Pith. sign in

REVIEW 4 major objections 5 minor 37 references

Multi-Robot Reliable Navigation in Uncertain Topological Environments with Graph Attention Networks

T0 review · 4 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read A graph-attention policy lets robot teams reroute as uncertain roads are revealed.

desk verdict A useful new combination for multi-agent CTP that deserves review once the undisclosed training expert is specified and ablated. read the letter →

arxiv 2411.16134 v1 pith:WTZPCXPT submitted 2024-11-25 cs.RO

Zhuoyuan Yu , Hongliang Guo , Albertus Hendrawan Adiwahono , Jianle Chan , Brina Shong Wey Tynn , Chee-Meng Chew , Wei-Yun Yau This is my paper · ORCID
classification cs.RO
keywords multi-robotnavigationCanadiantravellerproblemGraphAttentionNetworksPOMDPon-timearrivalprobabilityadaptiveembeddingmulti-agentreinforcementlearninguncertaintopological
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 team of robots can navigate road networks with unknown edge passability by learning to plan and reroute as information is revealed, without needing prior traversal probabilities at deployment. It reformulates the problem as a partially observable Markov decision process whose objective is the weighted on-time arrival probability of the whole team minus the information entropy of the explored graph. The proposed MARVEL pipeline uses cost-weighted shortest-path chains to embed graph nodes, a graph attention network to pick the next edge, and an online expert to supply training labels for faster convergence. If the claim holds, multi-robot systems can balance individual punctuality against team-level exploration in emergencies, time-critical deliveries, and other settings where road conditions change unexpectedly.

What carries the argument

The machinery is a three-part pipeline: (i) a dynamic adaptive graph embedding that replaces random walks with Dijkstra shortest-path chains weighted by edge costs and appends origin-destination positions, so node features reflect navigation value rather than raw topology; (ii) a multi-head Graph Attention Network policy that computes attention coefficients over neighbouring nodes and turns dot-product similarities into edge-selection probabilities; and (iii) an online expert-augmented policy gradient whose objective is the team's weighted stochastic on-time-arrival probability minus the information entropy of explored edges. The entropy term is the exploration lever: it rewards a robot for resolving uncertainty in edges whose passability would change teammates' decisions, while the cross-entropy loss against the expert's prior solutions accelerates convergence.

What would settle it

Run the centralized expert on the same test networks and compare its team on-time arrival probability with MARVEL's; if MARVEL does not match or beat the expert, the reported superiority may be inherited from the expert rather than learned. Then retrain MARVEL using training labels generated by a solver with access only to information the robots actually possess; if team on-time arrival probability drops sharply, the claimed no-prior-knowledge capability fails.

Watch

Extended reading notes

Core claim

MARVEL's central claim is that a team of robots, each with its own origin, destination, time budget, and task priority, can maximize the weighted on-time arrival probability of the whole team even when edge passability is unknown until a robot reaches one end of the edge. The key is to treat the evolving explored graph as a partially observable state, embed it with cost-weighted shortest-path chains rather than random walks, and select next edges with a Graph Attention Network whose self-attention weights shift as conditions are revealed. Training is guided by an online expert that supplies prior optimal solutions, and the objective subtracts the graph's information entropy from the team's success probability, which makes lower-priority robots sacrifice punctuality to probe uncertain edges when that probing helps the team. The paper reports that MARVEL outperforms three reliable-shortest-path baselines and three Canadian-traveller-problem solvers on four standard transportation networks in team on-time arrival probability, and that the behavior transfers to two physical robots in a maze.

Load-bearing premise

The load-bearing premise is that the centralized controller used in training can compute a genuinely optimal prior solution for every state a robot visits, and that imitating those labels is what transfers to unseen networks; if that expert relies on privileged traversal probabilities or is a weak solver, MARVEL's deployment-time claim is narrower than it appears.

Editorial extensions

If this is right

  • If MARVEL's claim is correct, robots can be sent into an unfamiliar network with no passability prior and still adapt their routes online as edges are revealed.
  • The mechanism implies a principled division of labor: low-priority robots absorb punctuality loss to probe uncertain edges, and the team's overall weighted success probability rises.
  • The reported results across four networks of different sizes suggest one trained policy can generalize to new topological maps without per-network retraining.
  • Physical maze experiments with two robots show the learned exploration-and-reroute behavior survives on real hardware.

