Rapid Quantification of Outdoor Object Visibility in Urban Setting Using Connected-Vehicle Fields of View

Adriana-Simona Mihaita; Artur Grigorev

REVIEW 4 major objections 6 minor 17 references

Rapid Quantification of Outdoor Object Visibility in Urban Setting Using Connected-Vehicle Fields of View

T0 review · 4 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read The paper claims that a data-driven pipeline using connected-vehicle trajectories, densified OpenStreetMap building vertices, and BallTree spatial indexing can objectively quantify how often each building corner is visible to passing…

desk verdict A sensible extension of the SmartAdP pipeline to connected-vehicle data, with an overclaimed log-normal fit and a circular FoV proxy that needs more scrutiny. read the letter →

arxiv 2506.03365 v3 pith:OCAFSAWS submitted 2025-06-03 eess.SY cs.CVcs.SYstat.CO

Artur Grigorev , Adriana-Simona Mihaita This is my paper

classification eess.SYcs.CVcs.SYstat.CO

keywords connectedvehicletrajectoriesvisibilityhotspotsBallTreespatialindexingOpenStreetMaplog-normaldistributionout-of-homeadvertisingurbananalyticsfield-of-viewapproximation

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 claims that a city-wide map of which building corners drivers actually see can be computed from two open data streams: connected-vehicle trajectories and OpenStreetMap building footprints. The method interpolates each vehicle's path to 5-second positions, draws a 100-metre-diameter circle 50 metres ahead of each position, and counts how many densified building vertices fall inside that circle, using a BallTree spatial index to keep the queries fast. Applied to roughly 2000 trips through an inner-Sydney suburb, the counts show that visibility is highly concentrated at a few 'visual hotspots' and that the distribution of counts across 34,495 vertices follows a log-normal curve. If correct, this gives advertisers, planners, and navigation services a quantitative, objective ranking of roadside exposure instead of traffic-volume proxies or subjective site surveys.

What carries the argument

The BallTree spatial index over radian-converted, densified OpenStreetMap building vertices, queried with query_radius under the haversine metric at angular radius R/EARTH_RADIUS for each viewing circle centre computed 50 m ahead of the vehicle bearing. This turns each visibility check into an O(logN) radius query instead of brute-force point-in-circle tests on all vertex pairs. The viewing circle itself is the second piece of machinery: a 100-metre-diameter circle centred 50 m ahead was selected as the best circular approximation of a 60-degree, 100-metre sensor sector by Sørensen-Dice coefficient (0.73).

What would settle it

Take a set of intersections (e.g., Bourke and Ebsworth Streets in the study area) and compare the paper's computed vertex visibility counts against ground-truth visibility ratings from a camera- or eye-tracking-based occlusion study of the same vertices; if the rank correlation between computed and observed visibility is weak, the circular field-of-view assumption fails.

Watch

Extended reading notes

Core claim

The central discovery is the combination of a specific geometric proxy and a scalable query structure: a forward-projected viewing circle (100 m diameter, centred 50 m ahead of each 5-second vehicle position, chosen by Dice-coefficient comparison against a 60-degree sector) checked against densified building vertices through a BallTree using the haversine metric. This yields per-vertex visibility counts whose distribution is heavily skewed: a small fraction of vertices captures a large share of total visibility (the top 10% accounts for about 39.5% of counts), and the aggregated counts are best described by a log-normal distribution (K-S statistic 0.0489) among the fitted candidates. The paper presents this as an objective, data-driven alternative to static traffic counts and Opportunity-To-See estimates, demonstrated in the Waterloo area of Sydney.

Load-bearing premise

The entire ranking rests on the assumption that a 100-metre-diameter circle centred 50 metres ahead of each 5-second vehicle position captures what a driver can actually see; the circle ignores occlusion, viewing angle, and attention, so if it is a poor proxy the hotspot rankings are systematically biased.

