Pith. sign in

REVIEW 3 major objections 5 minor 34 references

Planning Beyond the Sensing Horizon Using a Learned Context

T0 review · 3 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read A robot can find an unseen front door by learning to read driveway and walkway context from a partial semantic map.

desk verdict A clean, useful reformulation of context-aware exploration as image-to-image translation; the headline number is plausible but rests on unstated training-mask details and a single baseline. read the letter →

arxiv 1908.09171 v3 pith:YK2D4H76 submitted 2019-08-24 cs.RO cs.AIcs.LG

Michael Everett , Justin Miller , Jonathan P. How This is my paper · ORCID
classification cs.ROcs.AIcs.LG
keywords last-miledeliveryexplorationplanningcost-to-goestimationimage-to-imagetranslationsemanticgridmapfrontier-basedU-Netcontext-basednavigation
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 tries to establish that a robot can search efficiently for a goal it cannot see, such as a house's front door, by learning to read terrain context from a partially built map. The authors frame context use as an image-to-image translation problem: a U-Net takes the robot's current semantic gridmap and outputs an estimated cost-to-go image, a per-pixel value of how close each reachable point is to the unknown goal. This estimated cost-to-go then selects which frontier cell an exploration planner should head toward next. The claim is that this makes exploration goal-directed without goal coordinates, and the reported evidence is that across 42 real house layouts the planner reaches the goal within 63% of the prior-map optimal path and 189% faster than context-unaware frontier exploration. A symmetric benefit is that the network trains on a static dataset built from annotated satellite images, not on a simulator.

What carries the argument

The load-bearing object is the trained image-to-image translator: a U-Net with encoder-decoder layers and skip connections, trained with L1, GAN, or combined losses, that maps a 256x256 RGB semantic gridmap to a 256x256 RGB estimated cost-to-go image. Ground-truth labels come from Dijkstra's algorithm run on annotated satellite-view semantic maps, with element-wise observation masks applied to mimic partial observability, so the network learns the most likely cost-to-go given what has been seen so far. In online planning, Algorithm 2 filters the network output to traversable cells, considers only reachable cells that expand frontiers within the sensor field of view, and greedily selects the one with the highest estimated value; if the goal becomes reachable, it switches to a BFS shortest path. This machinery converts contextual cues such as a driveway connected to a road and a walkway leading to a door into a quantity a planner can optimize without knowing the goal coordinates.

What would settle it

Take the trained DC2G network to a neighborhood of houses with no driveways, with front doors opening directly onto the sidewalk or a pedestrian courtyard, and measure time-to-goal against Frontier; if the learned cost-to-go does not beat pure exploration there, the terrain-context assumption fails. Alternatively, build a test set of real partial maps from a forward-facing camera and compare planner performance to the same layouts' masked training maps; a large drop would show the synthetic masks do not capture online observability.

Watch

Extended reading notes

Core claim

The paper's central claim is that the abstract notion of context for planning can be made operational as a translation between two images: a partial semantic map (the robot's current belief) and a cost-to-go map (a per-cell estimate of the shortest distance to the high-level goal). DC2G, Deep Cost-to-Go, trains a U-Net generator offline on pairs produced by computing exact cost-to-go with Dijkstra on annotated satellite maps and masking them with simulated observation patterns, then uses the network's output online to rank frontier-expanding cells for a frontier-based explorer. Because the output is a full cost-to-go image rather than a single predicted goal location, it can serve as a receding-horizon heuristic and be combined with local collision avoidance. The paper reports that this scheme reaches the goal within 63% of optimal on 42 held-out houses and is 189% faster than pure frontier exploration, and demonstrates the full pipeline on a vehicle with a forward-facing RGB-D and semantic camera in a high-fidelity Unreal simulation.

Load-bearing premise

The method assumes that the artificially masked maps used in training look like the partial maps a real robot builds from its camera and SLAM system, and that front doors in new neighborhoods sit in the same driveway-and-walkway context as the training houses.

Editorial extensions

If this is right

  • A robot with no prior map can use learned terrain context to bias exploration toward a semantic goal, so delivery destinations can be specified as high-level concepts like "front door" instead of coordinates.
  • The approach transfers from static satellite-derived training data to a forward-facing camera pipeline in simulation, suggesting the same context signal survives realistic noisy semantic maps.
  • Because the learned output is a cost-to-go image rather than an action, the planner can be paired with standard local collision avoidance and can report failure if no frontier remains, which end-to-end policies cannot do.
  • The method degrades gracefully: if context misleads, the robot still eventually explores all frontiers, whereas an end-to-end learned policy could get stuck.
  • The planner generalizes to houses and neighborhoods not seen in training, including an urban neighborhood, though performance drops when layouts differ from training.