Reading between the lines

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

  • A natural extension the paper does not test: the entropy-driven exploration bonus could be added to single-agent CTP solvers as a curiosity reward, where it should improve information gain without needing a second robot.
  • Because the expert labels come from a centralized controller at training time only, the 'no prior knowledge' claim is about inference; a stricter test would train MARVEL with labels from a solver that sees only what robots observe, which would separate the learned policy's contribution from the expert's.
  • The POMDP formulation is generic enough to cover last-mile delivery or emergency response, but only if the time budgets and traversal distributions in those settings resemble the truncated Gaussian costs used here; that transfer is an open question.
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

4 major / 5 minor

Summary. The paper proposes MARVEL, a learning-based approach for multi-robot navigation in uncertain topological networks, where edge traversability is unknown until a robot reaches the edge's starting node. The problem is formulated as a POMDP with a team objective combining weighted on-time arrival probabilities and an information-entropy exploration bonus. The method combines a dynamic graph embedding step, a Graph Attention Network (GAT) policy, and an 'online expert' that provides supervision during training. The authors report that MARVEL outperforms several single-agent CTP and reliable-path-finding baselines on four standard transportation networks, and they include a physical experiment with two Pioneer 3-DX robots in a maze with an uncertain edge. The paper also provides an ablation study and a link to public source code.

Significance. If the central claims hold, MARVEL is a meaningful step toward multi-robot coordination under topological uncertainty, particularly because it operates at inference without prior knowledge of edge passability and can adapt routes in real time based on observations from teammates. The inclusion of real-robot experiments and a publicly available code repository is commendable and increases the paper's reproducibility. The ablation study provides useful evidence about the contributions of the attention mechanism and the entropy term. However, the credibility of the main empirical result hinges on the specification of the 'online expert' used for training, which is currently undefined, and on the absence of error bars and multi-agent baselines in the comparisons. These issues must be addressed before the performance claims can be considered fully supported.

major comments (4)
  1. [Section IV-C and Algorithm 1] The training procedure is not fully specified. The text states that 'a centralized controller' computes 'the prior optimal solution for each state' and that the model is trained by imitating the expert via a cross-entropy loss L, but Algorithm 1 only uses the policy gradient of Eq. (21) and never mentions L or the expert. This internal inconsistency makes it impossible to determine what was actually trained. Please define the expert precisely, including its inputs, its optimization objective, whether it has access to true traversal probabilities or other privileged information, and its computational cost. Also, reconcile Algorithm 1 with the described imitation loss, or explain how the cross-entropy loss enters the gradient in Eq. (21). Without this information, the 'no prior knowledge' claim at inference and the reported superiority over baselines may partly reflect the expert's performance rather than the learned policy.
  2. [Section V-D, Fig. 4] The central claim that MARVEL 'consistently outperforms' baselines is supported only by aggregate bar charts with no error bars, confidence intervals, or significance tests. Since on-time arrival probabilities are estimated from 10,000 simulations per OD pair, the variability across random seeds and across OD pairs should be reported. Please add standard deviations, box plots, or statistical significance tests (e.g., paired tests across the same OD pairs) to substantiate the comparisons.
  3. [Section V-D, Baselines] All baselines are single-agent algorithms evaluated independently for each agent; there is no multi-agent coordination baseline (e.g., a MA-CTP method, a multi-agent pathfinding variant, or a centralized greedy planner). The paper's central narrative emphasizes team collaboration and 'collective intelligence,' but without a multi-agent baseline it is unclear whether the gains come from coordination or simply from the GAT policy's representational power. Please add a multi-agent baseline or an ablation with independent (non-coordinated) policies to isolate the effect of coordination.
  4. [Equations (16)-(21)] The derivation of the policy gradient is not mathematically self-consistent. In Eq. (16), the SOTA probability appears as a weighting factor inside the expectation, but Eq. (17) replaces it with a sigmoid of the travel time against the budget, which is not derived. Moreover, Eq. (19) defines ΔH as a sum over explored edges, but the objective in Eq. (1) includes the graph entropy H(G) as a global term, not a trajectory-dependent one. Please clarify the relationship between the objective and the gradient estimators, and provide a clean derivation or cite the relevant variational policy gradient theorem appropriately.
minor comments (5)
  1. [Section III-C] In the text after Eq. (4), the information entropy H is defined in Eq. (4), but the sentence in Section III-C says 'H is defined in Eq. (5)', which is actually the normalized adjacency matrix. Please correct the cross-reference.
  2. [Section IV-B] The statement 'for ∀i ∈ E' should likely be 'for ∀i ∈ V' or 'for all vertices i', since attention coefficients are computed over vertices and their neighbors. Please fix the notation.
  3. [Section V-A.1] The description 'sampled from a Gaussian distribution with rejection' is vague. Please specify the exact truncation/rejection procedure, including the minimum bound and whether the rejection is independent across edges.
  4. [Algorithm 1] Algorithm 1 outputs 'at = SoftMax(πθ)', but Section IV-B states that during testing the action is selected as the edge with maximum probability. Please clarify whether the algorithm outputs a distribution or a deterministic action, and align the pseudocode with the text.
  5. [Section III-A] The POMDP formulation is presented as two tuples but does not explicitly define the state space, observation space, transition function, or reward function. Please provide a formal POMDP definition with these components, or state clearly how they map to the tuple notation.

