Dealing with uncertainty in agent-based models for short-term predictions

Alison Heppenstall; Le-Minh Kieu; Nicolas Malleson

REVIEW 2 major objections 6 minor 47 references

Dealing with uncertainty in agent-based models for short-term predictions

T0 review · 2 major / 6 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read This paper claims that pairing one-time parameter calibration with real-time particle filtering keeps agent-based bus models accurate for short-term predictions, and demonstrates the combination on synthetic GPS data.

desk verdict A clean synthetic demonstration that CEM calibration plus particle-filter DA improves ABM bus-location forecasts, but the missing PF-only control keeps the 'combination' story incomplete. read the letter →

arxiv 1908.08288 v1 pith:QDMF3YAS submitted 2019-08-22 cs.MA

Le-Minh Kieu , Nicolas Malleson , Alison Heppenstall This is my paper · ORCID

classification cs.MA

keywords agent-basedmodeldataassimilationparticlefilterparametercalibrationcross-entropymethodbuslocationpredictionshort-termforecastingidenticaltwinexperiment

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 tries to solve a known weakness of agent-based models: once calibrated on historical data, they drift from reality when the system changes. Its proposal is to add data assimilation on top of calibration, so the model is continuously corrected by streaming observations. Using a synthetic bus route, the paper shows that combining Cross-Entropy calibration with a Particle Filter yields lower location prediction error than no calibration or calibration alone. The intended payoff is a practical route to real-time forecasts of bus locations and arrival times using ABMs, not just one-time simulations.

What carries the argument

The mechanism is a Markovian state-space reformulation of the ABM: the state vector $X_t = [O_t, S_t]$ carries all agent information, and the Particle Filter maintains a weighted particle set $\mathcal{P}_t$ of hypotheses; each particle is one full ABM run state. After a predict step, importance weighting based on the observation vector, and systematic resampling (SIR), the method adds Gaussian roughening noise to the parameter sub-vector $S_t$ so that the calibrated parameters can continue to evolve with the data. The Cross-Entropy Method supplies the initial calibrated parameter distribution that the filter then tracks. This two-stage design is what carries the claim.

What would settle it

Apply the identical calibration-plus-filter pipeline to real bus GPS data where the underlying dwell-time and demand processes differ from the model family, and compare RMSE against calibration-only baselines; if the Particle Filter no longer improves predictions or the errors diverge, the central claim fails under structural misspecification. A cheaper version is to rerun the paper's own scenarios with BusSim-truth replaced by a truth model using a different boarding process or a different dwell-time formula.

Watch

Extended reading notes

Core claim

The central claim is that an ABM can be made dynamically optimisable for short-term prediction by calibrating its parameters with the Cross-Entropy Method against historical data and then applying a Particle Filter that assimilates real-time observations into the evolving state vector. In the controlled 'identical twin' experiments, where a richer stochastic-dynamic model (BusSim-truth) generates synthetic historical and real-time GPS data, the two simpler companion ABMs (BusSim-deterministic and BusSim-stochastic) achieve much lower RMSE of predicted bus location when the two stages are combined than in either baseline. For example, at one demand level the RMSE falls from about 335 metres with no calibration to about 125 metres after calibration, and to about 49 metres after calibration plus filtering. The paper interprets this as evidence that the drift caused by stochastic and dynamic change can be corrected at run time, and that the Particle Filter is a suitable data-assimilation method for the non-linear, non-Gaussian structure of ABMs.

Load-bearing premise

The weakest load-bearing premise is that the identical-twin test is diagnostic of real performance: the synthetic 'real-time' data are generated by a truth model whose passenger-boarding, dwell-time, and dynamic-change mechanisms are from the same family as the simpler companion models, so the experiment tests correction of known stochasticity, not reaction to a structurally different bus system.

Editorial extensions

If this is right

Real-time bus location and arrival-time prediction can be supported by an ABM that is continuously updated, rather than only by a static, once-calibrated simulator.
The method produces a particle ensemble, giving not a single forecast but a distribution that quantifies state uncertainty.
In the synthetic experiments, the combined calibration-plus-filter approach yields the lowest RMSE across a range of stochasticity and dynamicity settings.
The framework is positioned for use in passenger information systems and Intelligent Transport Systems, where live forecasts of bus positions are operationally valuable.
The paper's framing implies that any ABM expressible as a Markovian state-space model can inherit data-assimilation methods without requiring an analytical linearisation.

