Dynamic Estimators of Origin-Destination Matrices Using Traffic Counts

Cascetta, Ennio; Inaudi, Domenico; Marquis, G. A.

doi:10.1287/trsc.27.4.363

Cited by 285 publications

(146 citation statements)

References 4 publications

Supporting

Mentioning

145

Contrasting

Unclassified

Order By: Relevance

“…Within-day models are mainly approached from a reconstruction perspective via optimization models. For further reading on within-day dynamic models, we refer to the works of Willumsen (1984), Cremer and Keller (1987) , Cascetta (1993) and Ashok and Ben-Akiva (2002). In this review, we focus on day-to-day models, which have been approached from a statistical estimation perspective.…”

Section: Dynamic Modelsmentioning

confidence: 99%

“…Such data can be compiled and several statistics may be computed, such as the mean, standard deviation and con idence intervals. However, these direct methods require large samples to achieve a target statistical signi icance, which may be technically or economically infeasible (Cascetta, 2009). Another way of estimating OD matrices is by using trip generation and distribution models.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Statistical models for the estimation of the origin-destination matrix from traffic counts

2017

View full text Add to dashboard Cite

In transporta on planning, one of the first steps is to es mate the travel demand. The final product of the es ma on process is an origin-des na on (OD) matrix, whose entries correspond to the number of trips between pairs of origin-des na on zones in a study region. In this paper, we review the main sta s cal models proposed in the literature for the es ma on of the OD matrix based on traffic counts. Unlike reconstruc on models, sta s cal models do not aim at es ma ng the exact OD matrix corresponding to observed traffic volumes, but they rather aim at es ma ng the parameters of a sta s cal model of the popula on of OD matrices. Ini ally we define the es ma on problem, emphasizing its underspecified nature, which has lead to the development of several models based on different approaches. We describe sta c models whose parameters are es mated by means of maximum likelihood, the method of moments, and Bayesian inference. We also describe some recent dynamic models. Following that, we discuss research ques ons related to the underspecifica on problem, model assump ons and the es ma on of the route choice matrix, and indicate promising research direc ons. RESUMONo planejamento dos transportes, um dos primeiros passos é es mar a demanda por viagens. O produto final do processo de es mação é uma matriz origem-des no (OD), cujas entradas correspondem ao número de viagens entre pares de zonas de origemdes no em uma região de estudo. Neste ar go, revisamos os principais modelos estaAs cos propostos na literatura para a es mação da matriz OD com base em contagens de tráfego. Ao contrário dos modelos de reconstrução, os modelos estaAs cos não visam a es mar a matriz OD exata correspondente aos volumes de tráfego observados, mas sim a es mar os parâmetros de um modelo estaAs co da população de matrizes OD. Inicialmente, define-se o problema da es mação, enfa zando sua natureza subespecificada, o que levou ao desenvolvimento de vários modelos baseados em diferentes abordagens. Descrevem-se modelos está cos cujos parâmetros são es mados por meio da máxima verossimilhança, do método dos momentos e da inferência bayesiana. Descrevem-se também alguns modelos dinâmicos recentes. Em seguida, discutem-se questões de pesquisa relacionadas ao problema da subespecificação, às premissas adotadas nos modelos e à es ma va da matriz de escolha de rotas, e indicamse direções de pesquisa promissoras.

show abstract

Section: Dynamic Modelsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Statistical models for the estimation of the origin-destination matrix from traffic counts

2017

View full text Add to dashboard Cite

show abstract

“…This former implements calibration with respect to macroscopic traffic data, for example, flow-density data for the specific region, or estimates origin-destination matrices [3]. The latter treats vehicles as an individual entities and uses microscopic trajectory data [5,9].…”

