Pith. sign in

REVIEW 1 major objections 1 minor 28 references

Shape and Performance of Fastest Paths over Networks with Interacting Selfish Agents

T0 review · 1 major / 1 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 At a critical traffic level, fastest paths in eight cities simultaneously change shape: detours and path-area peaks spike, and the weak pull of city centers flips into strong repulsion.

desk verdict A useful path-based view of congestion transition, but the abstract's universal concurrent flip is not supported by the paper's own city-by-city results. read the letter →

arxiv 2412.17665 v2 pith:L43GJ6VK submitted 2024-12-23 physics.soc-ph cond-mat.stat-mech

Marco Cogoni , Giovanni Busonera , Enrico Gobbetti This is my paper · ORCID
classification physics.soc-phcond-mat.stat-mech
keywords fastestpathscongestiontransitionurbanroadnetworkspathshapedetourinnessperformanceindexGinicoefficient
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

The paper argues that the onset of congestion in urban road networks is not a gradual, edge-by-edge slowdown but a path-level qualitative transition. By simulating the morning rush hour as a sequence of vehicles that each choose the currently fastest path, the authors find that all eight cities they study show a critical traffic volume at which path shape, the share of rejected or incomplete trips, and the inequality of performance degradation change together. At that volume, average detour and the signed area between path and straight line (inness) peak sharply, and the weak pull of city centers on paths under light traffic reverses into a strong repulsion. The authors introduce a performance index that multiplies how fast a trip goes by how much of it can be completed, and use its Gini coefficient to show that congestion degrades a few paths far more than most. A sympathetic reader would care because this identifies a single observable early-warning level for each city and a path-based alternative to edge-based centrality measures.

What carries the argument

The load-bearing machinery is a sequential traffic-loading simulation on real road graphs combined with three path-geometry observables and one path-performance observable. Vehicles are added one by one; each computes the fastest path with full knowledge of current edge travel times, edge speed falls linearly with accumulated density following the standard single-regime speed-density relation (Eq. 1), and vehicles are never removed. Path shape is characterized by detour (maximum perpendicular distance from the straight origin-destination line) and inness (signed area between path and straight line, positive when the path leans toward the map center), both normalized by straight-line distance or its square. Path performance is captured by the Performance Index P = S·C, where the slowdown factor S compares congested travel time with empty-network travel time (rescaled by completeness) and the completeness factor C is the fraction of the path actually traversable before a dysfunctional edge or the time limit. The transition is located by the flex of the hard-rejected path ratio, and the same critical traffic level is where detour, inness, soft-reject ratio, and Gini all show concurrent qualitative changes.

What would settle it

Run the same eight-city protocol but let a fraction of vehicles reroute once or twice mid-trip, or remove vehicles when their trip completes, and check whether the detour and inness peaks and the inness sign reversal still occur at a well-defined critical volume; if the concurrent change smears out or the flip disappears, the central claim is an artifact of the static-loading assumption. A data-side check: compare the predicted origin-destination performance asymmetry maps against GPS-derived trip-time data for the same cities during morning peak; if neighborhoods predicted to be hard to leave are not, the performance index is not capturing real asymmetry.

Watch

Extended reading notes

Core claim

The central discovery is that the fastest-path structure of an urban network reorganizes at a well-defined critical traffic level, and that this reorganization is visible in path geometry before it is visible in connectivity. For every city examined, the average detour and the mean and variance of inness grow as traffic approaches the transition, peak at or just before it, then collapse; the inness sign flips from a slight attraction toward the city center in light traffic to a strong repulsion beyond the transition. The paper also shows that the flex of the hard-rejected path ratio localizes the transition, that soft-rejections peak just before it, and that the Gini coefficient of the performance index rises from about 0.1 to above 0.5 through the transition, meaning a minority of paths retain most of their performance while the rest degrade sharply. Finally, mapping the average performance index onto origins and destinations reveals neighborhoods that are easy to reach but hard to leave, and vice versa, under congestion.

