistic traversability map generation using 3D- LIDAR and camera”. In: 2016 IEEE International Conference on Robotics and Automation (ICRA) . 2016, pp. 5631–5637. DOI: 10 . 1109/ICRA.2016.7487782

    2016

  28. [37]

    CMX: Cross-Modal Fusion for RGB-X Semantic Segmentation With Transformers

    Jiaming Zhang et al. “CMX: Cross-Modal Fusion for RGB-X Semantic Segmentation With Transformers”. In: IEEE Transactions on Intelligent Transportation Systems 24.12 (2023), pp. 14679–14694. DOI: 10 . 1109/TITS.2023.3300537

    2023

  29. [38]

    Open3D: A Modern Library for 3D Data Processing

    Qian-Yi Zhou, Jaesik Park, and Vladlen Koltun. “Open3D: A Modern Library for 3D Data Processing”. In: arXiv:1801.09847 (2018)

    2018 arXiv

  30. [39]

    Positive- Unlabeled Learning in the Face of Labeling Bias

    Noah Youngs, Dennis Shasha, and Richard Bonneau. “Positive- Unlabeled Learning in the Face of Labeling Bias”. In: 2015 IEEE International Conference on Data Mining Workshop (ICDMW) . IEEE, Nov. 2015. DOI: 10 . 1109 / icdmw . 2015 . 207. URL: http : //dx.doi.org/10.1109/ICDMW.2015.207

    2015 doi

  31. [3774]

    URL: http://dx

    DOI: 10.1109/lra.2023.3234768 . URL: http://dx. doi.org/10.1109/LRA.2023.3234768

    2023

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.