pith. sign in

arxiv: 2606.23536 · v1 · pith:RR3G2AXNnew · submitted 2026-06-22 · 💻 cs.LG

Simulation-Free Estimation of Traffic Flows from Sparse Count Data

Davide Guastella , Gianluca Bontempi This is my paper

Pith reviewed 2026-06-26 09:04 UTC · model grok-4.3

classification 💻 cs.LG
keywords traffic flow estimationsparse count dataweighted least-squaressimulation-free methodregion-to-region routessensor contribution matrixBrussels road network
0
0 comments X

The pith

A weighted least-squares optimization on feasible region-to-region routes estimates time-varying traffic flows from sparse sensor counts without simulation.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper develops a method that divides a road network into spatial regions, generates possible routes between them, and solves an optimization problem to assign vehicle numbers to those routes so the resulting counts match sparse aggregated sensor readings. A contribution matrix weights how much each route is observed by each sensor, guiding the solver toward configurations consistent with the data. The resulting route flows then yield edge-level trajectories by matching against the temporal and volume profiles in the input counts. The approach is tested on real and synthetic data from the Brussels network, where it matches daily traffic shapes and beats baseline methods while using far less computation.

Core claim

The central claim is that partitioning the network into regions, enumerating feasible inter-region routes, and solving a weighted least-squares problem whose objective incorporates a sensor coverage matrix produces flow allocations that reproduce observed daily profiles and outperform existing methods at a fraction of the computational cost.

What carries the argument

The weighted contribution matrix that encodes sensor coverage and steers the optimizer toward flow configurations directly observable by sensors.

If this is right

  • Traffic estimation becomes feasible on networks where full microscopic simulation is too slow or too data-hungry.
  • Only aggregated regional counts are required as input, not individual vehicle trajectories.
  • Edge-level flow estimates follow directly from scoring the optimized routes against the observed temporal profiles.
  • Computational cost remains low enough for repeated daily or near-real-time use.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same route-allocation idea could be tested on other sparse-count problems such as pedestrian flows in buildings or commodity flows in supply networks.
  • If the enumerated route set systematically omits high-volume corridors, the optimizer would be forced to over-allocate on the included routes and the estimates would degrade.
  • Adding a small regularization term that favors smoother temporal profiles might improve robustness when sensor coverage is especially sparse.

Load-bearing premise

The chosen set of feasible routes together with the sensor contribution matrix supplies enough constraints for the optimizer to recover flows that match the true underlying traffic.

What would settle it

Apply the method to a network where the true route-level flows are known from a controlled simulation or detailed tracking, then measure whether the estimated flows deviate from the known values by more than a few percent on average.

Figures

Figures reproduced from arXiv: 2606.23536 by Davide Guastella, Gianluca Bontempi.

Figure 1
Figure 1. Figure 1: We use the open-source simulator SUMO to vali [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 1
Figure 1. Figure 1: The proposed method operates hierarchically: the [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Brussels road network used as a case study. Each [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 4
Figure 4. Figure 4: Average number of vehicles observed in all the re [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 3
Figure 3. Figure 3: Profile of the average loss for the considered sce [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figure 5
Figure 5. Figure 5: Absolute error (number of vehicles) per region and [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Average traffic counts for two spatial regions in the Brussels scenario. Figure 6a approximatively matches the average [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗
Figure 8
Figure 8. Figure 8: Spatial distribution of the traffic counts error (MAE) [PITH_FULL_IMAGE:figures/full_fig_p008_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Comparison of the traffic profiles obtained by our [PITH_FULL_IMAGE:figures/full_fig_p009_9.png] view at source ↗
Figure 11
Figure 11. Figure 11: Per-region Shannon entropy of the ground-truth [PITH_FULL_IMAGE:figures/full_fig_p010_11.png] view at source ↗
Figure 10
Figure 10. Figure 10: compares the aggregated (average over all time intervals) ground-truth transition matrix Q (left) with the es￾timate transition matrix Qˆ¯ (center), and their element-wise signed difference Q¯ − Q (right). 0_2 1_1 1_2 1_3 1_4 1_5 2_0 2_1 2_2 2_3 2_4 2_5 3_0 3_1 3_2 3_3 3_4 3_5 4_0 4_1 4_2 4_3 4_4 4_5 5_2 5_3 Destination region 0_2 1_1 1_2 1_3 1_4 1_5 2_0 2_1 2_2 2_3 2_4 2_5 3_0 3_1 3_2 3_3 3_4 3_5 4_0 4_1… view at source ↗
Figure 12
Figure 12. Figure 12: Comparison of trajectory log-likelihood distribu [PITH_FULL_IMAGE:figures/full_fig_p011_12.png] view at source ↗
read the original abstract