Reading between the lines

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

  • The same cost-to-go translation frame should apply to other structured outdoor goals, such as mailboxes, loading docks, or parking entrances, as long as their terrain adjacency statistics are stable and can be labeled in overhead imagery; this is a testable extension the paper does not run.
  • The real generalization condition hidden in the method is that front-door context is mostly terrain adjacency, so neighborhoods that violate that adjacency, such as alley-loaded houses or courtyard entrances, would require retraining or fallback to pure exploration.
  • A practical improvement suggested by the paper's own similarity analysis would be to weigh the learned heuristic by its estimated reliability: when the current map is far from the training distribution, decay toward frontier exploration, bounding worst-case performance.
  • Satellite imagery plus manual semantic annotation may be a general low-cost substitute for simulator interaction in other outdoor robot learning tasks, since the paper shows the learned representation transfers across data formats.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

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 DC2G, a planner for last-mile delivery to an unknown goal (a house's front door) using only a partial semantic gridmap built online. Context is captured by training a U-Net/GAN image-to-image translator offline on pairs of masked partial semantic maps and masked Dijkstra cost-to-go maps derived from satellite imagery of real houses. At run time, the planner selects the reachable frontier cell with the highest estimated cost-to-go, then executes a shortest path to that cell. The main evaluation is a gridworld simulation over 42 test houses from Bing Maps, reporting that DC2G reaches the goal within 63% of optimal and 189% faster than frontier-based exploration; a qualitative Unreal Engine demonstration with a forward-facing camera is also included. The paper is clearly written, releases source code, and frames the training as using a static dataset rather than a simulator.

Significance. If the empirical claim holds, the paper offers a practical way to inject terrain context into exploration without pre-mapping, and the image-to-image formulation is a clean reduction of the context-utilization problem. The cost-to-go representation is more informative for planning than a single goal-point estimate, and the evaluation across 42 real house layouts is more extensive than many single-scene robotics demos. The open-source release and the use of a satellite-derived static dataset are concrete strengths, and the authors are candid about the task-specific nature of context. However, the central quantitative result depends on an unstated match between the training-mask distribution and the online observation process, and the headline averages are reported without variance or significance tests. The Unreal transfer demonstration is qualitative, so the deployment claim is supported only weakly. The circularity burden is low because the network is trained on Dijkstra ground truth and evaluated on held-out houses with no parameters fitted to the final planner result.

major comments (3)
  1. [Section III-A, Algorithm 1; Section IV-B] The paper never specifies how the 256 observation masks M are generated, although this is the only mechanism that aligns the offline training distribution with the partial maps the planner sees online. In Section IV-B the online maps are accumulated from a 90-degree, 8-cell field of view along the robot's actual trajectory, so the training masks should be visibility wedges anchored at feasible robot poses with matching range and FOV. If the masks are instead sampled as arbitrary contiguous occlusions, the network can be well calibrated on the training masks but miscalibrated on the planner's true inputs, which would make the frontier ranking in Algorithm 2 unreliable even if per-pixel L1 error on full maps is low. Please state the exact mask-generation procedure and add a quantitative comparison, e.g., a distribution-distance measure or an ablation in which training masks are drawn from the online FOV process.
  2. [Section IV-B, Fig. 7] The headline result 'DC2G reaches the goal within 63% of optimal, and 189% faster than Frontier on average' is reported as a single average over 42 houses with no error bars, confidence intervals, or significance test. The text states that random starts are selected 100 times per neighborhood, so the data contain enough trials for a paired statistical comparison. Please report per-house and aggregate distributions with uncertainty, and perform a paired test (e.g., Wilcoxon signed-rank) against Frontier. Without this, the 189% figure cannot be distinguished from sampling noise, especially because Fig. 7 groups results by neighborhood without showing variability.
  3. [Section III-A] The dataset counts are internally inconsistent. With 31 train, 4 validation, and 42 test houses and 256 masks per house, the expected number of pairs is 7936, 1024, and 10752, respectively, yet the text reports 7936 train, 320 validation, and 615 test pairs. The test-pair count is not an integer multiple of 42 (615 = 41 × 15), so either the number of test houses, the number of masks, or the pair counts are misreported. This makes the exact composition of the evaluation set unclear and should be corrected.
minor comments (5)
  1. [Section IV-A.1] There is a typo, 'acheives', in the discussion of the GAN loss.
  2. [Abstract and Section IV-B] The phrase 'within 63% of optimal' is ambiguous because the metric defined in Section IV-B is percent extra time beyond the oracle path, (%te_goal). Recommend rewording to something like 'on average 63% longer than the optimal path'.
  3. [Section IV-D] The Unreal Engine demonstration is qualitative; reporting a quantitative success metric, such as completion rate and path-length ratio relative to an oracle, would strengthen the claim that the approach transfers to camera-built semantic maps.
  4. [Section IV-E] The statement that Frontier outperformed DC2G on 2/77 houses should be reconciled with the test-set claim of 42 houses; if the statistic includes training or validation houses, that should be stated explicitly.
  5. [Section III-C] The online mapping module is cited to a GitHub repository [32]; a peer-reviewed citation for semantic SLAM or a more detailed description of the projection would improve reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the learned cost-to-go is trained on independently computed Dijkstra targets and evaluated on held-out houses; no final planner result is fitted.

