Pith. sign in

REVIEW 4 major objections 10 minor 36 references

MacLight: Multi-scene Aggregation Convolutional Learning for Traffic Signal Control

T0 review · 4 major / 10 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 Traffic-light agents train under an hour via scene compression

desk verdict Plausible speed win and a useful new dynamic-blockage scenario, but the abstract overclaims stability and the baseline comparison is compromised by missing tuning/reward details. read the letter →

arxiv 2412.15703 v3 pith:NWZIQWNC submitted 2024-12-20 cs.MA cs.AIcs.LG

Sunbowen Lee , Hongqin Lyu , Yicheng Gong , Yingying Sun , Chao Deng This is my paper · ORCID
classification cs.MAcs.AIcs.LG
keywords TrafficsignalcontrolMulti-sceneconvolutionVariationalautoencoderMulti-agentreinforcementlearningPPODynamicsimulationSUMOGlobalstaterepresentation
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

MacLight aims to show that traffic-signal control becomes faster and more stable when global road information is compressed by a convolutional variational autoencoder and fed to the value network of a PPO controller, rather than modeled as a graph-attention deep Q-network. The paper tests the method on a 4x4 SUMO grid under normal, heavy, and dynamically blocked traffic, reporting training times of 39 to 58 minutes per 80 episodes against 137 to 456 minutes for DQN-based baselines. It also introduces a dynamic-traffic benchmark in which random roads close and vehicles reroute, which the authors say previous peer studies did not deploy. The practical interest is that training cost and policy collapse are the main barriers to putting learned signal control in the field, and the paper attacks both with a single architecture.

What carries the argument

The load-bearing object is the multi-scene aggregation matrix: each intersection's local feature vector is oriented and placed into its geographic slot in a tensor whose width and height are grid positions and whose channels are feature values. A CNN-based variational autoencoder compresses this tensor into a compact latent representation, and the PPO value function consumes the concatenation of that representation with the local observation while the policy network improves from local features alone, letting global context guide value estimation without graph attention or neighbor sampling.

What would settle it

Re-run Normal, Peak, and Block with CoLight, DuaLight, and IDQN using their originally reported pressure-based rewards and tuned hyperparameters on the same SUMO setup, and check whether MacLight's wall-clock and final waiting-time advantages persist; alternatively, test MacLight on a non-grid road network to see whether the CNN aggregation matrix assumption breaks.

Watch

Extended reading notes

Core claim

The central discovery is that separating the global scene representation from the local intersection view, and feeding the global code only into the PPO value function while the policy keeps local features, prevents the overfitting and policy collapse seen in DQN-based graph methods. A CNN-based variational autoencoder is trained online to compress the scene matrix into a compact latent code; the value network evaluates the concatenation of that code with the local observation, and training stays stable because PPO clips its updates. On the 4x4 grid, MacLight produces the best average waiting time and queue length in the normal scenario, competitive returns under peak load and road blockages, and wall-clock training of 39 to 58 minutes for 80 episodes, while IDQN needs 137 to 186 minutes and CoLight and DuaLight need 283 to 456 minutes. The authors attribute the speed to convolution's GPU parallelism and the stability to on-policy PPO updates, and they position the dynamic blockage environment as a first step toward emergency traffic scenarios.

Load-bearing premise

The speed-and-stability claim rests on the comparison being fair to the older DQN methods and on a regular 4x4 grid standing in for real road networks; if either gives way, the reported advantage could shrink.

Editorial extensions

If this is right

  • Training times below one hour per scenario on a single A100 make repeated tuning and large-scale grid experiments affordable with the same hardware.
  • On-policy PPO updates avoid the policy collapse that the paper reports for IDQN, CoLight, and DuaLight under sparse rewards, opening learned signal control to high-pressure and emergency conditions.
  • Because global context enters only the value evaluation module, the architecture keeps the parallel-friendly convolutional computation and avoids the sequential neighbor aggregation of graph models.
  • The random road-blockage and rerouting procedure gives a concrete stress test for whether a traffic-signal policy can absorb sudden changes in traffic distribution.

