Pith. sign in

REVIEW 4 major objections 5 minor 55 references

Multi-agent DRL-based Lane Change Decision Model for Cooperative Platooning in Mixed Traffic

T0 review · 4 major / 5 minor · reviewed 2026-08-03 · deepseek-v4-flash

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

Pith's one-line read A CNN-augmented QMIX lane-change policy forms CAV platoons in mixed traffic, beating rule-based baselines by up to 26.2%.

desk verdict Plausible idea for variable-size QMIX, but the write-up is inconsistent and the empirical support is under-reported; worth a close look after major fixes. read the letter →

arxiv 2601.11809 v2 pith:VLS5VZYW submitted 2026-01-16 cs.AI

Zeyu Mu , Shangtong Zhang , B. Brian Park This is my paper
classification cs.AI
keywords multi-agentreinforcementlearningQMIXconvolutionalneuralnetworklane-changedecisioncooperativeplatooningmixedtrafficconnectedautomatedvehiclesmarketpenetrationrate
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 single multi-agent reinforcement-learning model can make lane-change decisions for connected automated vehicles (CAVs) even when the number of CAVs on the road changes over time and across market penetration rates. The proposed CNN-QMIX model replaces QMIX's fixed-size mixing network with a convolutional network that reads a spatial grid of all vehicles, so the architecture no longer depends on a preset agent count. In microscopic simulation on a three-lane, 1200-meter highway segment, the trained model forms cooperative platoons at a higher rate than the MOBIL rule-based model, a greedy platoon-assignment heuristic, and per-penetration-rate QMIX models. The reported improvements in platoon rate reach 26.2% over the greedy baseline at a 37.5% CAV market penetration rate, while average speed rises and energy consumption falls. A sympathetic reading is that learning-based coordination can help jump-start platooning in the sparse, early-deployment phase of CAVs.

What carries the argument

The central object is CNN-QMIX: QMIX is a multi-agent deep reinforcement-learning algorithm that trains individual Q-networks centrally and executes them decentrally, with a mixing network enforcing a monotonicity condition that lets each agent greedily select its own action. The paper's modification feeds the mixing network a convolutional encoding of a grid-based traffic snapshot (three channels: longitudinal position, speed, vehicle type) instead of an ordered list of Q-values. That grid representation is what is supposed to make mixing-network processing independent of agent count. A quintic-polynomial trajectory planner and a model predictive controller carry out the lane changes that t

What would settle it

Compute the partial derivative ∂Q_total/∂Q_i of the trained CNN-QMIX mixing network across varied agent counts and states; if it is ever negative, the decentralized greedy action selection no longer follows from QMIX's theory. Alternatively, run the trained model in a simulator with human-driver data or on-ramp traffic that changes the agent count mid-episode and test whether platoon rates and safety hold.

Watch

Extended reading notes

Core claim

The authors' claim is that changing the input and mixing structure of QMIX — encoding traffic as a grid with channels for position, speed, and vehicle type and processing it with convolutional layers — makes cooperative lane-change decisions invariant to the number of agents. Under this design, each CAV still learns an individual Q-function, but the joint Q-value is formed by a CNN-based mixing network that can consume a variable-sized scene, rather than a fixed list of per-agent Q-values. The paper reports that this single model, trained once, achieves platoon rates of 0.21, 0.53, and 0.72 at CAV market penetration rates of 12.5%, 37.5%, and 50%, outperforming per-penetration QMIX models an

Load-bearing premise

The load-bearing premise is that replacing QMIX's fixed-size mixing network with a CNN that consumes a spatial grid of all vehicles preserves the monotonicity condition that justifies decentralized execution, and that the MOBIL/IDM simulation of human drivers faithfully represents real mixed traffic.

Editorial extensions

If this is right

  • A single trained CNN-QMIX policy can be deployed across CAV market penetration rates without retraining separate models for each rate.
  • CAVs using the learned policy form longer and more frequent cooperative platoons than rule-based MOBIL or greedy assignment, with improvements up to 26.2% in platoon rate.
  • Higher platoon participation is associated with higher average traffic speed and lower simulated energy consumption compared with rule-based baselines.
  • Because execution is decentralized, individual CAVs only need their local observations at deployment, while coordination is learned during centralized training.
  • The reward structure — an explicit platoon bonus plus speed and safety terms — implies that lane changes should sacrifice short-term speed gains if they would break an existing platoon.