Reading between the lines

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

Because the truth model and the companion models share the same family of boarding, dwell-time, and dynamic-change equations, the head-to-head comparison measures correction of known stochasticity, not reaction to structural misspecification; a real deployment would need to test against a differently structured bus system.
The roughening step means parameter uncertainty is represented by particle diversity, so the method's benefit likely grows with system volatility, at least until the jitter-noise scale is miscalibrated.
The same two-stage recipe could transfer to other streaming ABMs, such as crowd movement or epidemic spread, whenever a finite state vector and an observation vector can be defined.
The concluding section mentions a 'Scenario 4 (only Particle Filtering)' comparison, but no such experiment appears in the numerical section or in the reported sensitivity table, so the marginal contribution of calibration relative to filtering alone is not empirically pinned down by this paper.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

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 you should know up front: the paper does what it says for the experiments it actually runs. Calibration plus particle filtering (Scenario 3) beats both no-calibration (Scenario 1) and calibration-only (Scenario 2) across a sensitivity sweep, and Table 3 is consistent with that claim. The code and synthetic data are on GitHub, and the experimental design is carefully controlled. This is a reproducible, useful case study for ABM-based real-time forecasting.

The genuinely new piece is the explicit sequence: calibrate a simpler companion ABM once on historical synthetic data, then run a particle filter with parameter jittering to keep the model aligned with streaming observations. Neither ingredient is new—CEM calibration and PF for ABMs both exist in the literature—but the combination, applied to bus operations and evaluated with sensitivity analysis over demand and dynamicity, is a legitimate contribution.

The soft spots are real, though not fatal. The stress-test note holds up: Section 4 defines only Scenarios 1–3, Table 3 has no PF-only column, and yet the conclusion claims the framework outperforms an undefined 'only Particle Filtering scenario (Scenario 4)'. That is a dangling reference, probably a leftover from an earlier draft. It does not undermine the comparison that is actually reported, but it leaves open whether calibration is load-bearing once DA is in the loop. If an uncalibrated PF achieves similar RMSE, the 'combination' narrative is much weaker. The identical-twin design is a second limitation: because the truth model and the companion models share the same family of boarding, dwell-time, and movement assumptions, the experiment measures correction of known stochasticity, not robustness to structural misspecification. The paper acknowledges this, but it is still the ceiling on how much we can infer about real systems. Also, there are no error bars, and the jitter noise and particle count are not reported. Those are minor and fixable.

Who gets value from this? ABM practitioners working on transit or other real-time prediction problems will find a concrete, reproducible recipe. Methodologists will see it as a moderate incremental step, not a breakthrough. The paper deserves a serious referee: I would send it to review with the expectation of a revision that either adds the PF-only control or removes the Scenario 4 claim, reports the missing tuning parameters, and adds error bars. Engage with it, but read the conclusions with one eye on what the experiments actually covered.

Referee Report

2 major / 6 minor

Summary. This paper proposes a framework for short-term prediction with agent-based models under uncertainty, combining offline parameter calibration (Cross-Entropy Method) with online data assimilation (Particle Filter). The authors build a stochastic, dynamic 'BusSim-truth' model to generate synthetic historical and real-time GPS bus data, then calibrate two simpler companion models (BusSim-deterministic and BusSim-stochastic). They evaluate three scenarios: no calibration, calibration only, and calibration plus particle filtering, and report RMSE sensitivity analyses over passenger demand and dynamic change rate. Table 3 shows the calibration-plus-filter scenario has lower RMSE than the other two scenarios. The paper concludes that the combined framework outperforms both the no-calibration benchmark and a particle-filter-only scenario, and discusses applications to passenger information systems.