Load-bearing premise

The result stands on the assumption that the morning rush hour is faithfully represented as a sequence of vehicles that each choose a fastest path with complete network knowledge, never reroute, never leave, and face speeds that fall linearly with accumulated density; if real congestion involves rerouting, trip completion, or departure-time choices, the concurrent geometric transition and the center-attraction-to-repulsion flip could be an artifact of that loading protocol.

Editorial extensions

If this is right

  • If the central claim holds, each city has a predictable critical traffic volume that can be read off the flex of the hard-reject curve, with multiple independent indicators converging on the same volume.
  • Detour and inness peaks serve as precursors: they appear just before the hard-reject ratio grows, so geometry can warn of an approaching breakup before connectivity measures do.
  • The performance index P, averaged over origins and destinations, provides a map of source-sink asymmetry: areas that are attractive as destinations can be poor origins under congestion, so planners can identify neighborhoods that need redundant exits rather than only additional entry capacity.
  • The Gini coefficient of P can be monitored over time as an inequality-of-degradation meter; a rise past roughly 0.4 signals that the network is entering the transition regime.
  • Network fragmentation happens through the saturation of a very small fraction of edges (around 0.1% or less), so resilience measures should focus on redundancy of those few desirable edges rather than on average edge capacity.

Reading between the lines

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

  • A natural extension, not tested here, is to allow en-route rerouting or trip completion and removal, and to check whether the concurrent geometric peak and the center-attraction-to-repulsion flip survive; if they vanish, the transition is a property of the static loading protocol rather than of the cities.
  • The inness sign flip suggests an effective central potential that changes sign at the transition; extracting the average inness per origin-destination bin and comparing it across cities could reveal whether the repulsive strength correlates with measurable features such as river barriers, ring-road layout, or one-way street density.
  • The same path-based machinery transfers to other agent-competition systems such as packet routing or pedestrian flows, provided the completeness factor is reweighted to match how much value a partially completed trip has in that context.
  • A testable robustness check: rerun the eight cities with a nonlinear speed-density relation (for example, with a capacity drop) and see whether the critical volume and the inness reversal shift strongly; if they do, the linear relation is doing much of the work.
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

1 major / 1 minor

Summary. The paper studies how fastest paths (FPs) on the road networks of eight cities change as traffic is progressively loaded under a Greenshields-type interaction model. The authors define path-level metrics (effective length, detour, inness, and a new Performance Index combining slowdown and completeness) and report that, at a critical traffic level identified by a flex in the hard-rejection curve, detour and inness peak, inness changes from a slight attraction toward city centers to strong repulsion, and the Gini coefficient of performance inequality grows sharply. They also use the Performance Index to reveal origin/destination asymmetries. The central claim is that these changes occur concurrently and universally across the eight cities.

Significance. If the central claim were fully supported, this would be a useful contribution: it shifts congestion analysis from edge-based to path-based observables, introduces a practical performance index, and documents rich city-specific behavior. The authors are transparent about several limitations, including noisy fits for exponents beyond the transition and the non-power-law behavior of Beijing, and they explicitly report exceptions such as Rome's persistent center-repulsion. The interaction model is taken from a prior published derivation (Ref. [9]), which reduces circularity concerns. However, the headline universality and concurrency statements are not actually supported by the paper's own qualitative descriptions, and the transition is located by visual inspection rather than by an operational criterion. These issues affect the main contribution and require revision.

major comments (1)
  1. [Section III A2 and Section III C] The model's sensitivity to its free parameters is not addressed. Section II A introduces L (average space per vehicle) and tau (maximum allowed travel time), and the maximum target volume is a simulation choice. The location of the transition and the claimed peak alignment are likely sensitive to these parameters, yet no sensitivity analysis is reported. At minimum, a paragraph should discuss how the findings depend on L and tau, and why the chosen values (e.g., tau = 3600 s, which the authors note is 'too long for the smaller urban areas') do not alter the qualitative conclusions.
minor comments (1)
  1. [Section III A2] The paper references many supplementary figures (S1-S48) that are not available in the reviewed version; please make them accessible or at least summarize their content for the main text, since several city-level claims rely on them.