Reading between the lines

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

  • Editorial inference: The same grid-based architecture should in principle transfer to agent counts outside the training range, such as a 24-vehicle highway with an on-ramp adding vehicles mid-episode; the paper does not test this directly, and it is the strongest untested consequence.
  • Editorial inference: The reported benefit is small at 12.5% MPR (platoon rate 0.21 vs 0.16 for greedy), so the model's practical value may concentrate at moderate penetration rather than the sparsest early-deployment regimes.
  • Editorial inference: Because the paper never verifies that the CNN mixing network preserves QMIX's monotonicity condition when the agent count changes, the measured gains could partly come from the CNN's spatial feature extraction rather than from the cooperative-decomposition guarantee; an ablation separating CNN features from QMIX mixing would clarify this.
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 a hierarchical lane-change control framework for connected automated vehicles (CAVs) in mixed traffic, combining a multi-agent deep reinforcement learning decision module (CNN-QMIX) with a trajectory planner and model predictive controller. The central claim is that CNN-QMIX can handle a varying number of CAV agents without retraining for each market penetration rate (MPR), and that a reward term explicitly encourages cooperative platooning. In a 24-vehicle, three-lane, 1200-m microsimulation with MPRs of 12.5%, 37.5%, and 50%, the authors report that CNN-QMIX achieves higher platoon rates than per-MPR QMIX and than MOBIL and greedy rule-based baselines, with platoon-rate improvements up to 26.2%, as well as better average speed and lower energy consumption. The paper also describes the baselines, simulation setup, and training details.

Significance. The problem is relevant: in the early deployment of CAVs, sparse connectivity makes cooperative platooning difficult, and an adaptive lane-change policy could yield practical benefits. The paper has a sensible modular design—high-level DRL decision-making, a quintic-polynomial trajectory planner, and an MPC tracker—and it evaluates against both per-MPR QMIX and two rule-based methods. If the QMIX-based cooperative mechanism were confirmed and the empirical gains were statistically robust, this would be a useful applied contribution. However, as written, the manuscript does not establish that the CNN-QMIX variant is a principled QMIX method, and the reported platoon lengths imply that the platooning reward term is inactive. The claimed mechanism and the evidence therefore need substantial work before the conclusions can be accepted.

major comments (4)
  1. [§II-B.2, Eq. (1)] QMIX's decentralized execution is justified only by the individual-global-max (IGM) condition, which follows from Eq. (1) when Q_total is monotonically increasing in each per-agent Q_i. The paper states that CNN-QMIX replaces the traditional fixed-size mixing network with a CNN over a grid-based state representation, 'instead of using a list of individual Q-values as input.' If the mixing network consumes raw vehicle states, then monotonicity in per-agent Q-values is undefined and IGM is not guaranteed; no non-negative-weight architecture is specified. If Q-values are still fed into the mixing network (as the next sentence states), then the modification is in the observation encoder rather than the mixer, and the description should say so. The authors must either provide the exact architecture and show that IGM is preserved, or reposition the method as a heuristic extension. As written,
  2. [Eq. (7), Tables IV–V] The cooperative-platooning reward is r_c = max(0, log2(n/10)), where n is the number of connected vehicles ahead of the agent. All reported maximum platoon lengths in Tables IV and V are between 1.1 and 4.9, so n < 10 in every reported setting and r_c = 0. Thus the reward term explicitly designed to foster platooning never activates in the experiments. The conclusion in §V that the reward function 'explicitly incorporates elements of cooperative platooning' is therefore not supported by the presented data. The authors should report the distribution of n during training and testing, or correct the reward/threshold definition.
  3. [§IV-C, Tables IV–V] The evaluation is based on 100 test runs, but the tables report only point estimates. No standard deviations, confidence intervals, or statistical tests are given, so the word 'significantly' in the abstract and text is not justified. The narrative percentages also disagree with Table V: at 12.5% MPR, 0.21 versus 0.16 is a 31.3% increase, not 21.3%; at 50% MPR, 0.72 versus 0.62 is 16.1%, not 26.1%; and the performance discussion refers to 'MPR of 35.5%' instead of 37.5%. These inconsistencies must be corrected.
  4. [§III, §IV-C] The paper's headline contribution is handling 'fluctuating traffic agent numbers,' but the microsimulation is a fixed 24-vehicle, 1200-m segment without vehicles entering, exiting, or dropping out of communication range. The varying-MPR comparison trains one model across three MPR values and evaluates separately at each MPR; it does not demonstrate within-episode variation in the number of agents. If the contribution is a single model across MPRs, the claim should be stated more narrowly. If real-time fluctuation is claimed, an experiment in which agents enter or leave the sensing/communication range during an episode is needed.
minor comments (5)
  1. [Throughout] There are numerous typos and grammatical errors, e.g., 'utlize,' 'genearely,' 'raod,' 'safty,' 'comparisions,' and the header title says 'Cooperative Planning' while the abstract says 'Cooperative Platooning.' A careful proofread is needed.
  2. [Eq. (5)] Equation (5) has unbalanced parentheses in the definition of r_d, and the symbol r is used both for the reward and for the exponential decay constant. Please clarify the notation.
  3. [Eq. (9)] In the trajectory cost function, a_2 is described as 'control effort (acceleration),' but a_2 is also a coefficient of the quintic polynomial. This is confusing and should be renamed.
  4. [Table V] The 'Time' column for MOBIL is listed as 'N/A' without explanation, even though text discusses platoon-formation time. Please explain or remove.
  5. [General] No code or data availability statement is provided. Given the ambiguity in the architecture description, releasing code or pseudocode would materially improve reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: headline claims are empirical comparisons against external baselines; the CNN-QMIX monotonicity gap is a correctness/support issue, not a circular derivation.