full rationale

The paper's derivation chain is self-contained and empirically testable. Training data are produced by Algorithm 1: a manually annotated satellite-derived semantic map is separated into traversable and non-traversable regions, the ground-truth cost-to-go is computed by Dijkstra's algorithm from the goal, and 256 observation masks are applied to create partial-map training pairs (Section III-A). The network is then trained offline to translate the masked semantic map into the masked cost-to-go image (Section III-B). At planning time, the DC2G planner feeds the robot's current partial semantic map into the fixed network and uses the estimated cost-to-go to rank frontier-expanding cells (Algorithm 2). The headline result in Section IV-B is an average over 42 test houses whose layouts and masks were not used in training, measured against a frontier-exploration baseline and an oracle with a prior map. No parameter is fitted to the final planner result, and no claimed prediction is equivalent by construction to the training objective. The only substantive concern is distributional: whether the Algorithm 1 masks truly match the partial maps produced online by the forward-facing camera and semantic SLAM pipeline. That is a generalization risk about unseen sensor distributions, not a circularity, because the training targets and evaluation metric are independently defined. Self-citations are not load-bearing: reference [22] motivates U-Nets for context extraction but the planner's success is not derived from it. Therefore the paper warrants a circularity score of 0.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The ledger shows the central result is an empirical ML claim. The learned network weights are the main fitted content, but they are fit on training houses and evaluated on held-out houses, so the 189% figure is not a fit to the test result. The load-bearing assumptions are about transfer from clean annotated maps to noisy online maps and about task-specific context, both acknowledged in the discussion.

free parameters (3)
  • U-Net encoder-decoder weights = Learned on 7,936 training image pairs (not enumerated)
    The planner's cost-to-go estimates and the 189% result depend on these trained weights. They are fitted to the training set, not to the test houses, so they are legitimate learned parameters rather than test-set fitting.
  • Observation masks per training map = 256
    Chosen by hand in Section III-A to simulate partial observability. The distribution of masked maps shapes what the network learns; changing this number could change planner performance.
  • Gridworld sensor FOV and range = 90 degrees horizontal, 8-cell radial range
    Chosen to approximate a common RGB-D sensor in the low-fidelity planner evaluation (Section IV-B). The planner's feasible frontiers and results depend on this sensing model.
assumptions (5)
  • domain assumption Traversability is determined by terrain class in the semantic map, with roads and driveways traversable and grass and houses not.
    Used in Algorithm 1 and Algorithm 2 to compute cost-to-go and reachable sets. If terrain classes do not track real drivability, the learned cost-to-go and the planner's actions are wrong.
  • domain assumption The 256 synthetic observation masks per map produce training pairs representative of partial maps the robot will see online.
    Section III-A generates training data by masking full maps. Section IV-D validates transfer only qualitatively in one Unreal scenario.
  • domain assumption Online semantic SLAM maps from a forward-facing RGB-D camera are close enough to the clean training maps for the trained network to transfer.
    Section IV-D demonstrates the pipeline in Unreal, but reports no quantitative planning performance on these noisier maps.
  • standard math Dijkstra's algorithm on the traversable grid gives the true cost-to-go to the goal.
    Algorithm 1 Line 5 uses Dijkstra to create ground-truth training targets. This is standard and unproblematic.
  • domain assumption Front doors have consistent contextual structure, such as driveway and walkway adjacency to roads, across the 42 test houses.
    The network learns contextual clues from annotated maps; the authors acknowledge context is task-specific in Section IV-E.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Planning Beyond the Sensing Horizon Using a Learned Context." pith.science (2026). https://pith.science/paper/YK2D4H76