Circularity Check

1 steps flagged · score 4.0 of 10

Performance claim is partly fitted to an undisclosed 'online expert' that optimizes the same objective, though ablation shows the expert is not the sole source.

  1. fitted input called prediction [Section IV-C 'Online Expert-Augmented Policy Gradient' and Algorithm 1]
    "Next, we use the centralized controller to compute the prior optimal solution for each state. The model is trained by imitating the decisions of the “expert”. As such, the objective is to minimize the loss L between the predicted solution and the optimal solution, which essentially corresponds to maximizing the objective function Jθ."

    The 'prior optimal solution' labels are generated by a centralized controller solving the identical objective Jθ (Eq. 1) on which MARVEL is later evaluated. Training fits the GAT policy to those labels via cross-entropy, so the reported team on-time arrival probability can be inherited from the undisclosed expert rather than derived from the learned policy. The paper never specifies the expert's inputs, so it may use the true edge traversal probabilities that MARVEL claims not to need. Algorithm 1 also omits the cross-entropy loss L, making it impossible to tell which objective was actually optimized. However, the ablation 'No Cross Entropy Loss' retains 62% vs 63% SOTA, so the expert is not the only source of performance; hence partial circularity.

full rationale

MARVEL's policy-gradient update (Eq. 21) is a standard variational derivation from the stated objective (Eq. 1), and the graph embedding and GAT components are defined independently of the evaluation metric. The main circularity burden is the online expert: the paper fits the policy to the expert's 'prior optimal solutions' of the same objective and then reports the policy's score on that objective, with the expert undefined and unbenchmarked. This is a fitted-input-called-prediction pattern, but it is partial rather than total because the paper's own ablation removes the cross-entropy (expert) loss with only a 1-point SOTA drop on Sioux Falls, and the method is also tested against external baselines and on physical robots. The self-cited baselines (RAO*, FMA, DDP, GP4) are used as comparators or for parameter settings, not as load-bearing justification for the method, so they do not raise the score further. Overall: 4 of 10, reflecting a meaningful but not complete circularity.