Editorial extensions

If this is right

A city can rank every roadside building edge by driver exposure using only GPS traces and building footprints, with no traffic counters or manual surveys.
The log-normal fit gives a statistical baseline: locations in the upper tail (roughly top 10%) account for about 39.5% of all exposure events, so placement decisions can focus on a short list of candidate vertices.
The O(logN) query structure makes the computation feasible at city scale: millions of trajectory points against thousands of building vertices run in sub-second per-point time, enabling repeated re-runs as new trip data arrives.
If the circle proxy is accepted, the same pipeline yields a dynamic visibility metric that can be recomputed for different time windows or traffic conditions, supporting temporal placement strategies.
The method supplies a quantitative basis for calibrating or replacing OTS-style exposure metrics used in out-of-home media planning.

Reading between the lines

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

The log-normal shape suggests visibility arises from multiplicative, compounding factors (traffic volume, geometry, speed, building prominence). A practical consequence the paper does not draw: the same functional form could be used as a null model to flag vertices whose observed visibility departs from the fitted curve, signalling occlusions or data errors.
Because the circle ignores occlusion, the method systematically overstates visibility behind tall structures; coupling the same trajectory data with a digital elevation model or simple building-height layer would produce a corrected ranking, and the Bourke/Ebsworth intersection example in the paper is a natural test case.
The pipeline transfers directly to other road users: pedestrian or micromobility traces could yield foot-traffic exposure maps, and the field-of-view proxy would need only a different radius and forward offset.
A sensitivity test the paper leaves implicit: recomputing the rankings with a 60-degree sector instead of the best-fit circle would show whether hotspot identification is robust to the FoV geometry; the Dice coefficient alone does not guarantee rank stability.

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.

Desk Editor's Note

read the letter

The thing to know about this paper is that the engineering is fine and the headline claim is wobbly. It extends the SmartAdP recipe—trajectory data plus spatial indexing for billboard placement—to connected-vehicle trajectories and building vertices from OSM, adding a forward-projected circular FoV proxy. That is a legitimate, clearly described extension. The pipeline is sensible, the algorithm is explicit, and the hotspot maps look plausible. The authors also deserve credit for a candid limitations section that admits the 2D circle ignores occlusion and even names a specific intersection where a tall building hides a high-count vertex.

What is actually new: the combination of Compass IoT connected-vehicle data, 5-second interpolated trajectories with bearing, densified OSM building vertices (<=10m spacing), and BallTree radius queries is not in the cited prior work. SmartAdP used taxi data and different exposure metrics. The efficiency story is credible.

The soft spots are real but not fatal. First, the log-normal claim is overstated. Table 4 reports K-S p-values below 1e-7 for every fitted distribution, including log-normal. Strictly, that rejects all of them. The paper should have said "log-normal gives the best fit among poor fits" rather than "conforms to a log-normal model." That is a correctable statistical overclaim.

Second, the circular FoV proxy is the more fundamental issue. The stress-test note is right: the chosen circle at xend=100m has 58.6% excess area over the 60-degree sector, and the Bourke/Ebsworth example is a concrete counterexample to "objective visibility" as the abstract phrases it. This can reorder hotspot rankings. The paper acknowledges the limitation, but then still calls the metric objective in the abstract and conclusion. The method is a decent 2D baseline for exposure, not a true visibility measure, and that should be said up front.

Third, there is no code, no data, and no external validation, and the five free parameters (circle radius, forward distance, interpolation interval, densification spacing, distribution parameters) get no sensitivity analysis. For an applied methods paper, that is a moderate gap, not a fatal one.

Overall, the paper is a useful contribution within urban analytics and OOH placement. It deserves a serious referee: the pipeline is sound, the data source is interesting, and the flaws are fixable. I would send it to review with requests to (1) reframe the distributional claim, (2) add sensitivity analysis on the FoV parameters, and (3) soften "objective" to "baseline 2D proxy." I wouldn't cite it myself in the next year unless I work on this exact problem, but I'd bring it to a reading group to discuss the fit-rejection trap.

Referee Report

4 major / 6 minor

Summary. The paper proposes a computational pipeline for quantifying the visibility of building vertices from connected vehicle trajectory data. Vehicle positions are interpolated to 5-second intervals, a circular approximation of a driver's forward field of view is projected 50 meters ahead, and building vertices densified from OpenStreetMap footprints are counted within these circles using a BallTree spatial index. Aggregated counts are then used to identify 'visual hotspots' and to fit a log-normal distribution. The paper reports two empirical findings (spatial concentration and log-normality) and claims a scalability advantage for the BallTree approach. The authors explicitly acknowledge that the 2D circular approximation ignores 3D occlusion, and they give a concrete example where a tall building blocks visibility to a point that nonetheless receives a high count.