@misc{pith2026190809171,
  author       = {Pith},
  title        = {Pith review of: Planning Beyond the Sensing Horizon Using a Learned Context},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YK2D4H76}},
  note         = {Machine review of arXiv:1908.09171}
}
read the original abstract

Last-mile delivery systems commonly propose the use of autonomous robotic vehicles to increase scalability and efficiency. The economic inefficiency of collecting accurate prior maps for navigation motivates the use of planning algorithms that operate in unmapped environments. However, these algorithms typically waste time exploring regions that are unlikely to contain the delivery destination. Context is key information about structured environments that could guide exploration toward the unknown goal location, but the abstract idea is difficult to quantify for use in a planning algorithm. Some approaches specifically consider contextual relationships between objects, but would perform poorly in object-sparse environments like outdoors. Recent deep learning-based approaches consider context too generally, making training/transferability difficult. Therefore, this work proposes a novel formulation of utilizing context for planning as an image-to-image translation problem, which is shown to extract terrain context from semantic gridmaps, into a metric that an exploration-based planner can use. The proposed framework has the benefit of training on a static dataset instead of requiring a time-consuming simulator. Across 42 test houses with layouts from satellite images, the trained algorithm enables a robot to reach its goal 189\% faster than with a context-unaware planner, and within 63\% of the optimal path computed with a prior map. The proposed algorithm is also implemented on a vehicle with a forward-facing camera in a high-fidelity, Unreal simulation of neighborhood houses.

Figures

Figures reproduced from arXiv: 1908.09171 by the authors.

Figure 1
Figure 1. Robot delivers package to front door. If the robot has no prior [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. System architecture. To plan a path to an unknown goal position beyond the sensing horizon, a robot’s sensor data is used to build a semantically [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Training data creation. A satellite view of a house’s front yard is manually converted to a semantic map (top left), with colors for different [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Qualitative Assessment. Network’s predicted cost-to-go on 10 [PITH_FULL_IMAGE:figures/full_fig_p004_4.png]
Figure 5
Figure 5. Figure 5: Comparing network loss functions. The performance on the validation [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 7
Figure 7. Figure 7: Planner performance across neighborhoods. DC2G reaches the goal [PITH_FULL_IMAGE:figures/full_fig_p006_7.png]
Figure 8
Figure 8. Figure 8: Sample gridworld scenario. The top row is the agent’s observed semantic map (network input); the middle is its current estimate of the cost-to-go [PITH_FULL_IMAGE:figures/full_fig_p007_8.png]
Figure 9
Figure 9. Figure 9: Unreal Simulation. Using a network trained only on aerial images, [PITH_FULL_IMAGE:figures/full_fig_p007_9.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

34 extracted references · 26 canonical work pages

  1. [1]

    Frontier-based exploration using multiple robots,

    B. Yamauchi, “Frontier-based exploration using multiple robots,” in Proceedings of the second international conference on Autonomous agents. ACM, 1998, pp. 47–53

    work page 1998

  2. [2]

    Information gain-based exploration using rao-blackwellized particle filters

    C. Stachniss, G. Grisetti, and W. Burgard, “Information gain-based exploration using rao-blackwellized particle filters.” in Robotics: Science and Systems , vol. 2, 2005, pp. 65–72

    work page 2005

  3. [3]

    Learning search heuristics for finding objects in structured environments,

    D. Joho, M. Senk, and W. Burgard, “Learning search heuristics for finding objects in structured environments,” Robotics and Autonomous Systems, vol. 59, no. 5, pp. 319–328, 2011

    work page 2011

  4. [4]

    Using the web to interactively learn to find objects

    M. Samadi, T. Kollar, and M. M. Veloso, “Using the web to interactively learn to find objects.” in AAAI, 2012, pp. 2074–2080

    work page 2012

  5. [5]