Significance. If the central claim holds, the framework is a useful step toward using ABMs with streaming data, addressing a recognized limitation of the field. The paper's strengths include a controlled synthetic evaluation with a more complex truth model, a sensitivity analysis over key dynamic and stochastic parameters, explicit model documentation in appendices, and public release of source code and data. The main gap is that the reported experiments do not include a particle-filter-only control, so the specific claim that the combination of calibration and DA is superior to DA alone is not directly supported. Nevertheless, the study provides a clear demonstration that calibration plus particle filtering improves prediction relative to a calibrated static model, which is valuable for future work in dynamic ABM calibration.

major comments (2)

[Section 6 and Section 4] The conclusion states that the framework outperforms 'the no calibration scenario (Scenario 1) and only Particle Filtering scenario (Scenario 4)', but Scenario 4 is never defined in Section 4 and no results for it appear in Table 3. The methodology and sensitivity analysis cover only Scenarios 1-3, so the paper contains no experiment in which a particle filter is applied without prior CEM calibration. As a result, the claimed advantage of the combination over a particle filter alone is not established; adding this control and reporting its RMSE is necessary to support the central claim.
[Section 4.5 and Eq. (8)] The evaluation in Section 4.5 and the RMSE defined in Eq. (8) do not clearly separate state estimation (filtering) from short-term forecasting. If the particle filter at time t is updated with the observation at time t, then the reported RMSE largely measures the accuracy of the filtered state rather than the skill of predictions made ahead of time. The paper should either clarify that the reported errors are one-step-ahead or multiple-step-ahead forecast errors, or re-run the evaluation at explicit lead times, so that the 'short-term predictions' claim in the title and abstract is directly supported.

minor comments (6)

[Abstract] The keyword line reads 'First keyword· Second keyword· More' and should be replaced with actual keywords.
[Section 3.3] In the last paragraph of Section 3.3, 'should be the best guest of the system state' should read 'best guess'.
[Appendix A, Eqs. (16)-(17)] Equation (17) is missing a closing parenthesis, and the text says that xi > 0 'represents an increase in passenger demand and trafﬁc speed' while the equations and Figure 5 show that positive xi reduces traffic speed and increases arrival rate; please reconcile the wording.
[Table 2] The 'FleetSize' row in Table 2 gives a parameter name but no value; please provide the number of bus agents used in the experiments.
[Eq. (8)] The RMSE formula in Eq. (8) has no index for the bus or replication; clarify whether the error is averaged over buses and over the ten replications, or reported per bus.
[Section 4.6, Table 3] Table 3 reports only mean RMSE over 10 replications; please include standard deviations or confidence intervals so the reader can judge the significance of the differences between scenarios.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the headline comparison is against separately generated synthetic real-time data, and no reported prediction reduces by construction to fitted parameters or a self-citation.

full rationale

The derivation chain is not circular in the sense defined by the review criteria. The companion models are calibrated with CEM to one synthetic 'historical' dataset (Section 3.2, Eq. 3), while the RMSE values in Table 3 (Section 4.6, Eq. 8) are computed against a separately generated synthetic 'real-time' dataset produced by BusSim-truth. The paper states that the companion models 'will not know the parameters of BusSim-truth', so the calibrated parameters are not definitionally the target of the prediction. The PF step assimilates observations into the state-space model, but no equation sets the predicted trajectory equal to the assimilated observations; the improvement is reported relative to explicit Scenario 1 (no calibration) and Scenario 2 (calibration only) baselines. The only relevant self-citation is [47] (Ward, Evans, Malleson), used as background on dynamic calibration of ABMs; it is not the load-bearing justification for the proposed combination, and no uniqueness theorem or ansatz is imported from it. Two weaknesses exist, but they are not circularity: the identical-twin design means the synthetic truth and the candidate models share the same modelling family, limiting external validity; and Section 6 claims outperformance over a 'Scenario 4' (PF only) that is never defined or reported in Section 4 or Table 3. These are evidence/completeness gaps that weaken the conclusion, not steps where a prediction reduces to its inputs by definition.