full rationale

The paper's central claims are empirical: after training, CNN-QMIX yields the platoon rates in Tables IV-VI, compared with MOBIL [51], greedy [19], and per-MPR QMIX baselines. These baselines are external and the metrics are direct simulation outputs, so there is no fitted parameter renamed as a prediction and no self-referential loop. The reward function does include a platooning incentive (Eq. 7, r_c = max(0, log2(n/10))) and the headline metric is 'platoon rate', but the paper does not derive the measured platoon rate from the reward; it reports the optimized policy's behavior. Optimizing an objective and then measuring the same construct is an evaluation loop, not a circular derivation, and the claimed contribution is the margin over baselines that do not use this reward. The main concern in Section II-B.2 is that the paper replaces 'the traditional fixed-size mixing network' with a grid-based CNN yet later states 'Q-values are fed into the mixing network', and it does not prove that QMIX's monotonicity/IGM guarantee survives; this is an omitted proof / architectural coherence problem, not one of the enumerated circularity patterns. No load-bearing self-citation is present; references used for baselines and human-driving parameters are external. Thus the derivation chain is self-contained for circularity purposes, and the score is 0.

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

No new physical or conceptual entities are postulated: CNN-QMIX rearranges existing components (QMIX plus a CNN grid encoder), and the platooning reward is a numeric function, not an entity. The ledger is dominated by hand-tuned or unspecified numeric parameters (reward weights, decay constants, controller weights, baseline parameters) and by domain assumptions about the realism of IDM/MOBIL human-driver modeling and the acceleration-integral energy proxy. The discount factor and network layout are also free choices.

free parameters (6)
  • Reward weights w1, w2, w3 = 1, 0.5, 2
    Hand-tuned via empirical testing (Table II); these set the relative priority of platooning, speed, and safety rewards, and the paper states in §V that parameters are configured manually.
  • Reward decay constants r (Eq. 5) and m (Eq. 6) = unspecified
    Constants in the exponential safety and speed rewards; values are never given, so the reward shaping cannot be replicated.
  • Discount factor gamma = 0.5
    Table II; unusually low for a planning task with a multi-second horizon, and it strongly shapes the learned policy's short-sightedness.
  • Neural network layout (CNN filters, FC units, strides) = 16/32/16 filters; 128/64 FC units
    Table III; manually tuned through grid-based exploration per §III-D, not derived from any principle.
  • Greedy baseline parameters alpha, m, r = unspecified
    Weights and tolerances in Eqs (15)–(17) of the baseline assignment method; values are omitted, leaving the baseline under-specified.
  • Planner and MPC weights c1-c3, Q, R = unspecified
    Cost weights in Eqs (9)–(10); values not reported, so the execution layer cannot be reproduced.
assumptions (5)
  • domain assumption Human-driven vehicles are adequately modeled by IDM longitudinal control (Table I) and MOBIL lane changes (Eq. 12) with parameters from refs [51], [55].
    The entire comparison rests on this model of human behavior; its realism determines the significance of the reported improvements.
  • domain assumption The kinematic bicycle model with first-order acceleration lag (Eq. 11) and the printed A/B matrices represent CAV dynamics for the MPC layer.
    Used in §II-C; as printed, the A and B matrices are dimensionally inconsistent (rows with 5, 6, and 2–3 entries), so the control layer is not verifiable as written.
  • domain assumption Energy consumption is measured as the time-integral of acceleration over the run (§IV-A).
    A proxy, not a physical energy model (no mass, drag, or powertrain); energy-saving claims inherit this limitation.
  • standard math QMIX convergence guarantees (monotonic mixing, CTDE, TD update, Eqs 1–2) carry over to the traffic environment and to the CNN-grid modification.
    Invoked without proof as background; the extension to a grid-based mixing input is asserted, not derived.
  • domain assumption The custom microsimulation (24 vehicles, 1200-m segment, random spawn in [100, 300] m) is a faithful proxy for real mixed traffic.
    §III-D. All results are simulation-internal; no validation against field data is presented.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Multi-agent DRL-based Lane Change Decision Model for Cooperative Platooning in Mixed Traffic." pith.science (2026). https://pith.science/paper/VLS5VZYW