Circularity Check

0 steps flagged · score 2.0 of 10

No circular derivation: the transition peaks are emergent simulation findings, with only non-load-bearing self-citations.

full rationale

The paper's central claim—concurrent peaks in detour and inness, a peak in their variance, and a sign change of inness near a critical traffic level—is obtained by running a stated sequential-loading simulation (Greenshields speed-density relation, Eq. 1, with fastest paths recomputed at each step) and then measuring geometric path properties (detour, inness) and the derived Performance Index P = S·C. No parameter is fitted to reproduce the reported peaks. The critical traffic level is identified independently by the flex of the hard-reject ratio curve (Section III A2: 'This transition traffic volume ... coincides also with the position of the flex in the curve of the hard-rejected path ratio'), and the peaking of detour and inness is then observed to occur 'just before the transition'; this is an empirical coincidence, not a definitional one. The inness sign convention (positive toward the city center) is a definition, but the finding that the sign changes with traffic is data-driven. The interaction model is taken from the authors' prior PRE 2024 paper [9], which is a self-citation; however, the model equations are restated in Section II A and the present results do not reduce to any result imported from [9]. The Gini coefficient and soft-reject ratio are called 'precursor[s]' based on their observed ordering in traffic, not by construction. The internal inconsistency between the abstract's 'all eight cities' wording and Section III A2's statement that Rome 'strongly repels all paths but those with small s' is a correctness or overstatement risk, not a circularity. Therefore no circular step can be exhibited with a specific equation-to-equation reduction; the score of 2 reflects only the presence of non-load-bearing self-citations, not a circular derivation.

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

The model imports standard and simplified traffic-flow assumptions from the authors' prior work [9] and from the traffic-flow literature; the central claims rest on these modeling choices plus a set of hand-chosen parameters (L, tau, target volume) whose values are not fully reported. No new physical entities are introduced; the Performance Index is a derived metric, not an invented entity.

free parameters (3)
  • L (average space occupied by one vehicle) = not stated in main text
    Eq. (1) normalizes edge density; the value sets the traffic-volume scale at which the transition appears across cities. Reported results are in raw vehicle counts, so changing L shifts the critical level.
  • tau (maximum path travel time) = 3600 s
    Used for soft-reject criterion and occupancy normalization; authors admit it may be too long for smaller cities like Rome and Madrid, where soft-reject peaks stay below 10%.
  • Maximum target traffic volume = 2.0 x 10^6 OD pairs
    Chosen to drive all cities to near-gridlock; affects the deep-congestion regime but not the transition location itself.
assumptions (5)
  • domain assumption Single-regime Greenshields linear speed-density relation
    Invoked in Section II A: speed on an edge starts at free flow and decreases linearly to zero with maximum density. This standard but simplified flow model determines all congestion dynamics in the paper.
  • domain assumption Sequential greedy loading with no vehicle removal or re-routing
    Section II A: vehicles are added one at a time, each picks the current fastest path with complete knowledge, and 'vehicles are never removed from the network'. This substitutes a static superposition of path loads for time-dependent traffic.
  • domain assumption Uniform random OD sampling over all network nodes
    Section II A: 'O and D are chosen uniformly over the network nodes'. Real demand is spatially heterogeneous; the paper does not test sensitivity to demand patterns, and this choice affects rejection ratios and the P distribution.
  • domain assumption The critical scaling exponents apply to finite networks
    Section III A1: the authors note a strict definition of exponents holds only for infinite networks but treat them as useful for real finite networks; the fits are admitted to be noisy beyond the transition.
  • domain assumption A partially completed trip has linear value (completeness factor)
    Section II C: the Performance Index includes C, and the authors state this is tailored for vehicular traffic because partial transport is still valuable; other transport contexts would use an all-or-nothing rule.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Shape and Performance of Fastest Paths over Networks with Interacting Selfish Agents." pith.science (2026). https://pith.science/paper/L43GJ6VK