Reading between the lines

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

  • The paper does not ablate the VAE against a cheaper compressor, so a reader could test whether simple pooling of the scene matrix preserves the speed and stability gains; that would isolate whether the latent compression or the PPO backbone is the actual driver.
  • All algorithms in the comparison use the waiting-time reward, while CoLight and DuaLight were originally published with pressure-based rewards; re-running those baselines with their native rewards is the natural check on whether the reported margin is a property of MacLight or of the reward choice.
  • The grid assumption limits immediate transfer to non-Manhattan road networks; a multiscale convolution with imputed missing neighbors is a testable extension the authors themselves point toward.
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 / 10 minor

Summary. The paper proposes MacLight, a traffic signal control method that combines a CNN-based variational autoencoder for global state compression with a PPO backbone for decentralized control, and introduces a dynamic traffic scenario with random road blockages in the SUMO simulator. The method is evaluated against fixed-time, IPPO, MAPPO, IDQN, CoLight, and DuaLight on a 4x4 grid under Normal, Peak, and Block scenarios. The main reported advantages are faster wall-clock training (Table 5) and competitive or better waiting-time/queue/speed metrics in the Normal scenario, with acknowledged inferiority to IPPO in the Block scenario. The paper claims 'superior stability, optimized convergence levels and the highest time efficiency' relative to general and domain SOTA methods, but this claim is weakened by the Block-scenario result and by unresolved questions about baseline configuration and reward selection.

Significance. If the results are reproducible, the wall-clock training advantage in Table 5 is a practically meaningful contribution: reducing training time from several hours to under one hour on the same hardware directly addresses a known bottleneck of graph-attention DQN methods. The dynamic road-blockage scenario is also a useful extension to TSC benchmarking, and the authors provide a code link. The method itself is straightforward and its ablation against IPPO helps isolate the effect of the global representation. However, the headline superiority claims rest on a comparison whose fairness is not established, and the Block-scenario numbers contradict the stated 'superior stability' claim. The paper's contributions are defensible but require substantial additional evidence.