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

The central claim rests on a small number of modeled assumptions (Markovian state space, Poisson arrivals, linear dwell time, single xi for dynamics) and on several un-reported tuning parameters (jittering noise, particle count, CEM hyperparameters). No new physical entities are introduced. The identifiability of the parameters is not discussed.

free parameters (4)

jittering noise standard deviation (sigma_jitter) = not reported
Section 3.3 adds Gaussian white noise with a 'predefined standard deviation' to the model parameters during resampling; its value controls exploration of parameter space and likely affects prediction accuracy, but no value is given.
number of particles (NP) = not reported
Equation 5 defines a set of NP particles; the number used in Scenario 3 is never stated, and particle count determines filter accuracy and cost.
CEM hyperparameters (rho, alpha, initial mu/sigma) = not reported
Appendix B pseudocode lists elite ratio rho, step size alpha, and initial distribution parameters without specifying values used in the experiments.
calibration objective weighting (mean vs std) = equal weights
Equation 3 sums absolute differences in mean and standard deviation of bus locations without justification for equal weighting; changing the weights changes the calibrated parameters.

assumptions (5)

domain assumption The bus route system can be represented as a Markovian state-space model with a fixed-size state vector.
Section 3.3 formulates the ABM as a state-space model and requires Markovian structure for the particle filter; real bus systems include non-Markovian elements such as driver memory and schedule recovery.
domain assumption Boarding passenger counts follow a homogeneous Poisson process.
Equation 11 uses a Poisson process with a fixed arrival rate per stop; real passenger arrivals are time-dependent (non-homogeneous), as acknowledged in Appendix A.
domain assumption Dwell time is a linear function of boarding and alighting counts with fixed coefficients.
Equation 14 adopts TCQSM's linear dwell time model with theta = [3, 1, 0.85] seconds; the coefficients are taken from prior transit literature and assumed valid for this bus system.
ad hoc to paper Dynamic changes in demand and traffic speed are captured by a single deterministic parameter xi with linear equations.
Equations 16 and 17 linearly scale traffic speed and arrival rate by xi; this is a simplification made for this study, and the paper notes a time-dependent xi would be more realistic.
standard math The particle filter with SIR resampling and Gaussian parameter jittering is a valid approximation for this non-linear, non-Gaussian ABM.
Section 3.3 justifies PF over Kalman filters for non-linear ABMs; this is a standard result in the DA literature.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Dealing with uncertainty in agent-based models for short-term predictions." pith.science (2026). https://pith.science/paper/QDMF3YAS

@misc{pith2026190808288,
  author       = {Pith},
  title        = {Pith review of: Dealing with uncertainty in agent-based models for short-term predictions},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/QDMF3YAS}},
  note         = {Machine review of arXiv:1908.08288}
}

read the original abstract

Agent-based models (ABM) are gaining traction as one of the most powerful modelling tools within the social sciences. They are particularly suited to simulating complex systems. Despite many methodological advances within ABM, one of the major drawbacks is their inability to incorporate real-time data to make accurate short-term predictions. This paper presents an approach that allows ABMs to be dynamically optimised. Through a combination of parameter calibration and data assimilation (DA), the accuracy of model-based predictions using ABM in real time is increased. We use the exemplar of a bus route system to explore these methods. The bus route ABMs developed in this research are examples of ABMs that can be dynamically optimised by a combination of parameter calibration and DA. The proposed model and framework can also be used in an passenger information system, or in an Intelligent Transport Systems to provide forecasts of bus locations and arrival times.

Figures

Figures reproduced from arXiv: 1908.08288 by the authors.

**Figure 1.** Study workflow. The study workflow generally consists of 2 major steps. It starts with the development of a Markovian ABM of bus route operation that will be referred as BusSim-truth. BusSim-truth is a hypothetical version of reality and will be used to generate synthetic GPS data of bus locations with timestamps. Two sets of data will be generated. The first represents ‘historical’ GPS data, which are essentially t… view at source ↗

**Figure 2.** Flowchart of BusSim-truth. of the model. New solution candidates are drawn from this distribution and are evaluated. The best candidates are then selected to form a new improved probability distribution of the optimal parameters, until certain criteria are met. CEM is chosen over some other popular optimisation methods in parameter calibration of ABMs, such as Genetic Algorithm [17] and simulated annealing [37], bec… view at source ↗

**Figure 3.** Synthetic ‘historical’ versus ‘real-time’ GPS bus location data. Each coloured line shows the trajectory of one [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗

Figures from the paper (5 more)

**Figure 4.** Figure 4: Synthetic bus GPS trajectory at low and high passenger demand. Red, dashed lines are bus trajectories when [PITH_FULL_IMAGE:figures/full_fig_p010_4.png]

**Figure 5.** Figure 5: Synthetic bus GPS trajectory with two different value of [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]

**Figure 6.** Figure 6: Prediction results from Scenario 1: no calibration [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]

**Figure 7.** Figure 7: Prediction results from Scenario 2: Parameter calibration [PITH_FULL_IMAGE:figures/full_fig_p012_7.png]

**Figure 8.** Figure 8: Prediction results from Scenario 3: Parameter calibration and Particle Filtering [PITH_FULL_IMAGE:figures/full_fig_p013_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

47 extracted references · 46 canonical work pages

[1]

Matsim-t: Architecture and simulation times

Michael Balmer, Marcel Rieser, Konrad Meister, David Charypar, Nicolas Lefebvre, and Kai Nagel. Matsim-t: Architecture and simulation times. In Multi-agent systems for trafﬁc and transportation engineering, pages 57–78. IGI Global, 2009

work page 2009
[2]

Modeling transit trip time using archived bus dispatch system data

Robert L Bertini and Ahmed M El-Geneidy. Modeling transit trip time using archived bus dispatch system data. Journal of transportation engineering, 130(1):56–67, 2004

work page 2004
[3]

Bus arrival time prediction using support vector machines.Journal of Intelligent Transportation Systems, 10(4):151–158, 2006

Yu Bin, Yang Zhongzhen, and Yao Baozhen. Bus arrival time prediction using support vector machines.Journal of Intelligent Transportation Systems, 10(4):151–158, 2006

work page 2006
[4]

Agent based modeling: Methods and techniques for simulating human systems

E Bonabeau. Agent based modeling: Methods and techniques for simulating human systems. Proceedings of the National Academy of Sciences, 99(90003):7280–7287, 2002

work page 2002
[5]

Improved particle ﬁlter for nonlinear problems

James Carpenter, Peter Clifford, and Paul Fearnhead. Improved particle ﬁlter for nonlinear problems. IEE Proceedings-Radar, Sonar and Navigation, 146(1):2–7, 1999

work page 1999
[6]

Data assimilation in the geosciences: An overview of methods, issues, and perspectives

Alberto Carrassi, Marc Bocquet, Laurent Bertino, and Geir Evensen. Data assimilation in the geosciences: An overview of methods, issues, and perspectives. Wiley Interdisciplinary Reviews: Climate Change, 9(5):e535, September 2018

work page 2018
[7]

Mesoscopic modeling of bus public transportation

Oded Cats, Wilco Burghout, Tomer Toledo, and Haris Koutsopoulos. Mesoscopic modeling of bus public transportation. Transportation Research Record: Journal of the Transportation Research Board, (2188):9–18, 2010. 16 A PREPRINT - AUGUST 23, 2019

work page 2010
[8]

Dynamic bus arrival time prediction with artiﬁcial neural networks

Steven I-Jy Chien, Yuqing Ding, and Chienhung Wei. Dynamic bus arrival time prediction with artiﬁcial neural networks. Journal of Transportation Engineering, 128(5):429–438, 2002

work page 2002

Show all 47 references

[9]

Steady-states and kinetics of ordering in bus-route models: connection with the nagel-schreckenberg model

Debashish Chowdhury and Rashmi C Desai. Steady-states and kinetics of ordering in bus-route models: connection with the nagel-schreckenberg model. The European Physical Journal B-Condensed Matter and Complex Systems, 15(2):375–384, 2000

2000
[10]

Crooks and Sarah Wise

Andrew T. Crooks and Sarah Wise. GIS and agent-based models for humanitarian assistance. Computers, Environment and Urban Systems, 41:100–111, 2013

2013
[11]

A headway-based approach to eliminate bus bunching: Systematic analysis and comparisons

Carlos F Daganzo. A headway-based approach to eliminate bus bunching: Systematic analysis and comparisons. Transportation Research Part B: Methodological, 43(10):913–921, 2009

2009
[12]

Calibration/data assimilation approach for integrating grace data into the watergap global hydrology model (wghm) using an ensemble kalman ﬁlter: First results

Annette Eicker, Maike Schumacher, Jürgen Kusche, Petra Döll, and Hannes Müller Schmied. Calibration/data assimilation approach for integrating grace data into the watergap global hydrology model (wghm) using an ensemble kalman ﬁlter: First results. Surveys in Geophysics, 35(6)...