@misc{pith2026241217665,
  author       = {Pith},
  title        = {Pith review of: Shape and Performance of Fastest Paths over Networks with Interacting Selfish Agents},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/L43GJ6VK}},
  note         = {Machine review of arXiv:2412.17665}
}
read the original abstract

We study the evolution of the fastest paths (FP) in transportation networks under increasing congestion. Moving from the common edge-based to a path-based analysis, we examine the directed FPs connecting random origin-destination pairs as traffic grows. We describe their shape through effective length, detour (maximum distance of FP from a straight line), inness (signed area between FP and straight line), and their performance through a novel metric measuring how fast and how far an agent travels toward its destination. The entire network is characterized by analyzing the distribution of the performance metric (and its Gini coefficient) across uniformly sampled paths. The study focuses on the traffic loading phenomenon that takes place during the morning peak hour for eight major cities: Networks start with empty edges that are progressively populated by the FPs of single vehicles. As vehicle density grows, the interactions among selfish agents becomes stronger at edge level, and travel speed linearly decreases, thus optimal paths dynamically change with traffic. We fully characterize the transition to congestion and discuss the common aspects among the cities (and some peculiarities), in particular we were able to pinpoint a critical traffic level (or a sequence) for which path shape, rejection ratio, and inequality of the performance degradation, show a concurrent qualitative change. For all cities we observe large peaks for both detour and inness (and their variance) in the proximity of the critical traffic level. Inness shows that paths are slightly attracted by city centers with light traffic, but switch to a strong repulsion immediately beyond the transition. Finally, our path performance metric highlighted a strongly asymmetric behavior when the city neighborhoods act as origins or destinations.

Figures

Figures reproduced from arXiv: 2412.17665 by the authors.

Figure 1
Figure 1. FIG. 1. Between O and D we construct: a straight line [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. FIG. 2. Critical exponents [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. FIG. 3. The full distributions of inness and detour for London. [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (6 more)
Figure 5
Figure 5. Figure 5: FIG. 5. Path reject ratio: solid black line are the hard-rejects [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: FIG. 6. Normalized inness for four traffic levels (the third plot corresponds to the transition traffic level). Vertical dotted line [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: FIG. 7. In both panels: FPs between same OD (both at 15 km radius) at different traffic load. (The circle and diamond [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 8
Figure 8. Figure 8: FIG. 8. Distributions (counts of paths per bin) of [PITH_FULL_IMAGE:figures/full_fig_p009_8.png]
Figure 9
Figure 9. Figure 9: FIG. 9 [PITH_FULL_IMAGE:figures/full_fig_p010_9.png]
Figure 11
Figure 11. Figure 11: FIG. 11. The Gini coefficient as traffic load increases for the [PITH_FULL_IMAGE:figures/full_fig_p010_11.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

28 extracted references · 27 canonical work pages

  1. [9]

    Eugene Stanley, and Shlomo Havlin

    Guanwen Zeng, Jianxi Gao, Louis Shekhtman, Sheng- min Guo, Weifeng Lv, Jianjun Wu, Hao Liu, Orr Levy, Daqing Li, Ziyou Gao, H. Eugene Stanley, and Shlomo Havlin. Multiple metastable network states in urban traf- fic. Proc Natl Acad Sci USA , 117(30):17528–17534, July 2020

    work page 2020

  2. [1]

    The linear fit holds up only up to ∼ 3 km for low traffic, while linearity is regained deep into the congested regime

    Critical exponents in the low-congestion regime Path travel time scaling ( m) is well approximated by a linear fit in the log-log space for all cities except for Beijing where the curve does not follow a power law. The linear fit holds up only up to ∼ 3 km for low traffic, while linearity is regained deep into the congested regime. The results of the line...