    Utilizing object-object and object-scene context when planning to find things,

    T. Kollar and N. Roy, “Utilizing object-object and object-scene context when planning to find things,” in Robotics and Automation, 2009. ICRA’09. IEEE International Conference on . IEEE, 2009, pp. 2168– 2173

    work page 2009

  6. [6]

    Bootstrapping probabilistic models of qualitative spatial relations for active visual object search,

    L. Kunze, C. Burbridge, and N. Hawes, “Bootstrapping probabilistic models of qualitative spatial relations for active visual object search,” in AAAI Spring Symposium , 2014, pp. 24–26

    work page 2014

  7. [7]

    Using qualitative spatial relations for indirect object search,

    L. Kunze, K. K. Doreswamy, and N. Hawes, “Using qualitative spatial relations for indirect object search,” in Robotics and Automation (ICRA), 2014 IEEE International Conference on . IEEE, 2014, pp. 163–168

    work page 2014

  8. [8]

    Prior-assisted propagation of spatial information for object search,

    M. Lorbach, S. H ¨ofer, and O. Brock, “Prior-assisted propagation of spatial information for object search,” in Intelligent Robots and Systems (IROS 2014), 2014 IEEE/RSJ International Conference on . IEEE, 2014, pp. 2904–2909

    work page 2014

Show all 34 references
  1. [9]

    Learning to locate from demonstrated searches

    P. Vernaza and A. Stentz, “Learning to locate from demonstrated searches.” in Robotics: Science and Systems , 2014

    2014

  2. [10]

    Active visual object search in unknown environments using uncertain seman- tics,

    A. Aydemir, A. Pronobis, M. G ¨obelbecker, and P. Jensfelt, “Active visual object search in unknown environments using uncertain seman- tics,” IEEE Transactions on Robotics , vol. 29, no. 4, pp. 986–1002, 2013

    2013

  3. [11]

    Robot task planning and explanation in open and uncertain worlds,

    M. Hanheide, M. G ¨obelbecker, G. S. Horn, A. Pronobis, K. Sj ¨o¨o, A. Aydemir, P. Jensfelt, C. Gretton, R. Dearden, M. Janicek, et al. , “Robot task planning and explanation in open and uncertain worlds,” Artificial Intelligence, vol. 247, pp. 119–150, 2017

    2017

  4. [12]

    Deepnav: Learning to navigate large cities,

    S. Brahmbhatt and J. Hays, “Deepnav: Learning to navigate large cities,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. , 2017, pp. 3087–3096

    2017

  5. [14]

    Following high-level navigation instructions on a simulated quadcopter with imitation learning,

    V . Blukis, N. Brukhim, A. Bennett, R. A. Knepper, and Y . Artzi, “Following high-level navigation instructions on a simulated quadcopter with imitation learning,” arXiv preprint arXiv:1806.00047 , 2018

    2018 arXiv

  6. [15]

    Target-driven Visual Navigation in Indoor Scenes using Deep Reinforcement Learning,

    Y . Zhu, R. Mottaghi, E. Kolve, J. J. Lim, A. Gupta, L. Fei-Fei, and A. Farhadi, “Target-driven Visual Navigation in Indoor Scenes using Deep Reinforcement Learning,” in IEEE International Conference on Robotics and Automation , 2017

    2017

  7. [16]

    Visual semantic planning using deep successor representations,

    Y . Zhu, D. Gordon, E. Kolve, D. Fox, L. Fei-Fei, A. Gupta, R. Mottaghi, and A. Farhadi, “Visual semantic planning using deep successor representations,” in Proceedings of the IEEE International Conference on Computer Vision , 2017, pp. 483–492

    2017

  8. [17]

    Iqa: Visual question answering in interactive environments,

    D. Gordon, A. Kembhavi, M. Rastegari, J. Redmon, D. Fox, and A. Farhadi, “Iqa: Visual question answering in interactive environments,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018, pp. 4089–4098

    2018

  9. [18]

    Cognitive mapping and planning for visual navigation,

    S. Gupta, J. Davidson, S. Levine, R. Sukthankar, and J. Malik, “Cognitive mapping and planning for visual navigation,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , 2017, pp. 2616–2625

    2017

  10. [19]

    Em- bodied Question Answering,

    A. Das, S. Datta, G. Gkioxari, S. Lee, D. Parikh, and D. Batra, “Em- bodied Question Answering,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , 2018