Significance. If the pipeline is reliable, it offers a scalable, data-driven alternative to static traffic-count-based site selection, with potential applications in out-of-home advertising, urban design, and navigation. The use of real trajectory data and open building data, the explicit geometric comparison of candidate approximations, and the provision of a detailed algorithm are notable strengths. However, the empirical claims are not yet fully supported: the reported goodness-of-fit p-values statistically reject the log-normal model, the circular field-of-view proxy has 58.6% excess area and is contradicted by the paper's own occlusion example, and no sensitivity analysis or runtime benchmarks are provided. These gaps limit the current evidentiary support for the headline conclusions, though they are addressable in a revision.

major comments (4)

[§3.5, Table 4] The manuscript claims in the Abstract, §4, and §5 that aggregated visibility counts 'conform to' or 'strongly adhere to' a Log-Normal distribution, yet Table 4 reports K-S p-values effectively zero (<1e-7) for all fitted distributions, including Log-Normal. With N=34,495, a p-value at that level rejects the fitted model at any conventional significance level. The claim should be restricted to 'the best among the tested families' as stated in §3.5, and the Abstract and Conclusion wording should be revised accordingly. Additionally, the reported Log-Normal parameters (loc=-23.5784, scale=263.5925, s=1.0332) imply a location parameter far below the minimum visibility count of zero; the authors should explain how a shifted log-normal with a negative location parameter is compatible with non-negative count data.
[§3, Table 3; §5] The circular approximation of the 60-degree field of view has 58.6% excess area relative to the true sector (Table 3), meaning the viewing circle includes a large lateral region outside the driver's nominal field of view. At 50 meters ahead the circle is 100 meters wide, while the sector is about 57.7 meters wide. This is not a minor distortion in dense urban geometry; it can systematically overcount vertices at large lateral angles and reorder hotspot rankings. The paper's Section 5 example at Bourke Street and Ebsworth Street is a concrete instance: a high visibility count is assigned to a location that is physically occluded. Since all downstream results (hotspot ranking, Pareto-style statistics, and the distribution fit) are computed from these circle-based counts, the claim of an 'objective' visibility metric is not supported without either a sensitivity analysis across field-of-view definitions or an occlusion-aware refinement.
[§3.1, Algorithm 1] The claimed computational advantage of the BallTree (O(log N) radius queries and 'significantly outperforming brute-force geometric checks') is not empirically tested. Table 2 reports arithmetic operation counts for point-in-shape tests, but this does not demonstrate the end-to-end performance of the BallTree query versus a brute-force scan over all trajectory circles. Without a runtime benchmark (e.g., wall-clock time, queries per second, or scaling with the number of points), the central scalability contribution remains unverified. The authors should add a simple experiment comparing BallTree query times to naive point-in-circle loops, or at least provide a more rigorous complexity analysis that accounts for the actual query_radius behavior.
[§3.2, §3.3] No sensitivity analysis or uncertainty quantification is provided for the free parameters: viewing circle radius (50 m), forward projection distance (50 m), trajectory interpolation interval (5 s), and building-edge densification spacing (10 m). These choices are arbitrary to some degree, and because the visibility counts are deterministic functions of them, the reported hotspot rankings and the fitted distribution may change under alternative choices. For instance, varying the circle radius by ±20% or the interpolation interval between 1 s and 15 s could alter the top quantiles and the fitted distribution. The paper should demonstrate that the qualitative findings (hotspot concentration and the distributional shape) are robust to reasonable variations of these parameters.

minor comments (6)