2014
[13]

Waiting time perceptions at transit stops and stations: Effects of basic amenities, gender, and security

Yingling Fan, Andrew Guthrie, and David Levinson. Waiting time perceptions at transit stops and stations: Effects of basic amenities, gender, and security. Transportation Research Part A: Policy and Practice, 88:251–264, 2016

2016
[14]

Richiardi, and Mike Tsionas

Jakob Grazzini, Matteo G. Richiardi, and Mike Tsionas. Bayesian estimation of agent-based models. Journal of Economic Dynamics and Control, 77:26–47, April 2017

2017
[15]

Real-time bus route state forecasting using particle ﬁlter and mesoscopic modeling

Etienne Hans, Nicolas Chiabaut, Ludovic Leclercq, and Robert L Bertini. Real-time bus route state forecasting using particle ﬁlter and mesoscopic modeling. Transportation Research Part C: Emerging Technologies, 61:121– 140, 2015

2015
[16]

Forecasting, structural time series models and the Kalman ﬁlter

Andrew C Harvey. Forecasting, structural time series models and the Kalman ﬁlter. Cambridge university press, 1990

1990
[17]

A. J. Heppenstall, A.J. Evans, and M H Birkin. Genetic algorithm optimisation of an agent-based model for simulating a retail market. Environment and Planning B: Planning and Design, 34:1051–1070, 2007. Cited by 0018

2007
[18]

Numerical analysis of a time-headway bus route model

Scott A Hill. Numerical analysis of a time-headway bus route model. Physica A: Statistical Mechanics and its Applications, 328(1):261–273, 2003

2003
[19]

Analysis of a continuous car-following model for a bus route: existence, stability and bifurcations of synchronous motions

HJC Huijberts. Analysis of a continuous car-following model for a bus route: existence, stability and bifurcations of synchronous motions. Physica A: Statistical Mechanics and its Applications, 308(1):489–517, 2002

2002
[20]

Realistic bus route model considering the capacity of the bus

Rui Jiang, Mao-Bin Hu, Bin Jia, and Qing-Song Wu. Realistic bus route model considering the capacity of the bus. The European Physical Journal B-Condensed Matter and Complex Systems, 34(3):367–372, 2003

2003
[21]

Atmospheric Modeling, Data Assimilation and Predictability

Eugenia Kalnay. Atmospheric Modeling, Data Assimilation and Predictability. Cambridge University Press, 2003

2003
[22]

Prediction intervals to account for uncertainties in travel time prediction

Abbas Khosravi, Ehsan Mazloumi, Saeid Nahavandi, Doug Creighton, and JWC Van Lint. Prediction intervals to account for uncertainties in travel time prediction. IEEE Transactions on Intelligent Transportation Systems, 12(2):537–547, 2011

2011
[23]

Stochastic collective model of public transport passenger arrival process

Le Minh Kieu and Chen Cai. Stochastic collective model of public transport passenger arrival process. IET Intelligent Transport Systems, 12(9), 2018

2018
[24]

Liu, and Wing Hung Wong

Augustine Kong, Jun S. Liu, and Wing Hung Wong. Sequential Imputations and Bayesian Missing Data Problems. Journal of the American Statistical Association, 89(425):278–288, March 1994

1994
[25]

Bus travel time prediction using a time-space discretization approach

B Anil Kumar, Lelitha Vanajakshi, and Shankar C Subramanian. Bus travel time prediction using a time-space discretization approach. Transportation Research Part C: Emerging Technologies, 79:308–332, 2017

2017
[26]

Lewis, S

John M. Lewis, S. Lakshmivarahan, and Sudarshan Dhall.Dynamic Data Assimilation: A Least Squares Approach. Cambridge University Press, Cambridge, 2006