We propose a method for estimating time-varying traffic flow patterns from sparse aggregated vehicle counts. The method partitions the study area into spatial regions, constructs a set of feasible region-to-region routes, and solves a weighted least-squares optimization problem to determine the number of vehicles to allocate on each route. A weighted contribution matrix encodes sensor coverage, steering the optimizer toward flow configurations that are directly observable by sensors. Edge-level trajectories are then derived by scoring candidate routes against the temporal and volumetric profiles of aggregated regional sensor counts. The method is evaluated on the Brussels road network using real and synthetic traffic data. Results show that the proposed approach reproduces the daily traffic profile in the input data and outperforms the baseline methods at a fraction of the computational cost.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 1 minor

Summary. The paper proposes a simulation-free method for estimating time-varying traffic flows from sparse aggregated vehicle counts. The approach partitions the study area into regions, enumerates feasible region-to-region routes, constructs a weighted contribution matrix encoding sensor coverage, solves a weighted least-squares optimization to allocate vehicles to routes, and derives edge-level trajectories by scoring routes against sensor profiles. Evaluation on the Brussels road network with real and synthetic data claims that the method reproduces daily traffic profiles in the input data and outperforms baseline methods at a fraction of the computational cost.

Significance. If the method could be shown to recover ground-truth flows from underdetermined sparse counts (rather than merely fitting aggregates), it would offer a computationally efficient alternative to simulation-based traffic estimation. The simulation-free formulation and use of a weighted contribution matrix to steer the optimizer are conceptually appealing, but the manuscript provides no quantitative metrics or identifiability analysis to support these strengths.

major comments (3)
  1. [Abstract] Abstract: The claim that the method 'reproduces the daily traffic profile in the input data' only verifies consistency with observed sensor aggregates. With sparse counts the linear system is underdetermined, so many route-flow vectors produce identical readings; the manuscript supplies no identifiability argument, regularization analysis, or proof that the weighted least-squares selects the true underlying flows rather than any feasible fit.
  2. [Abstract] Abstract / Evaluation: The assertion that the approach 'outperforms the baseline methods' is unsupported because no quantitative error metrics, validation details on the real and synthetic datasets, or comparison tables are provided, preventing assessment of the magnitude or statistical significance of any improvement.
  3. [Method] Method description: The construction of the feasible route set and the weighted contribution matrix is presented at a high level with no discussion of exhaustiveness of the route enumeration or how the weighting matrix resolves degeneracies in the underdetermined system; this is load-bearing for the central claim that the optimizer is steered toward true flows.
minor comments (1)
  1. [Abstract] Abstract: The phrase 'at a fraction of the computational cost' is stated without naming the baselines or reporting any runtime numbers.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive feedback. The comments correctly identify areas where additional discussion and quantitative support would strengthen the manuscript. We address each major comment below and will revise accordingly.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The claim that the method 'reproduces the daily traffic profile in the input data' only verifies consistency with observed sensor aggregates. With sparse counts the linear system is underdetermined, so many route-flow vectors produce identical readings; the manuscript supplies no identifiability argument, regularization analysis, or proof that the weighted least-squares selects the true underlying flows rather than any feasible fit.

    Authors: We agree that consistency with aggregates does not establish recovery of unique true flows in an underdetermined system. The weighted contribution matrix is intended to prioritize observable configurations, but the manuscript lacks a formal identifiability analysis. We will add a dedicated discussion of these limitations and the role of weighting in the revised version. revision: yes

  2. Referee: [Abstract] Abstract / Evaluation: The assertion that the approach 'outperforms the baseline methods' is unsupported because no quantitative error metrics, validation details on the real and synthetic datasets, or comparison tables are provided, preventing assessment of the magnitude or statistical significance of any improvement.

    Authors: The evaluation section reports reproduction of profiles and computational advantages on the Brussels data, but we acknowledge the absence of detailed quantitative metrics and tables. We will add error metrics (e.g., MAE on flow profiles), validation details, and comparison tables with statistical information in the revision. revision: yes

  3. Referee: [Method] Method description: The construction of the feasible route set and the weighted contribution matrix is presented at a high level with no discussion of exhaustiveness of the route enumeration or how the weighting matrix resolves degeneracies in the underdetermined system; this is load-bearing for the central claim that the optimizer is steered toward true flows.