[Throughout] The manuscript contains numerous typographical ligature artifacts: 'e fficient', 'di ffers', 'tra ffic', 'o ffers' appear in the text; these should be corrected to 'efficient', 'differs', 'traffic', and 'offers'.
[§2.1] The sentence introducing the bounding box ends with a colon followed by an empty line before Table 1; the text should explicitly reference Table 1 or include the coordinate values in the sentence.
[Figures 1, 4, 5] In the provided manuscript text, Figures 1, 4, and 5 are represented only by captions without the embedded images. If the figures are present in the submission, please ensure the captions are placed adjacent to the images; otherwise, add the figures. Without the visualizations, the claims about spatial clustering and the distribution cannot be fully assessed.
[Table 5] Table 5 appears immediately after Table 4 with no caption; it appears to be a table note. It should be formatted as a table note directly under Table 4.
[References] Reference [15] (Xiao et al., pedestrian evacuation) does not appear to support the claim about Australia's MOVE system in §2; please verify the citation or replace it with an appropriate source on MOVE or OTS metrics.
[§3.4] The statement that 'top 10% of points gains around 39.5% of total visibility' should include the exact quantile definition and how the percentage is computed; ideally, a measure of uncertainty or a sensitivity check would strengthen this claim.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: visibility counts are computed directly from trajectory and building data, and the distributional claim is an in-sample fit rather than a derived prediction.

full rationale

The paper's derivation chain is self-contained: visibility counts are produced by BallTree radius queries over densified OSM building vertices against 50m-radius circles centered 50m ahead of interpolated vehicle bearings (Algorithm 1, Section 3.1). No fitted parameter is used as an input to compute these counts; the circle radius (50m) and projection distance (50m) are fixed modeling choices justified separately by a Dice-coefficient Monte Carlo analysis in Table 3, not by the visibility data. The hotspot/concentration claims are direct empirical summaries of the computed counts, not predictions forced by a fitted quantity. The log-normal 'finding' in Section 3.5 is a distribution fit to the same 34,495 counts it describes (Table 4), and the paper's own K-S p-values are effectively zero, so this is a descriptive overstatement rather than a circular derivation; it does not feed back into the visibility computation. The paper contains no self-citations that carry a load-bearing argument and no uniqueness theorems imported from prior work. Section 5 acknowledges the 2D, occlusion-free FoV limitation, which is a validity caveat, not a circularity.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The central contribution is a pipeline, so most free parameters are method choices rather than fitted physical constants. No novel physical entities are postulated. The main burden is the unvalidated mapping from geometric circles to human visibility.

free parameters (5)

viewing circle radius = 50 m
Hand-selected, informed only by a Dice coefficient comparison against a 60-degree, 100 m sector; no perceptual data.
forward projection distance = 50 m
Circle center placed 50 m ahead of each vehicle; tied to radius and not independently justified.
trajectory interpolation interval = 5 s
Data pre-processing choice; affects temporal density of viewing circles.
building edge densification spacing = 10 m
Spacing chosen for vertex representation; not sensitivity-tested.
lognormal fit parameters = loc=-23.5784, scale=263.5925, s=1.0332
Fitted to the same visibility counts used to claim log-normality.

assumptions (4)

domain assumption Five-second interpolated trajectory points with computed bearing capture relevant driver positions.
Invoked in Section 3 for all visibility queries.
domain assumption A circular FoV centered 50 m ahead is a valid proxy for driver forward visibility.
Section 3 Table 3; acknowledged limitation in Section 4.
domain assumption Densified building vertices represent visibility-relevant features of buildings.
Section 3; no validation against human perception.
domain assumption Visibility counts aggregate linearly across trips and are unbiased by traffic sampling.
Section 3.2 and Section 4 acknowledge sample representativeness as an open question.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Rapid Quantification of Outdoor Object Visibility in Urban Setting Using Connected-Vehicle Fields of View." pith.science (2026). https://pith.science/paper/OCAFSAWS

@misc{pith2026250603365,
  author       = {Pith},
  title        = {Pith review of: Rapid Quantification of Outdoor Object Visibility in Urban Setting Using Connected-Vehicle Fields of View},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OCAFSAWS}},
  note         = {Machine review of arXiv:2506.03365}
}