@misc{pith2026260111809,
  author       = {Pith},
  title        = {Pith review of: Multi-agent DRL-based Lane Change Decision Model for Cooperative Platooning in Mixed Traffic},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VLS5VZYW}},
  note         = {Machine review of arXiv:2601.11809}
}
read the original abstract

Connected automated vehicles (CAVs) possess the ability to communicate and coordinate with one another, enabling cooperative platooning that enhances both energy efficiency and traffic flow. However, during the initial stage of CAV deployment, the sparse distribution of CAVs among human-driven vehicles reduces the likelihood of forming effective cooperative platoons. To address this challenge, this study proposes a hybrid multi-agent lane change decision model aimed at increasing CAV participation in cooperative platooning and maximizing its associated benefits. The proposed model employs the QMIX framework, integrating traffic data processed through a convolutional neural network (CNN-QMIX). This architecture addresses a critical issue in dynamic traffic scenarios by enabling CAVs to make optimal decisions irrespective of the varying number of CAVs present in mixed traffic. Additionally, a trajectory planner and a model predictive controller are designed to ensure smooth and safe lane-change execution. The proposed model is trained and evaluated within a microsimulation environment under varying CAV market penetration rates. The results demonstrate that the proposed model efficiently manages fluctuating traffic agent numbers, significantly outperforming the baseline rule-based models. Notably, it enhances cooperative platooning rates up to 26.2\%, showcasing its potential to optimize CAV cooperation and traffic dynamics during the early stage of deployment.

Figures

Figures reproduced from arXiv: 2601.11809 by the authors.

Figure 1
Figure 1. Lane Change Decision for Speed Incentive and Sabotaging Cooper [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 3
Figure 3. The Architecture of the Multi-agent DRL-based Lane Change Decision [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figure 4
Figure 4. The Input Process for the Multi-agent DRL-based Lane Change [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figures from the paper (2 more)
Figure 5
Figure 5. Figure 5: Microsimulation Framework from longitudinal and lateral vehicle controls of commanded acceleration ua and steering angle uδ are collected as vehicle input. The matrices A and B are given based on the vehicle model, defined as, A =       0 0 0 1 0 0 0 v0 0 0 0 0 0…
Figure 6
Figure 6. Figure 6: Training results of Multi-agent DRL-based lane change decision [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

55 extracted references · 1 canonical work pages

  1. [1]

    Performance measurement evaluation framework and co-benefittradeoff analysis for connected and automated vehicles (cav) applications: A survey,

    D. Tian, G. Wu, K. Boriboonsomsin, and M. J. Barth, “Performance measurement evaluation framework and co-benefittradeoff analysis for connected and automated vehicles (cav) applications: A survey,”IEEE Intelligent Transportation Systems Magazine, vol. 10, no. 3, pp. 110– 122, 2018

    2018

  2. [2]

    A review of communication, driver characteristics, and controls aspects of cooperative adaptive cruise con- trol (cacc),

    K. C. Dey, L. Yan, X. Wang, Y . Wang, H. Shen, M. Chowdhury, L. Yu, C. Qiu, and V . Soundararaj, “A review of communication, driver characteristics, and controls aspects of cooperative adaptive cruise con- trol (cacc),”IEEE Transactions on Intelligent Transportation Systems, vol. 17, no. 2, pp. 491–509, 2016

    2016

  3. [3]

    Aerodynamic drag analysis of autonomous electric vehicle platoons,

    S. T. Kaluva, A. Pathak, and A. Ongel, “Aerodynamic drag analysis of autonomous electric vehicle platoons,”Energies, vol. 13, no. 15, 2020. [Online]. Available: https://www.mdpi.com/1996-1073/13/15/4028

    2020

  4. [4]

    Cooperative connected autonomous vehicles (cav): Research, applications and challenges,

    J. He, Z. Tang, X. Fu, S. Leng, F. Wu, K. Huang, J. Huang, J. Zhang, Y . Zhang, A. Radford, L. Li, and Z. Xiong, “Cooperative connected autonomous vehicles (cav): Research, applications and challenges,” in 2019 IEEE 27th International Conference on Network Protocols (ICNP), 2019, pp. 1–6

    2019

  5. [5]

    Unravelling effects of cooperative adaptive cruise control deactivation on traffic flow characteristics at merging bottlenecks,

    L. Xiao, M. Wang, W. Schakel, and B. van Arem, “Unravelling effects of cooperative adaptive cruise control deactivation on traffic flow characteristics at merging bottlenecks,”Transportation Research Part C: Emerging Technologies, vol. 96, pp. 380–397, 2018. [Online]. Available: https://www.sciencedirect.com/science/article/pii/S0968090X1830528X

    2018

  6. [6]

    Modeling impacts of cooperative adaptive cruise control on mixed traffic flow in multi-lane freeway facilities,