major comments (4)
  1. [§5.3, Table 3] The DQN-based baselines (IDQN, CoLight, DuaLight) report waiting times on the order of 1.6–2.6 million in the Normal scenario, which is 29–46 times worse than the fixed-time baseline (56,409) and 350–550 times worse than MacLight (4,737). Such values are far outside the range reported in the original CoLight/DuaLight papers for comparable grid networks, and strongly suggest a mismatch in reward function, state encoding, or hyperparameter configuration. The paper never states whether these baselines were retrained with the waiting-time reward r = W_{t-1} - W_t defined in §3, nor does it report their hyperparameters, network sizes, or tuning budget. Since the central claim of superiority over domain SOTA depends on these numbers, the authors must provide baseline configuration details, run the baselines with their original pressure-based rewards, or otherwise demonstrate that the comparison is fair. This is a load-bearing issue for the headline claim.
  2. [§5.3, Table 3] The abstract claims MacLight achieves 'superior stability' and 'optimized convergence levels' compared to general and domain SOTA methods, but in the Block scenario the ablation IPPO outperforms MacLight on every reported metric: return -12.0 vs -17.3, waiting time 13,144 vs 24,224, queue 221 vs 249, and speed 5.4 vs 5.2. The text acknowledges 'indicators are inferior to IPPO' in §5.3, yet the abstract and conclusion still assert general superiority. The paper attempts to explain this by saying that average metrics over the whole process hide the final performance, and Table 4 shows near-tie test returns (-1.18 vs -1.17). This is not sufficient: the training-phase metrics are the advertised results, and the dynamic Block scenario is a stated novelty of the paper. The authors should either temper the 'superior stability' claim to exclude Block, provide a mechanistic explanation for why the global representation degrades the dynamic scenario, or modify the method to recover the Block performance.
  3. [§5.1, Table 3] The statistical support for 'superior stability' is weak. With five seeds, the standard deviations are very large: in Peak, MacLight's return is -362.3 ± 423 while IPPO is -434.8 ± 451, and in Normal the returns are -4.02 ± 10 versus -6.6 ± 22. No significance tests or confidence intervals are reported. The visual stability claim from Fig. 7 is also not quantified. The authors should report per-seed results and statistical tests (e.g., matched-pair comparisons across the same seeds) to substantiate the stability claim, especially where the mean differences are smaller than the standard deviations.
  4. [§3, Table 1] The reward function is selected based on experiments in the 'ingolstadt21' scenario, which the authors state is 'completely different' from their own 4x4 grid, and the selection is made using only IPPO. This leaves open the possibility that the chosen waiting-time reward is specifically favorable to PPO-style methods and unfavorable to DQN-based baselines. The paper does not report reward ablation on the actual Normal/Peak/Block scenarios, nor does it compare against the pressure reward used by CoLight/DuaLight. Given that the baseline comparison is the basis for the SOTA claim, the reward choice is a free parameter that should be justified on the actual test scenarios, or the authors should show that the baselines perform comparably under their original reward.
minor comments (10)
  1. [Eq. (4)] The text says 'We first give the process of upsampling' but Eq. (1) describes the encoder (downsampling) path. The wording should be corrected to 'downsampling'.
  2. [Abstract] The notation 'ConvTrans33_3' is ambiguous; it likely means a transposed convolution with 33 output channels and kernel size 3, but the subscript/superscript convention is inconsistent with Eq. (1). Please define the notation unambiguously.
  3. [§4.1 and §6] There is a typo: 'domian' should be 'domain'.
  4. [Table 3] The grid-based limitation is acknowledged in the conclusion but not in the abstract. Since the abstract claims general 'multi-scene aggregation' without mentioning the regular-grid assumption, the authors should state this limitation more prominently, as it restricts the method to Manhattan-like networks.
  5. [§3] The caption says 'The specific values in the table include the mean of the current column indicator and the standard deviation in brackets,' but the Waiting Time columns in Table 3 do not show standard deviations. Please either include them or explain why they are omitted.
  6. [Appendix B] The local observation vector is said to have four components (one-hot action, switching boolean, vehicle density, waiting density), but the total dimension is later given as 33 in Appendix B. The mapping from these components to 33 is not explained; please provide the exact vector construction.
  7. [§4.4] In the ValueNet parameters, state_dim is listed as 33 and global_emb_dim as 16. If the value input is the concatenation of local and global representations, the input dimension should be 49. Please clarify whether state_dim refers only to the local part or whether the concatenation is handled elsewhere.
  8. [§1] The dynamic scenario description does not specify the duration of the road blockage, the rerouting mechanism (e.g., SUMO rerouter settings), or how many episodes experience the blockage. These details are necessary for reproducibility and for interpreting the Block-scenario results.
  9. [§5.3] The claim that 'none of the current peer studies have deployed dynamic traffic systems for experiments' is stated without a survey citation to support it. Several works have used varying demand patterns or incidents; please soften the claim or provide evidence that dynamic emergency scenarios have not been tested before.
  10. [§5.3] The statement 'if we check the final value, MacLight still has the best performance' in the Peak scenario is not supported by a table or figure that reports final-value metrics; Table 4 only reports test returns, not waiting time/queue/speed at the final episode. Please add explicitly reported final metrics.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: MacLight is an empirical RL systems paper; its claims rest on simulator and wall-clock measurements against independent baselines, not on any derivation that reduces to its own inputs.

full rationale

MacLight is an empirical systems paper. Its contributions are a CNN-based VAE global representation, a PPO backbone, and a dynamic SUMO scenario; the headline claims are supported by measured waiting times, queue lengths, speeds, returns, and wall-clock training times in Tables 3-5. The reward r_i^t = W_i^{t-1} - W_i^t is a standard proxy objective, and reporting average waiting time as an evaluation metric is not a circular derivation: nothing in the paper claims that the metric is logically forced by the reward definition alone. The VAE is trained online with reconstruction and KL losses (Eqs. 5-6) and is not fitted to the downstream evaluation metric. No load-bearing step is justified by a self-citation: the one-hop-neighbor finding attributed to Wei et al. [29] and the CoLight/DuaLight baselines are external prior work, with no author overlap that would make the cited result a renamed version of the present claim. Concerns about the DQN baselines' extremely high waiting times and the absence of baseline tuning or reward details are empirical validity risks, not circularity. The conclusion's admission that real road networks are irregular and cannot be directly written as pixel matrices is an applicability limitation, not a circular step. No equation in the paper reduces a predicted quantity to a fitted input by construction, so the appropriate finding is no significant circularity.