2006
[27]

Simulation of nonhomogeneous poisson processes by thinning.Naval research logistics quarterly, 26(3):403–413, 1979

PA W Lewis and Gerald S Shedler. Simulation of nonhomogeneous poisson processes by thinning.Naval research logistics quarterly, 26(3):403–413, 1979

1979
[28]

David J. B. Lloyd, Naratip Santitissadeekorn, and Martin B. Short. Exploring data assimilation and forecasting issues for an urban crime model. European Journal of Applied Mathematics, 27(Special Issue 03):451–478, 2016

2016
[29]

A realistic cellular automata model of bus route system based on open boundary

Yong-Ji Luo, Bin Jia, Xin-Gang Li, Can Wang, and Zi-You Gao. A realistic cellular automata model of bus route system based on open boundary. Transportation Research Part C: Emerging Technologies, 25:202–213, 2012

2012
[30]

Optimising an Agent-Based Model to Explore the Behaviour of Simulated Burglars

Nick Malleson, Linda See, Andrew Evans, and Alison Heppenstall. Optimising an Agent-Based Model to Explore the Behaviour of Simulated Burglars. In Vahid Dabbaghian and Vijay Kumar Mago, editors, Theories and Simulations of Complex Social Systems, number 52 in Intelligent Syste...

2014
[31]

Understanding the kalman ﬁlter

Richard J Meinhold and Nozer D Singpurwalla. Understanding the kalman ﬁlter. The American Statistician, 37(2):123–127, 1983

1983
[32]

Kinetic clustering and jamming transitions in a car-following model for bus route

Takashi Nagatani. Kinetic clustering and jamming transitions in a car-following model for bus route. Physica A: Statistical Mechanics and its Applications, 287(1):302–312, 2000

2000
[33]

Bunching transition in a time-headway model of a bus route

Takashi Nagatani. Bunching transition in a time-headway model of a bus route. Physical Review E, 63(3):036115, 2001

2001
[34]

Calibration of second order trafﬁc models using continuous cross entropy method

D Ngoduy and MJ Maher. Calibration of second order trafﬁc models using continuous cross entropy method. Transportation Research Part C: Emerging Technologies, 24:102–121, 2012

2012
[35]

Jamming transition in a homogeneous one-dimensional system: The bus route model

OJ O’loan, MR Evans, and ME Cates. Jamming transition in a homogeneous one-dimensional system: The bus route model. Physical Review E, 58(2):1404, 1998

1998
[36]

Montemayor

Juan Jose Pantrigo, Angel Sanchez, Kostas Gianikellis, and Antonio S. Montemayor. Combining Particle Filter and Population-based Metaheuristics for Visual Articulated Motion Tracking. ELCVIA Electronic Letters on Computer Vision and Image Analysis, 5(3):68–83, November 2005

2005
[37]

Pennisi, R

M. Pennisi, R. Catanuto, F. Pappalardo, and S. Motta. Optimal vaccination schedules using simulated annealing. Bioinformatics, 24(15):1740–1742, August 2008

2008
[38]

The cross-entropy method for combinatorial and continuous optimization

Reuven Rubinstein. The cross-entropy method for combinatorial and continuous optimization. Methodology and computing in applied probability, 1(2):127–190, 1999

1999
[39]

Design and Implementation of An Agent-Based Simulation for Emergency Response and Crisis Management

Timothy Schoenharl and Greg Madey. Design and Implementation of An Agent-Based Simulation for Emergency Response and Crisis Management. Journal of Algorithms & Computational Technology, 5(4):601–622, 2011

2011
[40]

The Use of Adjoint Equations in Numerical Modelling of the Atmospheric Circulation

Oliver Talagrand. The Use of Adjoint Equations in Numerical Modelling of the Atmospheric Circulation. In An- dreas Griewank and George F. Corliss, editors, Automatic Differentiation of Algorithms: Theory, Implementation, and Application, pages 169–180. SIAM, Philadelphia, PA, 1991

1991
[41]