    H. Liu, X. D. Kan, S. E. Shladover, X.-Y . Lu, and R. E. Ferlis, “Modeling impacts of cooperative adaptive cruise control on mixed traffic flow in multi-lane freeway facilities,”Transportation Research Part C: Emerging Technologies, vol. 95, pp. 261–279, 2018. [Online]. Available: https://www.sciencedirect.com/science/article/pii/S0968090X18310313

    2018

  7. [7]

    Influence of exclusive lanes for connected and autonomous vehicles on freeway traffic flow,

    K. Ma and H. Wang, “Influence of exclusive lanes for connected and autonomous vehicles on freeway traffic flow,”IEEE Access, vol. 7, pp. 50 168–50 178, 2019

    2019

  8. [8]

    Flow-aware platoon formation of connected automated vehicles in a mixed traffic with human-driven vehicles,

    S. Woo and A. Skabardonis, “Flow-aware platoon formation of connected automated vehicles in a mixed traffic with human-driven vehicles,”Transportation Research Part C: Emerging Technologies, vol. 133, p. 103442, 2021. [Online]. Available: https://www.sciencedirect. com/science/article/pii/S0968090X21004307

    2021

Show all 55 references
  1. [9]

    Impact of cav platoon management on traffic flow considering degradation of control mode,

    L. Zhou, T. Ruan, K. Ma, C. Dong, and H. Wang, “Impact of cav platoon management on traffic flow considering degradation of control mode,”Physica A: Statistical Mechanics and its Applications, vol. 581, p. 126193, 2021. [Online]. Available: https://www.sciencedirect.com/ scien...

    2021

  2. [10]

    Design and operation of dedicated lanes for connected and automated vehicles on motorways: A conceptual framework and research agenda,

    S. Razmi Rad, H. Farah, H. Taale, B. van Arem, and S. P. Hoogendoorn, “Design and operation of dedicated lanes for connected and automated vehicles on motorways: A conceptual framework and research agenda,”Transportation Research Part C: Emerging Technologies, vol. 117, p. 102...

    2020

  3. [11]

    Leveraging dynamic right-of- way allocation and tolling policy for cav dedicated lane management to promote cav and improve mobility,

    H. Chen, F. Wu, K. Hou, and T. Z. Qiu, “Leveraging dynamic right-of- way allocation and tolling policy for cav dedicated lane management to promote cav and improve mobility,”IEEE Transactions on Intelligent Transportation Systems, pp. 1–10, 2024

    2024

  4. [12]

    Optimal lane management policy for connected automated vehicles in mixed traffic flow,

    Z. Yao, L. Li, W. Liao, Y . Wang, and Y . Wu, “Optimal lane management policy for connected automated vehicles in mixed traffic flow,”Physica A: Statistical Mechanics and its Applications, vol. 637, p. 129520, 2024. [Online]. Available: https: //www.sciencedirect.com/science/a...

    2024

  5. [13]

    Lane-changing in traffic streams,

    J. A. Laval and C. F. Daganzo, “Lane-changing in traffic streams,” Transportation Research Part B: Methodological, vol. 40, no. 3, pp. 251–264, 2006. [Online]. Available: https://www.sciencedirect.com/ science/article/pii/S019126150500055X

    2006

  6. [14]

    Capacity-drop at extended bottlenecks: Merge, diverge, and weave,

    D. Chen and S. Ahn, “Capacity-drop at extended bottlenecks: Merge, diverge, and weave,”Transportation Research Part B: Methodological, vol. 108, pp. 1–20, 2018. [Online]. Available: https://www.sciencedirect.com/science/article/pii/S0191261517306938

    2018

  7. [15]

    Freeway vehicle fuel efficiency improvement via cooperative adaptive cruise control,

    H. Liu, S. E. Shladover, X.-Y . Lu, and X. D. Kan, “Freeway vehicle fuel efficiency improvement via cooperative adaptive cruise control,”Journal of Intelligent Transportation Systems, vol. 25, no. 6, pp. 574–586, 2021. [Online]. Available: https://www.sciencedirect.com/ scienc...

    2021

  8. [16]

    Heavy-duty vehicle platoon formation for fuel efficiency,

    K.-Y . Liang, J. M ˚artensson, and K. H. Johansson, “Heavy-duty vehicle platoon formation for fuel efficiency,”IEEE Transactions on Intelligent Transportation Systems, vol. 17, no. 4, pp. 1051–1061, 2016

    2016

  9. [17]

    Planning of truck platoons: A literature review and directions for future research,

    A. K. Bhoopalam, N. Agatz, and R. Zuidwijk, “Planning of truck platoons: A literature review and directions for future research,” Transportation Research Part B: Methodological, vol. 107, pp. 212– 228, 2018. [Online]. Available: https://www.sciencedirect.com/science/ article/p...