read the original abstract

Identifying locations that offer maximum visual exposure to passing vehicular traffic is a core problem in urban analytics, with applications spanning urban design, navigation, location-based services, and the placement of street-level assets. Traditional site selection methods often rely on static traffic counts or subjective assessments. This research introduces a data-driven methodology to objectively quantify location visibility by analyzing large-scale connected vehicle trajectory data within urban environments. We model the dynamic driver field-of-view using a forward-projected visibility area for each vehicle position derived from interpolated trajectories. By integrating this with building vertex locations extracted from OpenStreetMap, we quantify the cumulative visual exposure, or ``visibility count'', for thousands of potential points of interest along roadways. The core technical contribution involves the construction of a BallTree spatial index over building vertices. This enables highly efficient (O(logN) complexity) radius queries to determine which vertices fall within the viewing circles of millions of trajectory points across numerous trips, significantly outperforming brute-force geometric checks. Analysis reveals two key findings: 1) Visibility is highly concentrated, identifying distinct 'visual hotspots' receiving disproportionately high exposure compared to average locations. 2) The aggregated visibility counts across vertices conform to a Log-Normal distribution.

Figures

Figures reproduced from arXiv: 2506.03365 by the authors.

**Figure 2.** Visual comparison of circular approximations to the true sensor sector. The circle with [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

**Figure 3.** Example interpolated trajectory trace and associated viewing circles for a single trip (ID ending 2174) within the Waterloo study area. Blue line indicates [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗

Figures from the paper (2 more)

**Figure 4.** Figure 4: Frequency distribution of aggregated visibility counts per building [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]

**Figure 5.** Figure 5: Spatial distribution of Building Vertex Visibility, Highlighting Top Quantiles. Di [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

17 extracted references · 17 canonical work pages

[1]

Masters thesis, Ehrenberg-Bass Institute for Mar- keting Science, University of South Australia, 2021

Talbot, E., The Great Outdoors: An Investigation into the Value of Out- of-Home Advertising. Masters thesis, Ehrenberg-Bass Institute for Mar- keting Science, University of South Australia, 2021

work page 2021
[2]

Clow, K. E. and D. Baack, Integrated Advertising, Promotion, and Mar- keting Communications. Pearson, 10th ed., 2023

work page 2023
[3]

Postigo, N

Farah, H., I. Postigo, N. Reddy, Y . Dong, C. Rydergren, N. Raju, and J. Olstam, Modeling Automated Driving in Microscopic Tra ffic Simula- tions for Traffic Performance Evaluations: Aspects to Consider and State of the Practice. IEEE Transactions on Intelligent Transportation Systems, V ol. 24, No. 6, 2023, pp. 6558–6574

work page 2023
[4]

S., Development and evaluation of simulation models for as- sessing the impacts of connected and automated vehicles, 2024

Matin, A. S., Development and evaluation of simulation models for as- sessing the impacts of connected and automated vehicles, 2024

work page 2024
[5]

Huynh, N., V . L. Cao, R. Wickramasuriya, B. M., P. Perez, and J. Barth- lemy, An Agent Based Model for the Simulation of Road Tra ffic and Transport Demand in A Sydney Metropolitan Area, 2014

work page 2014
[6]

Senanayake, G. P., M. Kieu, Y . Zou, and K. Dirks, Agent-based simu- lation for pedestrian evacuation: A systematic literature review. Interna- tional Journal of Disaster Risk Reduction, V ol. 111, 2024, p. 104705

work page 2024
[7]

Jankowski, and M

Maecki, K., J. Jankowski, and M. Szkwarkowski, Modelling the Impact of Transit Media on Information Spreading in an Urban Space Using Cel- lular Automata. Symmetry, V ol. 11, 2019, p. 428

work page 2019
[8]

In Handbook of Mobility Data Mining (H

Handbook of Mobility Data Mining, V olume 3: Mobility Data-Driven Applications. In Handbook of Mobility Data Mining (H. Zhang, ed.), El- sevier, 2023

work page 2023

Show all 17 references

[9]

Phd the- sis, School of Information Technology and Electrical Engineering, 2020

Chao, P., A study on map-matching and map inference problems. Phd the- sis, School of Information Technology and Electrical Engineering, 2020

2020
[10]

Kumar, S

Rathore, P., D. Kumar, S. Rajasegarar, M. Palaniswami, and J. C. Bezdek, A Scalable Framework for Trajectory Prediction. IEEE Transactions on Intelligent Transportation Systems, V ol. 20, No. 10, 2019, pp. 3860–3874

2019
[11]