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

The paper introduces no new physical or conceptual entities; its contributions are a new architectural combination and a simulation scenario. The main free parameters are the reward choice and the scenario-specific blockage configuration, both selected by the authors and used for all algorithms.

free parameters (3)
  • reward function = r_t = W_{t-1} - W_t (waiting-time difference)
    Chosen as the best among several tested rewards using IPPO on the ingolstadt21 scenario, then applied to all algorithms in the main experiments.
  • blocked roads in dynamic scenario = D3C3, D3D2, D2C2, C3C2
    Chosen to demonstrate rerouting effects; the selection is a scenario design choice that defines the difficulty of the Block benchmark.
  • fixed-time baseline switching interval = 45 seconds
    Used as the non-RL baseline; a hand-set parameter that influences all comparisons.
assumptions (3)
  • domain assumption Road networks can be approximated as regular pixel grids for CNN aggregation.
    Invoked in Section 4.1 to justify the aggregation matrix; the authors concede in the conclusion that real networks are not regular.
  • domain assumption The SUMO simulator faithfully models traffic dynamics so that RL policies trained in simulation transfer to the stated scenarios.
    Implicit in all experiments; the paper does not validate against real-world data.
  • domain assumption Waiting-time-based reward is an appropriate proxy for the traffic control objective across all algorithms.
    Selected in Section 3 over pressure, queue, and speed rewards using IPPO on an external scenario.

how reviews work

0 comments
Cite this review

Pith. "Pith review of MacLight: Multi-scene Aggregation Convolutional Learning for Traffic Signal Control." pith.science (2026). https://pith.science/paper/NWZIQWNC

@misc{pith2026241215703,
  author       = {Pith},
  title        = {Pith review of: MacLight: Multi-scene Aggregation Convolutional Learning for Traffic Signal Control},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NWZIQWNC}},
  note         = {Machine review of arXiv:2412.15703}
}
read the original abstract

Reinforcement learning methods have proposed promising traffic signal control policy that can be trained on large road networks. Current SOTA methods model road networks as topological graph structures, incorporate graph attention into deep Q-learning, and merge local and global embeddings to improve policy. However, graph-based methods are difficult to parallelize, resulting in huge time overhead. Moreover, none of the current peer studies have deployed dynamic traffic systems for experiments, which is far from the actual situation. In this context, we propose Multi-Scene Aggregation Convolutional Learning for traffic signal control (MacLight), which offers faster training speeds and more stable performance. Our approach consists of two main components. The first is the global representation, where we utilize variational autoencoders to compactly compress and extract the global representation. The second component employs the proximal policy optimization algorithm as the backbone, allowing value evaluation to consider both local features and global embedding representations. This backbone model significantly reduces time overhead and ensures stability in policy updates. We validated our method across multiple traffic scenarios under both static and dynamic traffic systems. Experimental results demonstrate that, compared to general and domian SOTA methods, our approach achieves superior stability, optimized convergence levels and the highest time efficiency. The code is under https://github.com/Aegis1863/MacLight.

Figures

Figures reproduced from arXiv: 2412.15703 by the authors.