    2018

  10. [18]

    Decentralized coordination for truck platooning,

    Y . Zeng, M. Wang, and R. T. Rajan, “Decentralized coordination for truck platooning,”Computer-Aided Civil and Infrastructure Engineering, vol. 37, no. 15, pp. 1997–2015, 2022. [Online]. Available: https://onlinelibrary.wiley.com/doi/abs/10.1111/mice.12899

    1997 doi

  11. [19]

    Where to decide? centralized vs. dis- tributed vehicle assignment for platoon formation,

    J. Heinovski and F. Dressler, “Where to decide? centralized vs. dis- tributed vehicle assignment for platoon formation,” 2023

    2023

  12. [20]

    Review of Lane-Changing Maneuvers of Connected and Automated Vehicles: Models, Algorithms and Traffic Impact Analyses,

    Z. Wang, X. Shi, and X. Li, “Review of Lane-Changing Maneuvers of Connected and Automated Vehicles: Models, Algorithms and Traffic Impact Analyses,”Journal of the Indian Institute of Science, vol. 99, no. 4, pp. 589–599, Dec. 2019. [Online]. Available: https://doi.org/10.1007/...

    2019 doi

  13. [21]

    Analysis of lane-change crashes and near-crashes,

    D. Hs, “Analysis of lane-change crashes and near-crashes,” 2009. [On- line]. Available: https://api.semanticscholar.org/CorpusID:172779614

    2009

  14. [22]

    Studies of vehicle lane-changing dynamics and its effect on traffic efficiency, safety and environmental impact,

    X. Li and J.-Q. Sun, “Studies of vehicle lane-changing dynamics and its effect on traffic efficiency, safety and environmental impact,” Physica A: Statistical Mechanics and its Applications, vol. 467, pp. 41–58, 2017. [Online]. Available: https://www.sciencedirect.com/ science...

    2017

  15. [23]

    A model for the structure of lane-changing decisions,

    P. Gipps, “A model for the structure of lane-changing decisions,” Transportation Research Part B: Methodological, vol. 20, no. 5, pp. 403–414, 1986. [Online]. Available: https://www.sciencedirect.com/ science/article/pii/0191261586900123

    1986

  16. [25]

    Estimation of traffic densities for multilane roadways using a markov model approach,

    K. Singh and B. Li, “Estimation of traffic densities for multilane roadways using a markov model approach,”IEEE Transactions on Industrial Electronics, vol. 59, no. 11, pp. 4369–4376, 2012

    2012

  17. [26]

    Pay to change lanes: A cooperative lane-changing strategy for connected/automated driving,

    D. Lin, L. Li, and S. E. Jabari, “Pay to change lanes: A cooperative lane-changing strategy for connected/automated driving,” Transportation Research Part C: Emerging Technologies, vol. 105, pp. 550–564, 2019. [Online]. Available: https://www.sciencedirect.com/ science/article...

    2019

  18. [27]

    Review of deep learning: concepts, CNN architectures, challenges, applications, future directions,

    L. Alzubaidi, J. Zhang, A. J. Humaidi, A. Al-Dujaili, Y . Duan, O. Al-Shamma, J. Santamar ´ıa, M. A. Fadhel, M. Al-Amidie, and L. Farhan, “Review of deep learning: concepts, CNN architectures, challenges, applications, future directions,”Journal of Big Data, vol. 8, no. 1, p. ...

    2021 doi

  19. [28]

    Human-level control through deep reinforcement learning,

    V . Mnih, K. Kavukcuoglu, D. Silver, A. A. Rusu, J. Veness, M. G. Bellemare, A. Graves, M. Riedmiller, A. K. Fidjeland, G. Ostrovski, S. Petersen, C. Beattie, A. Sadik, I. Antonoglou, H. King, D. Kumaran, D. Wierstra, S. Legg, and D. Hassabis, “Human-level control through deep...

    doi

  20. [29]

    Mastering the game of go without human knowledge,

    D. Silver, J. Schrittwieser, K. Simonyan, I. Antonoglou, A. Huang, A. Guez, T. Hubert, L. Baker, M. Lai, A. Bolton, Y . Chen, T. Lillicrap, F. Hui, L. Sifre, G. van den Driessche, T. Graepel, and D. Hassabis, “Mastering the game of go without human knowledge,” vol. 550, no. 76...

    doi

  21. [30]

    Mastering chess and shogi by self-play with a general reinforcement learning algorithm,

    D. Silver, T. Hubert, J. Schrittwieser, I. Antonoglou, M. Lai, A. Guez, M. Lanctot, L. Sifre, D. Kumaran, T. Graepel, T. Lillicrap, K. Simonyan, and D. Hassabis, “Mastering chess and shogi by self-play with a general reinforcement learning algorithm,” 2017. [Online]. Available...