    work page

  3. [2]

    sub-transition

    Morphological changes and central force effects The ξ and χ exponents describe the scaling of the de- tour and path-length variability on average, but more in- formation may be extracted from their full distributions at all traffic levels. For the city of London, the two dis- tributions P(I) and P(d) are shown in fig. 3 where the color indicates the numbe...

    work page

  4. [3]

    Traffic and related self-driven many- particle systems

    Dirk Helbing. Traffic and related self-driven many- particle systems. Rev. Mod. Phys., 73(4):1067–1141, De- cember 2001. Publisher: American Physical Society

    work page 2001

  5. [4]

    Eugene Stanley, and Shlomo Havlin

    Guanwen Zeng, Daqing Li, Shengmin Guo, Liang Gao, Ziyou Gao, H. Eugene Stanley, and Shlomo Havlin. Switch between critical percolation modes in city traf- fic dynamics. Proc Natl Acad Sci USA , 116(1):23–28, January 2019

    work page 2019

  6. [5]

    Complex-network-based traffic network analy- sis and dynamics: A comprehensive review

    Mengyao Zhang, Tao Huang, Zhaoxia Guo, and Zheng- gang He. Complex-network-based traffic network analy- sis and dynamics: A comprehensive review. Physica A: Statistical Mechanics and its Applications , 607:128063, 2022

    work page 2022

  7. [6]

    Impact of random failures and attacks on Poisson and power-law random networks

    Cl´ emence Magnien, Matthieu Latapy, and Jean-Loup Guillaume. Impact of random failures and attacks on Poisson and power-law random networks. ACM Comput. Surv., 43(3), April 2011. 12

    work page 2011

  8. [7]

    Graph metrics for network robustness—a survey

    Milena Oehlers and Benjamin Fabian. Graph metrics for network robustness—a survey. Mathematics, 9(8):895, 2021

    work page 2021

Show all 28 references
  1. [8]

    Structural robustness and vulnerability of networks

    Alice C Schwarze, Jessica Jiang, Jonny Wray, and Ma- son A Porter. Structural robustness and vulnerability of networks. arXiv preprint arXiv:2409.07498 , 2024

    2024 arXiv

  2. [10]

    Stability of traffic breakup patterns in urban networks

    Marco Cogoni and Giovanni Busonera. Stability of traffic breakup patterns in urban networks. Physical Review E, 104(1):L012301, 2021

    2021

  3. [11]

    Predicting net- work congestion by extending betweenness centrality to interacting agents

    Marco Cogoni and Giovanni Busonera. Predicting net- work congestion by extending betweenness centrality to interacting agents. Physical Review E , 109(4):L044302, 2024

    2024

  4. [12]

    Unveiling the inter-relations be- tween the urban streets network and its dynamic traf- fic flows: Planning implication

    Nimrod Serok, Orr Levy, Shlomo Havlin, and Efrat Blumenfeld-Lieberthal. Unveiling the inter-relations be- tween the urban streets network and its dynamic traf- fic flows: Planning implication. Environment and Plan- ning B: Urban Analytics and City Science , 46(7):1362– 1376, S...

    2019

  5. [13]

    Congestion and centrality in traffic flow on complex networks

    Petter Holme. Congestion and centrality in traffic flow on complex networks. Advances in Complex Systems , 6(02):163–176, 2003

    2003

  6. [14]

    The worldwide air transportation net- work: Anomalous centrality, community structure, and cities’ global roles

    Roger Guimera, Stefano Mossa, Adrian Turtschi, and LA Nunes Amaral. The worldwide air transportation net- work: Anomalous centrality, community structure, and cities’ global roles. Proceedings of the National Academy of Sciences, 102(22):7794–7799, 2005

    2005

  7. [15]

    Can betweenness centrality explain traffic flow

    Aisan Kazerani and Stephan Winter. Can betweenness centrality explain traffic flow. In 12th AGILE inter- national conference on geographic information science , pages 1–9, 2009

    2009

  8. [16]

    Competition for short- est paths on sparse graphs

    Chi Ho Yeung and David Saad. Competition for short- est paths on sparse graphs. Physical review letters , 108(20):208701, 2012