    Authors: We will expand the method section with specifics on the route enumeration procedure, its exhaustiveness for the studied network, and further explanation of how the weighting resolves (or mitigates) degeneracies in the underdetermined system. revision: yes

Circularity Check

0 steps flagged

No significant circularity in the derivation chain

full rationale

The paper describes a method that partitions space, enumerates feasible routes, constructs a sensor contribution matrix, and solves a weighted least-squares problem whose objective is explicitly to match observed aggregated counts. Reproduction of the daily profile is the intended and direct consequence of this data-driven fit rather than a separate derived claim. Evaluation includes synthetic data (allowing ground-truth comparison) and runtime comparisons to baselines, providing external benchmarks. No equations, self-citations, or uniqueness arguments are shown that would reduce the outputs to the inputs by construction; the derivation remains an explicit optimization procedure driven by external sensor data.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Only the abstract is available; no explicit free parameters, axioms, or invented entities are described.

pith-pipeline@v0.9.1-grok · 5645 in / 1038 out tokens · 25042 ms · 2026-06-26T09:04:47.842400+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

17 extracted references

  1. [1]

    Behrisch and P

    M. Behrisch and P. Hartwig. A comparison of sumo’s count based and countless demand generation tools. SUMO Conference Proceedings, 2:125–131, Jun. 2022

    2022

  2. [2]

    Y . Cao, H. Van Lint, P. Krishnakumari, and M. Bliemer. Data driven origin–destination matrix estimation on large networks-a joint origin-destination-path-choice formulation.Transportation Research Part C: Emerg- ing Technologies, 168:104850, 2024

    2024

  3. [3]

    R. F. Daguano, L. R. Yoshioka, M. L. Netto, C. L. Marte, C. A. Isler, M. M. D. Santos, and J. F. Justo. Automatic calibration of microscopic traffic simulation models us- ing artificial neural networks.Sensors, 23(21):8798, 2023

    2023

  4. [4]

    Englezou, S

    Y . Englezou, S. Timotheou, and C. G. Panayiotou. Dy- namic origin-destination matrix estimation for networks operating under free-flow conditions using macroscopic flow dynamics.IFAC-PapersOnLine, 58(10):213–218,

  5. [5]

    17th IFAC Symposium on Control of Transporta- tion Systems CTS 2024

    2024

  6. [6]

    Galliani, P

    G. Galliani, P. Secchi, and F. Ieva. Estimation of dy- namic origin–destination matrices in a railway trans- portation network integrating ticket sales and passenger count data.Transportation Research Part A: Policy and Practice, 190:104246, 2024

    2024

  7. [7]

    D. A. Guastella, B. Cornelis, and G. Bontempi. Traf- fic simulation with incomplete data: the case of brus- sels. InProceedings of the 1st ACM SIGSPATIAL In- ternational Workshop on Methods for Enriched Mobility Data: Emerging Issues and Ethical Perspectives 2023, EMODE ’23, page 15–24, New York, NY , USA, 2023. Association for Computing Machinery

    2023

  8. [8]

    D. A. Guastella, A. Morales-Hern ´andez, B. Cornelis, and G. Bontempi. Calibration of vehicular traffic sim- ulation models by local optimization.Transportation, 2025

    2025

  9. [9]

    Nguyen and C

    S. Nguyen and C. Dupuis. An efficient method for com- puting traffic equilibria in networks with asymmetric transportation costs.Transportation Science, 18(2):185– 202, 1984

    1984

  10. [10]

    C. Osorio. High-dimensional offline origin-destination (od) demand calibration for stochastic traffic simulators 11 of large-scale road networks.Transportation Research Part B: Methodological, 124:18–43, 2019

    2019

  11. [11]

    P. K.B. Rangaiah, B.P. Pradeep kumar, and R. Augus- tine. Improving burn diagnosis in medical image re- trieval from grafting burn samples using b-coefficients and the clahe algorithm.Biomedical Signal Processing and Control, 99:106814, 2025

    2025

  12. [12]

    Roocroft, G

    A. Roocroft, G. Punzo, and M. A. Ramli. Flow count data-driven static traffic assignment models through network modularity partitioning.Transporta- tion, 52(1):185–214, 2025

    2025

  13. [13]

    Sadiq, M

    M. Sadiq, M. N. Kadhim, D. Al-Shammary, and M. Mi- lanova. Novel eeg classification based on hellinger distance for seizure epilepsy detection.IEEE Access, 12:127357–127367, 2024

    2024

  14. [14]

    L. Tang, D. Zhang, Y . Han, A. Fu, H. Zhang, Y . Tian, L. Yue, D. Wang, and J. Sun. Parallel-computing- based calibration for microscopic traffic simulation model.Transportation Research Record, 2678(4):279– 294, 2024

    2024

  15. [15]

    G. Wei, D. Gundleg ˚ard, and C. Rydergren. Consis- tent origin-destination and link flow estimation based on data-driven network assignment.Transportation Re- search Procedia, 86:668–675, 2025

    2025

  16. [16]

    Zhang, N

    C. Zhang, N. Arora, C. Bian, Y . Li, W. Ng, A. Tomkins, B. Yan, J. Zhang, and C. Osorio. Origin-destination travel demand estimation: An approach that scales worldwide, and its application to five metropolitan high- way networks, 2025

    2025

  17. [17]

    Zhang, G

    Z. Zhang, G. Yuan, Z. Qin, and Q. Luo. An improve- ment by introducing lbfgs idea into the adam optimizer for machine learning.Expert Systems with Applications, 296:129002, 2026. 12

    2026