Assumptions & free parameters 5 free parameters · 6 assumptions · 0 invented entities

The ledger shows the core contribution is an architecture and training recipe, not a first-principles derivation. The result depends on a hidden expert, a distributional travel-cost model, and an unvalidated entropy proxy.

free parameters (5)
  • Time budget multipliers for low-priority OD pairs = 1.2 times tLET
    Section V-D sets low-priority time budgets to 1.2 times least expected time; this choice directly shapes the reported SOTA values and has no sensitivity analysis.
  • Time budget multipliers for high-priority OD pairs = 0.95, 1.0, and 1.05 times tLET
    Section V-D adopts three use cases from [37]; the tight budget makes baselines look worse and MARVEL's margin larger.
  • Importance weights lambda = [0.7, 0.3] and [0.3, 0.7] in the toy network; not specified for benchmark networks
    Section V-C and V-D; objective Eq. (1) depends on these weights, but benchmark assignments are not reported.
  • Entropy versus SOTA balance coefficient = Implicitly 1.0
    Eq. (21) subtracts delta-H with no tunable coefficient; the scale between a sigmoid probability and a log-probability entropy term is arbitrary and untested.
  • Network hyperparameters = feature 128, heads 8, lr 1e-3, decay 0.95
    Section V-A.3; standard choices, no sensitivity sweep reported.
assumptions (6)
  • domain assumption Edge passability is revealed only when a robot reaches the source node of the edge; no other communication about that edge is available.
    Section III-A defines the CTP-style uncertainty; this is the paper's core information model, adopted without empirical justification.
  • domain assumption Travel costs on each edge are independent Gaussian variables with known mean and variance, truncated by rejection to enforce minimum travel times.
    Section V-A.1; all simulation results and the SOTA estimates depend on this distributional assumption.
  • ad hoc to paper A centralized online expert can compute a prior optimal solution for every encountered state.
    Section IV-C; the expert is the teacher for the policy, but its algorithm, cost, and information access are not disclosed.
  • standard math The variational policy gradient theorem applies to the POMDP with the SOTA surrogate and entropy-shaping reward defined in Section IV-C.
    The theorem is standard, but the POMDP state space, observation function, and reward are never formalized, so the application is asserted rather than justified.
  • domain assumption Shortest paths to all destinations form a corpus whose skip-gram embeddings preserve the information needed for CTP navigation.
    Section IV-A replaces random walks with Dijkstra chains; the paper gives a gradient argument against Node2Vec but no proof for the new sampling method.
  • ad hoc to paper The explored-edge entropy delta-H = sum p(e) log p(e) is a valid proxy for the value of team exploration.
    Eq. (19) is introduced without derivation and with a sign that depends on the convention in Eq. (4); no ablation isolates this term alone.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Multi-Robot Reliable Navigation in Uncertain Topological Environments with Graph Attention Networks." pith.science (2026). https://pith.science/paper/WTZPCXPT

@misc{pith2026241116134,
  author       = {Pith},
  title        = {Pith review of: Multi-Robot Reliable Navigation in Uncertain Topological Environments with Graph Attention Networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WTZPCXPT}},
  note         = {Machine review of arXiv:2411.16134}
}
read the original abstract

This paper studies the multi-robot reliable navigation problem in uncertain topological networks, which aims at maximizing the robot team's on-time arrival probabilities in the face of road network uncertainties. The uncertainty in these networks stems from the unknown edge traversability, which is only revealed to the robot upon its arrival at the edge's starting node. Existing approaches often struggle to adapt to real-time network topology changes, making them unsuitable for varying topological environments. To address the challenge, we reformulate the problem into a Partially Observable Markov Decision Process (POMDP) framework and introduce the Dynamic Adaptive Graph Embedding method to capture the evolving nature of the navigation task. We further enhance each robot's policy learning process by integrating deep reinforcement learning with Graph Attention Networks (GATs), leveraging self-attention to focus on critical graph features. The proposed approach, namely Multi-Agent Routing in Variable Environments with Learning (MARVEL) employs the generalized policy gradient algorithm to optimize the robots' real-time decision-making process iteratively. We compare the performance of MARVEL with state-of-the-art reliable navigation algorithms as well as Canadian traveller problem solutions in a range of canonical transportation networks, demonstrating improved adaptability and performance in uncertain topological networks. Additionally, real-world experiments with two robots navigating within a self-constructed indoor environment with uncertain topological structures demonstrate MARVEL's practicality.

Figures

Figures reproduced from arXiv: 2411.16134 by the authors.