    2012

  9. [17]

    Vu, and Lewi Stone

    Homayoun Hamedmoghadam, Mahdi Jalili, Hai L. Vu, and Lewi Stone. Percolation of heterogeneous flows un- covers the bottlenecks of infrastructure networks. Nat Commun, 12(1):1254, February 2021

    2021

  10. [18]

    Olmos, Serdar C ¸ olak, Sajjad Shafiei, Meead Saberi, and Marta C

    Luis E. Olmos, Serdar C ¸ olak, Sajjad Shafiei, Meead Saberi, and Marta C. Gonz´ alez. Macroscopic dynam- ics and the collapse of urban traffic. Proceedings of the National Academy of Sciences, 115(50):12654–12661, De- cember 2018. Publisher: Proceedings of the National Academ...

    2018

  11. [19]

    Gonz´ alez

    Serdar C ¸ olak, Antonio Lima, and Marta C. Gonz´ alez. Understanding congested travel in urban areas. Nat Commun, 7(1):10793, April 2016

    2016

  12. [20]

    Towards a clas- sification of planar maps

    Alexandre Diet and Marc Barthelemy. Towards a clas- sification of planar maps. Phys. Rev. E , 98(6):062304, December 2018. Publisher: American Physical Society

    2018

  13. [21]

    True scale-invariant random spatial networks

    David Aldous and Karthik Ganesan. True scale-invariant random spatial networks. Proceedings of the National Academy of Sciences, 110(22):8782–8785, 2013

    2013

  14. [22]

    Eugene Stanley, and Shlomo Havlin

    Daqing Li, Bowen Fu, Yunpeng Wang, Guangquan Lu, Yehiel Berezin, H. Eugene Stanley, and Shlomo Havlin. Percolation transition in dynamical traffic network with evolving critical bottlenecks. Proc Natl Acad Sci USA , 112(3):669–672, January 2015

    2015

  15. [23]

    From the betweenness centrality in street networks to structural invariants in random planar graphs

    Alec Kirkley, Hugo Barbosa, Marc Barthelemy, and Gourab Ghoshal. From the betweenness centrality in street networks to structural invariants in random planar graphs. Nat Commun , 9(1):2501, June 2018. Number: 1 Publisher: Nature Publishing Group

    2018

  16. [24]

    Morphology of travel routes and the organization of cities

    Minjin Lee, Hugo Barbosa, Hyejin Youn, Petter Holme, and Gourab Ghoshal. Morphology of travel routes and the organization of cities. Nat Commun , 8(1):2229, De- cember 2017. Number: 1 Publisher: Nature Publishing Group

    2017

  17. [25]

    Comparison of greenshields, pipes, and van aerde car-following and traffic stream models

    Hesham Rakha and Brent Crowther. Comparison of greenshields, pipes, and van aerde car-following and traffic stream models. Transportation Research Record, 1802(1):248–262, 2002

    2002

  18. [26]

    Detecting the urban traffic network struc- ture dynamics through the growth and analysis of multi- layer networks

    Rui Ding, Norsidah Ujang, Hussain bin Hamid, Mohd Shahrudin Abd Manan, Yuou He, Rong Li, and Jianjun Wu. Detecting the urban traffic network struc- ture dynamics through the growth and analysis of multi- layer networks. Physica A: Statistical Mechanics and its Applications, 50...

    2018

  19. [27]

    Shape of shortest paths in random spatial networks

    Alexander P Kartun-Giles, Marc Barthelemy, and Carl P Dettmann. Shape of shortest paths in random spatial networks. Physical Review E , 100(3):032315, 2019

    2019

  20. [28]

    Ultrametricity of optimal transport substates for multiple interacting paths over a square lattice net- work

    Marco Cogoni, Giovanni Busonera, and Gianluigi Zanetti. Ultrametricity of optimal transport substates for multiple interacting paths over a square lattice net- work. Physical Review E , 95(3):030108, 2017

    2017

Pith tools

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