Figure 1
Figure 1. General right-hand 2-way 6-lane intersection with [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. MacLight framework. The first row shows how to construct the aggregation matrix, and the second row introduces [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. The road network of the simulation environment [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (5 more)
Figure 6
Figure 6. Figure 6: Statistics of different experimental scenarios [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]
Figure 5
Figure 5. Figure 5: Some regular road traffic flow distributions [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 7
Figure 7. Figure 7: Training details of cumulative rewards CoLight. Referring to [29], a strong algorithm for applying RL to TSC tasks using GAT, built on top of DQN. DuaLight. Reference [15], a SOTA based on CoLight, adds fea￾ture weight matrix and neighborhood weight matrix for differen…
Figure 1
Figure 1. Figure 1: Complete statistics on the changes in the distribution of traffic flow before and after the implementation of road [PITH_FULL_IMAGE:figures/full_fig_p010_1.png]
Figure 2
Figure 2. Figure 2: All experimental results. For the Return and Speed indicators, the larger, the better; for the Queue and Waiting [PITH_FULL_IMAGE:figures/full_fig_p012_2.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

36 extracted references · 17 canonical work pages

  1. [1]

    Lucas N. Alegre. 2019. SUMO-RL. https://github.com/LucasAlegre/sumo-rl

    work page 2019

  2. [2]

    Alegre, Ana L

    Lucas N. Alegre, Ana L. C. Bazzan, and Bruno C. da Silva. 2021. Quantifying the impact of non-stationarity in reinforcement learning-based traffic signal control. PeerJ Computer Science 7 (2021), e575. https://doi.org/10.7717/peerj-cs.575

    work page doi:10.7717/peerj-cs.575 2021

  3. [3]

    James Ault and Guni Sharon. 2021. Reinforcement Learning Benchmarks for Traffic Signal Control. In Proceedings of the Thirty-fifth Conference on Neural Information Processing Systems (NeurIPS 2021) Datasets and Benchmarks Track

    work page 2021

  4. [4]

    Chacha Chen, Hua Wei, Nan Xu, Guanjie Zheng, Ming Yang, Yuanhao Xiong, Kai Xu, and Zhenhui Li. 2020. Toward A Thousand Lights: Decentralized Deep Reinforcement Learning for Large-Scale Traffic Signal Control. Proceedings of the AAAI Conference on Artificial Intelligence 34, 04 (Apr. 2020), 3414–3421. https://doi.org/10.1609/aaai.v34i04.5744

    doi:10.1609/aaai.v34i04.5744 2020

  5. [5]

    Christian Schroeder de Witt, Tarun Gupta, Denys Makoviichuk, Viktor Makoviy- chuk, Philip H. S. Torr, Mingfei Sun, and Shimon Whiteson. 2020. Is In- dependent Learning All You Need in the StarCraft Multi-Agent Challenge? arXiv:2011.09533 [cs.AI] https://arxiv.org/abs/2011.09533

    arXiv 2020

  6. [6]

    Christina Diakaki, Markos Papageorgiou, and Kostas Aboudolas. 2002. A multi- variable regulator approach to traffic-responsive network-wide signal control. Control Engineering Practice 10, 2 (2002), 183–195. https://doi.org/10.1016/S0967- 0661(01)00121-6

    work page doi:10.1016/s0967- 2002

  7. [7]

    Carlos Gershenson. 2005. Self-Organizing Traffic Lights. arXiv:nlin/0411066 [nlin.AO] https://arxiv.org/abs/nlin/0411066

    arXiv 2005

  8. [8]

    Sepp Hochreiter and Jürgen Schmidhuber. 1997. Long Short-Term Memory. Neural Comput. 9, 8 (Nov. 1997), 1735–1780. https://doi.org/10.1162/neco.1997.9. 8.1735

    doi:10.1162/neco.1997.9 1997

Show all 36 references
  1. [9]

    Industrial Solution

    Haoyuan Jiang, Xuantang Xiong, Ziyue Li, Hangyu Mao, Guanghu Sui, Jingqing Ruan, Yuheng Cheng, Hua Wei, Wolfgang Ketter, and Rui Zhao. 2024. GuideLight: "Industrial Solution" Guidance for More Practical Traffic Signal Control Agents. arXiv:2407.10811 [cs.MA] https://arxiv.org/...