    2017 arXiv

  22. [31]

    Safe, multi-agent, reinforcement learning for autonomous driving,

    S. Shalev-Shwartz, S. Shammah, and A. Shashua, “Safe, multi-agent, reinforcement learning for autonomous driving,” 2016. [Online]. Available: https://arxiv.org/abs/1610.03295

    2016 arXiv

  23. [32]

    Deep reinforcement learning framework for autonomous driving,

    A. E. Sallab, M. Abdou, E. Perot, and S. Yogamani, “Deep reinforcement learning framework for autonomous driving,”Electronic Imaging, vol. 29, no. 19, p. 70–76, Jan. 2017. [Online]. Available: http://dx.doi.org/10.2352/ISSN.2470-1173.2017.19.A VM-023

    2017

  24. [33]

    Lane change decision- making through deep reinforcement learning with rule-based con- straints,

    J. Wang, Q. Zhang, D. Zhao, and Y . Chen, “Lane change decision- making through deep reinforcement learning with rule-based con- straints,” in2019 International Joint Conference on Neural Networks (IJCNN), 2019, pp. 1–6

    2019

  25. [34]

    An intelli- gent lane-changing behavior prediction and decision-making strategy for an autonomous vehicle,

    W. Wang, T. Qie, C. Yang, W. Liu, C. Xiang, and K. Huang, “An intelli- gent lane-changing behavior prediction and decision-making strategy for an autonomous vehicle,”IEEE Transactions on Industrial Electronics, vol. 69, no. 3, pp. 2927–2937, 2022

    2022

  26. [35]

    Multi-agent drl-based lane change with right-of-way collaboration awareness,

    J. Zhang, C. Chang, X. Zeng, and L. Li, “Multi-agent drl-based lane change with right-of-way collaboration awareness,”IEEE Transactions on Intelligent Transportation Systems, vol. 24, no. 1, pp. 854–869, 2023

    2023

  27. [36]

    Driving decision and control for automated lane change behavior based on deep reinforcement learning,

    T. Shi, P. Wang, X. Cheng, C.-Y . Chan, and D. Huang, “Driving decision and control for automated lane change behavior based on deep reinforcement learning,” in2019 IEEE Intelligent Transportation Systems Conference (ITSC), 2019, pp. 2895–2900

    2019

  28. [37]

    An integrated model for autonomous speed and lane change decision-making based on deep re- inforcement learning,

    J. Peng, S. Zhang, Y . Zhou, and Z. Li, “An integrated model for autonomous speed and lane change decision-making based on deep re- inforcement learning,”IEEE Transactions on Intelligent Transportation Systems, vol. 23, no. 11, pp. 21 848–21 860, 2022

    2022

  29. [38]

    Decision making of autonomous vehicles in lane change scenarios: Deep reinforcement learning approaches with risk awareness,

    G. Li, Y . Yang, S. Li, X. Qu, N. Lyu, and S. E. Li, “Decision making of autonomous vehicles in lane change scenarios: Deep reinforcement learning approaches with risk awareness,”Transportation Research Part C: Emerging Technologies, vol. 134, p. 103452, 2022. [Online]. Availa...

    2022

  30. [39]

    Attention-based hierarchical deep reinforcement learning for lane change behaviors in autonomous driving,

    Y . Chen, C. Dong, P. Palanisamy, P. Mudalige, K. Muelling, and J. M. Dolan, “Attention-based hierarchical deep reinforcement learning for lane change behaviors in autonomous driving,” in2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), 201...

    2019

  31. [40]

    Highway lane change decision- making via attention-based deep reinforcement learning,

    J. Wang, Q. Zhang, and D. Zhao, “Highway lane change decision- making via attention-based deep reinforcement learning,”IEEE/CAA Journal of Automatica Sinica, vol. 9, no. 3, pp. 567–569, 2022

    2022

  32. [41]

    Multi-reward architecture based reinforcement learning for highway driving policies,

    W. Yuan, M. Yang, Y . He, C. Wang, and B. Wang, “Multi-reward architecture based reinforcement learning for highway driving policies,” in2019 IEEE Intelligent Transportation Systems Conference (ITSC), 2019, pp. 3810–3815

    2019

  33. [42]

    Harmonious lane changing via deep reinforcement learning,

    G. Wang, J. Hu, Z. Li, and L. Li, “Harmonious lane changing via deep reinforcement learning,”IEEE Transactions on Intelligent Transporta- tion Systems, vol. 23, no. 5, pp. 4642–4650, 2022

    2022

  34. [43]

    High-level decision making for safe and reasonable autonomous lane changing using reinforcement learning,

    B. Mirchevska, C. Pek, M. Werling, M. Althoff, and J. Boedecker, “High-level decision making for safe and reasonable autonomous lane changing using reinforcement learning,” in2018 21st International Conference on Intelligent Transportation Systems (ITSC), 2018, pp. 2156–2162

    2018

  35. [44]