Figure 1
Figure 1. The illustrative example: In the topological network, the blue and green [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. The MARVEL framework outlines a processing pipeline composed of a graph embedder, a graph neural network module, and a policy update module based on the online expert. (i) The feature matrix and adjacency matrix are dynamically updated by the real-time structural and positional information. The corpus is generated for each node based on neighbors and shortest paths to all destination nodes. (ii) Our policy network c… view at source ↗
Figure 3
Figure 3. The SOTA probability performance comparison of different scenarios. [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: The SOTA probability performance comparison with the baseline algorithms on different canonical transportation networks. [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Constructed physical environments and Rviz Virtual Maps. [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

37 extracted references · 27 canonical work pages

  1. [1]

    The robust Canadian traveller problem applied to robot routing,

    H. Guo and T. D. Barfoot, “The robust Canadian traveller problem applied to robot routing,” in 2019 International Conference on Robotics and Automation (ICRA) . IEEE, 2019, pp. 5523–5529

    work page 2019

  2. [2]

    Using reinforcement learning to minimize the probability of delay oc- currence in transportation,

    Z. Cao, H. Guo, W. Song, K. Gao, Z. Chen, L. Zhang, and X. Zhang, “Using reinforcement learning to minimize the probability of delay oc- currence in transportation,” IEEE Transactions on Vehicular Technology, vol. 69, no. 3, pp. 2424–2436, 2020

    work page 2020

  3. [3]

    Multi-robot path planning with due times,

    H. Wang and W. Chen, “Multi-robot path planning with due times,” IEEE Robotics and Automation Letters , vol. 7, no. 2, pp. 4829–4836, 2022

    work page 2022

  4. [4]

    Attention-based priority learning for limited time multi-agent path finding,

    Y . Yang, M. Fan, C. He, J. Wang, H. Huang, and G. Sartoretti, “Attention-based priority learning for limited time multi-agent path finding,” in Proceedings of the 23rd International Conference on Au- tonomous Agents and Multiagent Systems , 2024, pp. 1993–2001

    work page 2024

  5. [5]

    Multiagent-based route guidance for increasing the chance of arrival on time,

    Z. Cao, H. Guo, J. Zhang, and U. Fastenrath, “Multiagent-based route guidance for increasing the chance of arrival on time,” in Proceedings of the AAAI Conference on Artificial Intelligence , vol. 30, no. 1, 2016

    work page 2016

  6. [6]

    Multi-agent path finding with delay probabilities,

    H. Ma, T. S. Kumar, and S. Koenig, “Multi-agent path finding with delay probabilities,” in Proceedings of the AAAI Conference on Artificial Intelligence, vol. 31, no. 1, 2017

    work page 2017

  7. [7]

    On the online multi-agent o–d k-Canadian traveller problem,

    D. Shiri and F. S. Salman, “On the online multi-agent o–d k-Canadian traveller problem,” Journal of Combinatorial Optimization , vol. 34, pp. 453–461, 2017

    work page 2017

  8. [8]

    Shortest paths without a map,

    C. H. Papadimitriou and M. Yannakakis, “Shortest paths without a map,” Theoretical Computer Science , vol. 84, no. 1, pp. 127–150, 1991

    work page 1991

Show all 37 references
  1. [9]

    On the impact of uncertainty for path planning,

    J. Guzzi, R. O. Chavez-Garcia, L. M. Gambardella, and A. Giusti, “On the impact of uncertainty for path planning,” in 2019 International Conference on Robotics and Automation (ICRA) . IEEE, 2019, pp. 5929–5935

    2019

  2. [10]

    Generalized Canadian traveller problems,

    C.-S. Liao and Y . Huang, “Generalized Canadian traveller problems,” Journal of Combinatorial Optimization , vol. 29, pp. 701–712, 2015

    2015

  3. [11]

    Route planning under uncertainty: The Canadian traveller problem

    E. Nikolova and D. R. Karger, “Route planning under uncertainty: The Canadian traveller problem.” in AAAI, 2008, pp. 969–974

    2008

  4. [12]

    An optimal randomized online algorithm for the k-Canadian traveller problem on node-disjoint paths,

    M. Bender and S. Westphal, “An optimal randomized online algorithm for the k-Canadian traveller problem on node-disjoint paths,” Journal of Combinatorial Optimization, vol. 30, no. 1, pp. 87–96, 2015