    2024 arXiv

  2. [10]

    Kingma and Max Welling

    Diederik P. Kingma and Max Welling. 2014. Auto-Encoding Variational Bayes. In 2nd International Conference on Learning Representations, ICLR 2014, Banff, AB, Canada, April 14-16, 2014, Conference Track Proceedings , Yoshua Bengio and Yann LeCun (Eds.). http://arxiv.org/abs/1312.6114

    2014 arXiv

  3. [11]

    Alex Krizhevsky, Ilya Sutskever, and Geoffrey E Hinton. 2012. ImageNet Clas- sification with Deep Convolutional Neural Networks. In Advances in Neural Information Processing Systems, F. Pereira, C.J. Burges, L. Bottou, and K.Q. Wein- berger (Eds.), Vol. 25. Curran Associates, Inc

    2012

  4. [12]

    Colin Lea, René Vidal, Austin Reiter, and Gregory D. Hager. 2016. Temporal Convolutional Networks: A Unified Approach to Action Segmentation. CoRR abs/1608.08242 (2016). arXiv:1608.08242

    2016 arXiv

  5. [13]

    Pablo Alvarez Lopez, Michael Behrisch, Laura Bieker-Walz, Jakob Erdmann, Yun- Pang Flötteröd, Robert Hilbrich, Leonhard Lücken, Johannes Rummel, Peter Wagner, and Evamarie Wießner. 2018. Microscopic Traffic Simulation using SUMO, In The 21st IEEE International Conference on In...

    2018

  6. [14]

    Yican Lou, Jia Wu, and Yunchuan Ran. 2022. Meta-Reinforcement Learning for Multiple Traffic Signals Control. In Proceedings of the 31st ACM International Conference on Information & Knowledge Management (Atlanta, GA, USA) (CIKM ’22). Association for Computing Machinery, New Yo...

    2022

  7. [15]

    Jiaming Lu, Jingqing Ruan, Haoyuan Jiang, Ziyue Li, Hangyu Mao, and Rui Zhao

  8. [16]

    Jinming Ma and Feng Wu. 2020. Feudal Multi-Agent Deep Reinforcement Learn- ing for Traffic Signal Control. InProceedings of the 19th International Conference on Autonomous Agents and MultiAgent Systems (Auckland, New Zealand) (AAMAS ’20). International Foundation for Autonomou...

    2020

  9. [17]

    Volodymyr Mnih, Koray Kavukcuoglu, David Silver, Alex Graves, Ioannis Antonoglou, Daan Wierstra, and Martin Riedmiller. 2013. Playing Atari with Deep Reinforcement Learning. arXiv:1312.5602 [cs.LG]

    2013 arXiv

  10. [18]

    Afshin Oroojlooy, Mohammadreza Nazari, Davood Hajinezhad, and Jorge Silva

  11. [19]

    John Schulman, Filip Wolski, Prafulla Dhariwal, Alec Radford, and Oleg Klimov

  12. [20]

    Joar Skalse, Nikolaus H. R. Howe, Dmitrii Krasheninnikov, and David Krueger

  13. [21]

    Richard S. Sutton. 1988. Learning to predict by the methods of temporal dif- ferences. Machine Learning 3, 1 (01 Aug 1988), 9–44. https://doi.org/10.1007/ BF00115009

    1988

  14. [22]

    Richard S Sutton and Andrew G Barto. 2018. Reinforcement learning: An intro- duction. MIT press

    2018

  15. [23]

    Sutton, David McAllester, Satinder Singh, and Yishay Mansour

    Richard S. Sutton, David McAllester, Satinder Singh, and Yishay Mansour. 1999. Policy Gradient Methods for Reinforcement Learning with Function Approxi- mation. In Proceedings of the 12th International Conference on Neural Information Processing Systems (Denver, CO) (NIPS’99)....