Xue, A. Y ., J. Qi, X. Xie, R. Zhang, J. Huang, and Y . Li, Solving the data sparsity problem in destination prediction. The VLDB Journal , V ol. 24, No. 2, 2015, pp. 219–243

2015
[12]

Mesbah, J

Faroqi, H., M. Mesbah, J. Kim, and A. Khodaii, Targeted Advertising in the Public Transit Network Using Smart Card Data.Networks and Spatial Economics, V ol. 22, No. 1, 2022, pp. 97–124

2022
[13]

Fong, S., K. Cho, W. Ip, and E. Liu, Identifying Optimal Spatial Groups for Maximum Coverage in Ubiquitous Sensor Network by Using Clus- tering Algorithms. International Journal of Distributed Sensor Networks, V ol. 2013, 2013

2013
[14]

Zheng, X

Liu, S., Y . Zheng, X. Liu, W. Ke, V . W. Zheng, and H. Qu, SmartAdP: Visual Analytics of Large-scale Taxi Trajectories for Selecting Billboard Locations. In 2016 IEEE Pacific Visualization Symposium (PacificVis) , 2016, pp. 48–55. 8 Figure 5: Spatial distribution of Building ...

2016
[15]

Xiao, Y ., J. Xu, M. Chraibi, J. Zhang, and C. Gou, A generalized trajectories-based evaluation approach for pedestrian evacuation models. Safety science, V ol. 147, 2022, p. 105574

2022
[16]

Jolivet, C

Kerouanton, C., L. Jolivet, C. Perrin-Malterre, and A. Loison, Eye- catching or breath-catching: Role and landscape attributes of pauses dif- fers among hikers profile when rambling in a French mountainous area. Journal of Outdoor Recreation and Tourism, V ol. 46, 2024, p. 100734

2024
[17]

Gschwend, C., Relating movement to geographic context: e ffects of preprocessing, relation methods and scale . Ph.D. thesis, University of Zurich, 2015. 10

2015

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.

[1] [1]

Masters thesis, Ehrenberg-Bass Institute for Mar- keting Science, University of South Australia, 2021

Talbot, E., The Great Outdoors: An Investigation into the Value of Out- of-Home Advertising. Masters thesis, Ehrenberg-Bass Institute for Mar- keting Science, University of South Australia, 2021

work page 2021

[2] [2]

Clow, K. E. and D. Baack, Integrated Advertising, Promotion, and Mar- keting Communications. Pearson, 10th ed., 2023

work page 2023

[3] [3]

Postigo, N

Farah, H., I. Postigo, N. Reddy, Y . Dong, C. Rydergren, N. Raju, and J. Olstam, Modeling Automated Driving in Microscopic Tra ffic Simula- tions for Traffic Performance Evaluations: Aspects to Consider and State of the Practice. IEEE Transactions on Intelligent Transportation Systems, V ol. 24, No. 6, 2023, pp. 6558–6574

work page 2023

[4] [4]

S., Development and evaluation of simulation models for as- sessing the impacts of connected and automated vehicles, 2024

Matin, A. S., Development and evaluation of simulation models for as- sessing the impacts of connected and automated vehicles, 2024

work page 2024

[5] [5]

Huynh, N., V . L. Cao, R. Wickramasuriya, B. M., P. Perez, and J. Barth- lemy, An Agent Based Model for the Simulation of Road Tra ffic and Transport Demand in A Sydney Metropolitan Area, 2014

work page 2014

[6] [6]

Senanayake, G. P., M. Kieu, Y . Zou, and K. Dirks, Agent-based simu- lation for pedestrian evacuation: A systematic literature review. Interna- tional Journal of Disaster Risk Reduction, V ol. 111, 2024, p. 104705

work page 2024

[7] [7]

Jankowski, and M

Maecki, K., J. Jankowski, and M. Szkwarkowski, Modelling the Impact of Transit Media on Information Spreading in an Urban Space Using Cel- lular Automata. Symmetry, V ol. 11, 2019, p. 428

work page 2019

[8] [8]

In Handbook of Mobility Data Mining (H

Handbook of Mobility Data Mining, V olume 3: Mobility Data-Driven Applications. In Handbook of Mobility Data Mining (H. Zhang, ed.), El- sevier, 2023

work page 2023