    A reinforcement learning approach to autonomous decision making of intelligent vehicles on highways,

    X. Xu, L. Zuo, X. Li, L. Qian, J. Ren, and Z. Sun, “A reinforcement learning approach to autonomous decision making of intelligent vehicles on highways,”IEEE Transactions on Systems, Man, and Cybernetics: Systems, vol. 50, no. 10, pp. 3884–3897, 2020

    2020

  36. [45]

    A real-world reinforcement learning framework for safe and human-like tactical decision-making,

    M. U. Yavas, T. Kumbasar, and N. K. Ure, “A real-world reinforcement learning framework for safe and human-like tactical decision-making,” IEEE Transactions on Intelligent Transportation Systems, vol. 24, no. 11, pp. 11 773–11 784, 2023

    2023

  37. [46]

    Modeling coupled driving behavior during lane change: A multi-agent transformer reinforcement learning approach,

    H. Guo, M. Keyvan-Ekbatani, and K. Xie, “Modeling coupled driving behavior during lane change: A multi-agent transformer reinforcement learning approach,”Transportation Research Part C: Emerging Technologies, vol. 165, p. 104703, 2024. [Online]. Available: https://www.scienced...

    2024

  38. [47]

    Coordination for con- nected and automated vehicles at non-signalized intersections: A value decomposition-based multiagent deep reinforcement learning approach,

    Z. Guo, Y . Wu, L. Wang, and J. Zhang, “Coordination for con- nected and automated vehicles at non-signalized intersections: A value decomposition-based multiagent deep reinforcement learning approach,” IEEE Transactions on V ehicular Technology, vol. 72, no. 3, pp. 3025– 3034, 2023

    2023

  39. [48]

    Multi- agent actor-critic for mixed cooperative-competitive environments,

    R. Lowe, Y . Wu, A. Tamar, J. Harb, P. Abbeel, and I. Mordatch, “Multi- agent actor-critic for mixed cooperative-competitive environments,”

  40. [49]

    Fundamentals of recurrent neural network (rnn) and long short-term memory (lstm) network,

    A. Sherstinsky, “Fundamentals of recurrent neural network (rnn) and long short-term memory (lstm) network,”Physica D: Nonlinear Phenomena, vol. 404, p. 132306, 2020. [Online]. Available: https: //www.sciencedirect.com/science/article/pii/S0167278919305974

    2020

  41. [50]

    Kinematic and dynamic vehicle models for autonomous driving control design,

    J. Kong, M. Pfeiffer, G. Schildbach, and F. Borrelli, “Kinematic and dynamic vehicle models for autonomous driving control design,” in2015 IEEE Intelligent V ehicles Symposium (IV), 2015, pp. 1094–1099

    2015

  42. [51]

    General lane-changing model mobil for car-following models,

    A. Kesting, M. Treiber, and D. Helbing, “General lane-changing model mobil for car-following models,”Transportation Research Record, vol. 1999, no. 1, pp. 86–94, 2007. [Online]. Available: https://doi.org/10.3141/1999-10

    1999 doi

  43. [52]

    Congested traffic states in empirical observations and microscopic simulations,

    M. Treiber, A. Hennecke, and D. Helbing, “Congested traffic states in empirical observations and microscopic simulations,”Phys. Rev. E, vol. 62, pp. 1805–1824, Aug 2000. [Online]. Available: https://link.aps.org/doi/10.1103/PhysRevE.62.1805

    doi

  44. [53]

    String-stable cacc design and experimental validation, a frequency- domain approach,

    G. Naus, R. Vugts, J. Ploeg, M. Molengraft, van de, and M. Steinbuch, “String-stable cacc design and experimental validation, a frequency- domain approach,”IEEE Transactions on V ehicular Technology, vol. 59, no. 9, pp. 4268–4279, 2010

    2010

  45. [54]

    Impacts of commercially available adaptive cruise control vehicles on highway stability and throughput,

    M. Shang and R. E. Stern, “Impacts of commercially available adaptive cruise control vehicles on highway stability and throughput,” Transportation Research Part C: Emerging Technologies, vol. 122, p. 102897, 2021. [Online]. Available: https://www.sciencedirect.com/ science/art...

    2021

  46. [55]

    Calibration and evaluation of car following models using real-world driving data,

    M. Pourabdollah, E. Bj ¨arkvik, F. F¨urer, B. Lindenberg, and K. Burgdorf, “Calibration and evaluation of car following models using real-world driving data,” in2017 IEEE 20th International Conference on Intelligent Transportation Systems (ITSC), 2017, pp. 1–6

    2017

  47. [2020]

    Available: https://arxiv.org/abs/1706.02275

    [Online]. Available: https://arxiv.org/abs/1706.02275

    arXiv

Pith tools

Reviewed August 3, 2026 · model on record in the stance chip above.