Section: Model Calibrationmentioning

confidence: 99%

Traffic and Granular Flow '13

Flow¹,

Chraibi²,

Boltes³

et al. 2015

View full text Add to dashboard Cite

The Intelligent Driver Model (IDM) is studied and several drawbacks with respect to driving simulators are defined. We present two modifications of the IDM. The first one gives any predefined distance to the leading vehicle in a steady state. The second modification is a combination of the first one and the optimal velocity model. It takes into account driver's reaction time explicitly and is described by delay differential equation. This model always results in realistic vehicles accelerations what allows simulating real traffic collisions.Necessary and sufficient conditions are obtained, that guarantee a non-oscillating solution near the equilibrium for the vehicle platoon. We suggest the calibrating framework based on a numerical solution of the constrained optimization problem. Nonlinear constraints are generated by the numerical integration scheme. The suggested procedure incorporates the local stability conditions obtained and takes into account vehicle dynamics, drivers' behavior and weather conditions.

show abstract

“…Recently, Cascetta et al (1993) has proposed a generalized least squares dynamic O-D estimator. This model was tested on a linear section of freeway in Italy, having 19 zones, 19 nodes, and 54 links.…”

Section: Static Vs Dynamic Approachesmentioning

confidence: 99%

Estimating dynamic origin-destination demands from link and probe counts

1997

Transportation Research Part A: Policy and Practice

View full text Add to dashboard Cite

Many traffic engineering problems consist of evaluating a number of alternate schemes in terms of some measures of performance, such as total travel time, volume to capacity ratio, average speed, total fuel consumption, total emissions, etc. These evaluations generally require prior knowledge of the temporal trip making behaviour of drivers by origin and destination. This thesis presents the development, application, and evaluation of two models capable of inferring these temporal origin -destination (O-D) traffic demands on the basis of observed link traffic flows and assumptions regarding drivers' route choices. In particular, this thesis presents the development and evaluation of a Least Squared Error (LSE) and a Least Relative Error (LRE) model, each of which is capable of estimating either static demands, a time series of static demands, or dynamic demands. Furthermore, the potential of using probe data, from route guidance system (RGS) equipped vehicles, to enhance these estimated dynamic O-D demands is examined.Both of the LSE and LRE models' mathematical formulations are presented. The LSE model formulation parallels that of a least squared regression as the error function is composed of the sum of the squared absolute difference between the observed and estimated link flows. In contrast, the LRE model is formulated on the basis that the link flow error, when measured relative to the observed flow, is to be minimized instead. Iterative solution algorithms, that are modifications of the Jacobi and Gauss-Seidel techniques, are proposed to solve each of the model formulations. It is shown, by way of the application of these iterative algorithms to several example networks, that the estimated O-D demands, which result from these iterative solution techniques, are consistent with the model formulations and with the analytical solutions. Furthermore, it is shown for several examples that, when multiple solutions exist which each exactly replicate the observed link flows, and no prior O-D demand information is specified, both the LSE and LRE models estimate demands that closely approximate the maximum likelihood solution.The proposed iterative solution algorithms have been incorporated into a computer model called QUEENSOD. This model can be practically applied to real networks using current computer memory constraints. This thesis describes the application of the LSE and LRE models to a 35 km section of multilane urban freeway in Toronto, Canada, in which alternate parallel routes exist. Dynamic 15 minute O-D demands were estimated for the eastbound direction for the period from 5 am to 11 am. Despite FTMS detector data being available for only 45% of the network, a correlation coefficient of approximately 98% was obtained for both models. This value reflects the high linear correlation between estimated and observed link traffic flow data for this network. The statistical analysis of the expected quality of O-D demands, which are estimated solely on the basis of RGS probe vehicle data, indicated that ...

show abstract

Dynamic Estimators of Origin-Destination Matrices Using Traffic Counts

Cited by 285 publications

References 4 publications

Statistical models for the estimation of the origin-destination matrix from traffic counts

Statistical models for the estimation of the origin-destination matrix from traffic counts

Traffic and Granular Flow '13

Estimating dynamic origin-destination demands from link and probe counts

Contact Info

Product

Resources

About