    2015

  5. [13]

    The recoverable Canadian traveller problem based on limited provision information,

    B. Su and X. LAN, “The recoverable Canadian traveller problem based on limited provision information,” Syst. Eng., vol. 9, pp. 102–107, 2009

    2009

  6. [14]

    The covering Canadian traveller problem,

    C.-S. Liao and Y . Huang, “The covering Canadian traveller problem,” Theoretical Computer Science , vol. 530, pp. 80–88, 2014

    2014

  7. [15]

    Network-flow-problem- based approach to multi-agent path finding for connected autonomous vehicles,

    A. Okoso, B. Okumura, K. Otaki, and T. Nishi, “Network-flow-problem- based approach to multi-agent path finding for connected autonomous vehicles,” in 2021 IEEE International Intelligent Transportation Systems Conference (ITSC). IEEE, 2021, pp. 1946–1953

    2021

  8. [16]

    Distributed heuristic multi-agent path finding with communication,

    Z. Ma, Y . Luo, and H. Ma, “Distributed heuristic multi-agent path finding with communication,” in 2021 IEEE International Conference on Robotics and Automation (ICRA) . IEEE, 2021, pp. 8699–8705

    2021

  9. [17]

    Social behavior as a key to learning-based multi-agent pathfinding dilemmas,

    C. He, T. Duhan, P. Tulsyan, P. Kim, and G. Sartoretti, “Social behavior as a key to learning-based multi-agent pathfinding dilemmas,” arXiv preprint arXiv:2408.03063, 2024

    2024 arXiv

  10. [18]

    Feasibility study: Moving non-homogeneous teams in congested video game environ- ments,

    H. Ma, J. Yang, L. Cohen, T. Kumar, and S. Koenig, “Feasibility study: Moving non-homogeneous teams in congested video game environ- ments,” in Proceedings of the AAAI Conference on Artificial Intelligence and Interactive Digital Entertainment, vol. 13, no. 1, 2017, pp. 270–272

    2017

  11. [19]

    A comprehensive review on leveraging machine learning for multi-agent path finding,

    J.-M. Alkazzi and K. Okumura, “A comprehensive review on leveraging machine learning for multi-agent path finding,” IEEE Access, 2024

    2024

  12. [20]

    Alpha: Attention- based long-horizon pathfinding in highly-structured areas,

    C. He, T. Yang, T. Duhan, Y . Wang, and G. Sartoretti, “Alpha: Attention- based long-horizon pathfinding in highly-structured areas,” in2024 IEEE International Conference on Robotics and Automation (ICRA) . IEEE, 2024, pp. 14 576–14 582

    2024

  13. [21]

    PRIMAL 2: Pathfind- ing via reinforcement and imitation multi-agent learning-lifelong,

    M. Damani, Z. Luo, E. Wenzel, and G. Sartoretti, “PRIMAL 2: Pathfind- ing via reinforcement and imitation multi-agent learning-lifelong,” IEEE Robotics and Automation Letters , vol. 6, no. 2, pp. 2666–2673, 2021

    2021

  14. [22]

    SCRIMP: Scalable communication for reinforcement-and imitation-learning-based multi- agent pathfinding,

    Y . Wang, B. Xiang, S. Huang, and G. Sartoretti, “SCRIMP: Scalable communication for reinforcement-and imitation-learning-based multi- agent pathfinding,” in 2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) . IEEE, 2023, pp. 9301–9308

    2023

  15. [23]

    Multi-agent path finding with prioritized communication learning,

    W. Li, H. Chen, B. Jin, W. Tan, H. Zha, and X. Wang, “Multi-agent path finding with prioritized communication learning,” in 2022 International Conference on Robotics and Automation (ICRA) . IEEE, 2022, pp. 10 695–10 701

    2022

  16. [24]

    Graph neural networks for decentralized multi-robot path planning,

    Q. Li, F. Gama, A. Ribeiro, and A. Prorok, “Graph neural networks for decentralized multi-robot path planning,” in 2020 IEEE/RSJ Inter- national Conference on Intelligent Robots and Systems (IROS) . IEEE, 2020, pp. 11 785–11 792