    1999

  16. [24]

    In Proceedings of the 36th International Conference on Neural Information Processing Systems (New Orleans, LA, USA) (NIPS ’22)

    Defining and characterizing reward hacking. In Proceedings of the 36th International Conference on Neural Information Processing Systems (New Orleans, LA, USA) (NIPS ’22). Curran Associates Inc., Red Hook, NY, USA, Article 687, 12 pages

  17. [25]

    Petar Veličković, Guillem Cucurull, Arantxa Casanova, Adriana Romero, Pietro Liò, and Yoshua Bengio. 2018. Graph Attention Networks. International Confer- ence on Learning Representations (2018)

    2018

  18. [26]

    Y. Wang, T. Xu, X. Niu, C. Tan, E. Chen, and H. Xiong. 2022. STMARL: A Spatio-Temporal Multi-Agent Reinforcement Learning Approach for Cooperative Traffic Light Control. IEEE Transactions on Mobile Computing 21, 06 (jun 2022), 2228–2242. https://doi.org/10.1109/TMC.2020.3033782

    2022

  19. [27]

    Christopher J. C. H. Watkins and Peter Dayan. 1992. Q-learning. Machine Learning 8, 3 (01 May 1992), 279–292. https://doi.org/10.1007/BF00992698

    1992 doi

  20. [28]

    Pravin Varaiya. 2013. The Max-Pressure Controller for Arbitrary Networks of Signalized Intersections. Springer New York, New York, NY, 27–66. https://doi. org/10.1007/978-1-4614-6243-9_2

    2013 doi

  21. [29]

    Hua Wei, Nan Xu, Huichu Zhang, Guanjie Zheng, Xinshi Zang, Chacha Chen, Weinan Zhang, Yanmin Zhu, Kai Xu, and Zhenhui Li. 2019. CoLight: Learning Network-level Cooperation for Traffic Signal Control. In Proceedings of the 28th ACM International Conference on Information and Kn...

    2019

  22. [30]

    Libing Wu, Min Wang, Dan Wu, and Jia Wu. 2021. DynSTGAT: Dynamic Spatial- Temporal Graph Attention Network for Traffic Signal Control. In Proceedings of the 30th ACM International Conference on Information & Knowledge Management (Virtual Event, Queensland, Australia) (CIKM ’21...

    2021 doi

  23. [31]

    Chao Yu, Akash Velu, Eugene Vinitsky, Jiaxuan Gao, Yu Wang, Alexandre Bayen, and Yi Wu. 2022. The Surprising Effectiveness of PPO in Cooperative Multi- Agent Games. In Thirty-sixth Conference on Neural Information Processing Systems Datasets and Benchmarks Track

    2022

  24. [32]

    Hua Wei, Chacha Chen, Guanjie Zheng, Kan Wu, Vikash Gayah, Kai Xu, and Zhenhui Li. 2019. PressLight: Learning Max Pressure Control to Coordinate Traffic Signals in Arterial Network. In Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mi...

    2019

  25. [36]

    Dongbin Zhao, Yujie Dai, and Zhen Zhang. 2012. Computational Intelligence in Urban Traffic Signal Control: A Survey. IEEE Transactions on Systems, Man, and Cybernetics, Part C (Applications and Reviews) 42, 4 (2012), 485–494. https: //doi.org/10.1109/TSMCC.2011.2161577 A DYNAM...

    2012

  26. [2017]

    arXiv:1707.06347 [cs.LG]

    Proximal Policy Optimization Algorithms. arXiv:1707.06347 [cs.LG]

    arXiv

  27. [2020]

    In Proceedings of the 34th International Conference on Neural Information Processing Systems (Vancouver, BC, Canada)(NIPS ’20)

    AttendLight: universal attention-based reinforcement learning model for traffic signal control. In Proceedings of the 34th International Conference on Neural Information Processing Systems (Vancouver, BC, Canada)(NIPS ’20). Curran Associates Inc., Red Hook, NY, USA, Article 34...

  28. [2024]

    In Proceedings of the 23rd International Confer- ence on Autonomous Agents and Multiagent Systems (Auckland, New Zealand) (AAMAS ’24)

    DuaLight: Enhancing Traffic Signal Control by Leveraging Scenario-Specific and Scenario-Shared Knowledge. In Proceedings of the 23rd International Confer- ence on Autonomous Agents and Multiagent Systems (Auckland, New Zealand) (AAMAS ’24). International Foundation for Autonom...

Pith tools

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