A bus-following model with an on-line bus station

Tieqiao Tang, Yanfeng Shi, Yunpeng Wang, and Guizhen Yu. A bus-following model with an on-line bus station. Nonlinear Dynamics, 70(1):209–215, 2012

2012
[42]

Mesoscopic simulation for transit operations

Tomer Toledo, Oded Cats, Wilco Burghout, and Haris N Koutsopoulos. Mesoscopic simulation for transit operations. Transportation Research Part C: Emerging Technologies, 18(6):896–908, 2010

2010
[43]

Transit capacity and quality of service manual

TRB. Transit capacity and quality of service manual. Transit Cooperative Highway Research Program (TCRP) Report 165, 2013

2013
[44]

Vadakkepat and L

P. Vadakkepat and L. Jing. Improved Particle Filter in Sensor Fusion for Tracking Randomly Moving Object. IEEE Transactions on Instrumentation and Measurement, 55(5):1823–1832, October 2006

2006
[45]

Data assimilation and its applications

Bin Wang, Xiaolei Zou, and Jiang Zhu. Data assimilation and its applications. Proceedings of the National Academy of Sciences, 97(21):11143–11144, 2000

2000
[46]

Data assimilation in agent based simulation of smart environments using particle ﬁlters

Minghao Wang and Xiaolin Hu. Data assimilation in agent based simulation of smart environments using particle ﬁlters. Simulation Modelling Practice and Theory, 56:36–54, 2015

2015
[47]

Ward, Andrew J

Jonathan A. Ward, Andrew J. Evans, and Nicolas S. Malleson. Dynamic calibration of agent-based models using data assimilation. Royal Society Open Science, 3(4), 2016. 18

2016

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [1]

Matsim-t: Architecture and simulation times

Michael Balmer, Marcel Rieser, Konrad Meister, David Charypar, Nicolas Lefebvre, and Kai Nagel. Matsim-t: Architecture and simulation times. In Multi-agent systems for trafﬁc and transportation engineering, pages 57–78. IGI Global, 2009

work page 2009

[2] [2]

Modeling transit trip time using archived bus dispatch system data

Robert L Bertini and Ahmed M El-Geneidy. Modeling transit trip time using archived bus dispatch system data. Journal of transportation engineering, 130(1):56–67, 2004

work page 2004

[3] [3]

Bus arrival time prediction using support vector machines.Journal of Intelligent Transportation Systems, 10(4):151–158, 2006

Yu Bin, Yang Zhongzhen, and Yao Baozhen. Bus arrival time prediction using support vector machines.Journal of Intelligent Transportation Systems, 10(4):151–158, 2006

work page 2006

[4] [4]

Agent based modeling: Methods and techniques for simulating human systems

E Bonabeau. Agent based modeling: Methods and techniques for simulating human systems. Proceedings of the National Academy of Sciences, 99(90003):7280–7287, 2002

work page 2002

[5] [5]

Improved particle ﬁlter for nonlinear problems

James Carpenter, Peter Clifford, and Paul Fearnhead. Improved particle ﬁlter for nonlinear problems. IEE Proceedings-Radar, Sonar and Navigation, 146(1):2–7, 1999

work page 1999

[6] [6]

Data assimilation in the geosciences: An overview of methods, issues, and perspectives

Alberto Carrassi, Marc Bocquet, Laurent Bertino, and Geir Evensen. Data assimilation in the geosciences: An overview of methods, issues, and perspectives. Wiley Interdisciplinary Reviews: Climate Change, 9(5):e535, September 2018

work page 2018

[7] [7]

Mesoscopic modeling of bus public transportation

Oded Cats, Wilco Burghout, Tomer Toledo, and Haris Koutsopoulos. Mesoscopic modeling of bus public transportation. Transportation Research Record: Journal of the Transportation Research Board, (2188):9–18, 2010. 16 A PREPRINT - AUGUST 23, 2019

work page 2010

[8] [8]

Dynamic bus arrival time prediction with artiﬁcial neural networks

Steven I-Jy Chien, Yuqing Ding, and Chienhung Wei. Dynamic bus arrival time prediction with artiﬁcial neural networks. Journal of Transportation Engineering, 128(5):429–438, 2002

work page 2002