    2020

  17. [25]

    Message-aware graph attention networks for large-scale multi-robot path planning,

    Q. Li, W. Lin, Z. Liu, and A. Prorok, “Message-aware graph attention networks for large-scale multi-robot path planning,” IEEE Robotics and Automation Letters, vol. 6, no. 3, pp. 5533–5540, 2021

    2021

  18. [26]

    Deep reinforcement learning for multiagent systems: A review of challenges, solutions, and applications,

    T. T. Nguyen, N. D. Nguyen, and S. Nahavandi, “Deep reinforcement learning for multiagent systems: A review of challenges, solutions, and applications,” IEEE transactions on cybernetics, vol. 50, no. 9, pp. 3826– 3839, 2020

    2020

  19. [27]

    Deep reinforcement learning for autonomous driving: A survey,

    B. R. Kiran, I. Sobh, V . Talpaert, P. Mannion, A. A. Al Sallab, S. Yo- gamani, and P. Pérez, “Deep reinforcement learning for autonomous driving: A survey,” IEEE Transactions on Intelligent Transportation Systems, vol. 23, no. 6, pp. 4909–4926, 2021

    2021

  20. [28]

    Efficient estimation of word representations in vector space,

    T. Mikolov, K. Chen, G. Corrado, and J. Dean, “Efficient estimation of word representations in vector space,” arXiv preprint arXiv:1301.3781 , 2013

    2013 arXiv

  21. [29]

    A mathematical theory of communication,

    C. E. Shannon, “A mathematical theory of communication,” The Bell system technical journal , vol. 27, no. 3, pp. 379–423, 1948

    1948

  22. [30]

    Semi-supervised classification with graph convolutional networks,

    T. N. Kipf and M. Welling, “Semi-supervised classification with graph convolutional networks,” arXiv preprint arXiv:1609.02907 , 2016

    2016 arXiv

  23. [31]

    Link prediction based on graph neural net- works,

    M. Zhang and Y . Chen, “Link prediction based on graph neural net- works,” Advances in Neural Information Processing Systems , vol. 31, 2018

    2018

  24. [32]

    Transportation network test problems,

    H. Bar-Gera, “Transportation network test problems,” Last Accessed July 19, 2024. [Online]. Available: https://github.com/bstabler/ TransportationNetworks

    2024

  25. [33]

    Navigation with time limits in transportation networks: A fourth moment approach,

    H. Guo, Z. He, C. Gao, and D. Rus, “Navigation with time limits in transportation networks: A fourth moment approach,”IEEE Transactions on Intelligent Transportation Systems, vol. 23, no. 12, pp. 23 781–23 796, 2022

    2022

  26. [34]

    Algorithms for most reliable routes on stochastic and time-dependent networks,

    A. A. Prakash, “Algorithms for most reliable routes on stochastic and time-dependent networks,” Transportation Research Part B: Method- ological, vol. 138, pp. 202–220, 2020

    2020

  27. [35]

    Most reliable path-finding algorithm for maximizing on-time arrival probability,

    B. Y . Chen, C. Shi, J. Zhang, W. H. Lam, Q. Li, and S. Xiang, “Most reliable path-finding algorithm for maximizing on-time arrival probability,” Transportmetrica B: Transport Dynamics, vol. 5, no. 3, pp. 248–264, 2017

    2017

  28. [36]

    Dual dynamic programming for the mean standard deviation Canadian traveller problem,

    H. Guo, R. Shi, D. Rus, and W.-Y . Yau, “Dual dynamic programming for the mean standard deviation Canadian traveller problem,” IEEE Transactions on Vehicular Technology , vol. 71, no. 11, pp. 11 465– 11 479, 2022

    2022

  29. [37]

    GP4: Gaussian process proactive path planning for the stochastic on time arrival problem,

    C. Gao, H. Guo, and W. Sheng, “GP4: Gaussian process proactive path planning for the stochastic on time arrival problem,” IEEE Transactions on Vehicular Technology, vol. 70, no. 10, pp. 9849–9862, 2021

    2021

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.