    2018

  11. [20]

    U-net: Convolutional net- works for biomedical image segmentation,

    O. Ronneberger, P. Fischer, and T. Brox, “U-net: Convolutional net- works for biomedical image segmentation,” in International Conference on Medical image computing and computer-assisted intervention . Springer, 2015, pp. 234–241

    2015

  12. [21]

    Image-to-image translation with conditional adversarial networks,

    P. Isola, J.-Y . Zhu, T. Zhou, and A. A. Efros, “Image-to-image translation with conditional adversarial networks,” CVPR, 2017

    2017

  13. [22]

    Learning deep generative spatial models for mobile robots,

    A. Pronobis and R. P. Rao, “Learning deep generative spatial models for mobile robots,” in Intelligent Robots and Systems (IROS), 2017 IEEE/RSJ International Conference on . IEEE, 2017, pp. 755–762

    2017

  14. [23]

    High-speed autonomous navigation of unknown environments using learned probabilities of collision,

    C. Richter, J. Ware, and N. Roy, “High-speed autonomous navigation of unknown environments using learned probabilities of collision,” in 2014 IEEE International Conference on Robotics and Automation (ICRA), May 2014, pp. 6114–6121

    2014

  15. [24]

    Learning to plan for visibility in navigation of unknown environments,

    C. Richter and N. Roy, “Learning to plan for visibility in navigation of unknown environments,” in 2016 International Symposium on Experimental Robotics , D. Kuli ´c, Y . Nakamura, O. Khatib, and G. Venture, Eds. Cham: Springer International Publishing, 2017, pp. 387–398

    2016

  16. [25]

    Richter, W

    C. Richter, W. Vega-Brown, and N. Roy, Bayesian Learning for Safe High-Speed Navigation in Unknown Environments . Cham: Springer International Publishing, 2018, pp. 325–341. [Online]. Available: https://doi.org/10.1007/978-3-319-60916-4 19

    2018 doi

  17. [26]

    Learning heuristic search via imitation,

    M. Bhardwaj, S. Choudhury, and S. Scherer, “Learning heuristic search via imitation,” arXiv preprint arXiv:1707.03034 , 2017

    2017 arXiv

  18. [28]

    Image-to-image translation in tensorflow,

    C. Hesse, “Image-to-image translation in tensorflow,” https://affinelayer. com/pix2pix/, accessed: 2018-08-28

    2018

  19. [29]

    A note on two problems in connexion with graphs,

    E. W. Dijkstra, “A note on two problems in connexion with graphs,” Numerische mathematik, vol. 1, no. 1, pp. 269–271, 1959

    1959

  20. [30]

    Bing maps,

    Microsoft, “Bing maps,” https://www.bing.com/maps, accessed: 2019- 03-01

    2019

  21. [31]

    Mask r-cnn,

    K. He, G. Gkioxari, P. Doll ´ar, and R. Girshick, “Mask r-cnn,” IEEE transactions on pattern analysis and machine intelligence , 2018

    2018

  22. [32]

    Real-time voxel based 3d semantic mapping with a hand held rgb-d camera,

    Z. Xuan and F. David, “Real-time voxel based 3d semantic mapping with a hand held rgb-d camera,” https://github.com/floatlazer/semantic slam, 2018

    2018

  23. [33]

    Improved techniques for training gans,

    T. Salimans, I. Goodfellow, W. Zaremba, V . Cheung, A. Radford, and X. Chen, “Improved techniques for training gans,” in Advances in Neural Information Processing Systems , 2016, pp. 2234–2242

    2016

  24. [34]

    Minimalistic gridworld environment for openai gym,

    L. W. Maxime Chevalier-Boisvert, “Minimalistic gridworld environment for openai gym,” https://github.com/maximecb/gym-minigrid, 2018

    2018

  25. [35]

    Airsim: High-fidelity visual and physical simulation for autonomous vehicles,

    S. Shah, D. Dey, C. Lovett, and A. Kapoor, “Airsim: High-fidelity visual and physical simulation for autonomous vehicles,” in Field and service robotics. Springer, 2018, pp. 621–635

    2018

  26. [2018]

    Available: http://arxiv.org/abs/1801.02209

    [Online]. Available: http://arxiv.org/abs/1801.02209

    arXiv

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.