Modelling South African traffic for large networks - an extension of the gravity model for traffic demand modelling

Swarts, Stefanus Jacobus (2017-03)

Thesis (MCom)--Stellenbosch University, 2017.

Thesis

ENGLISH SUMMARY : Traffic congestion is a growing burden on society. Due to the complexity of modelling transportation, many approaches do not scale efficiently. Assumptions are made to estimate the locality and quantity of traffic that passes through, leaves or enters a study area. Infrastructure investment based on inaccurate transportation modelling could potentially increase congestion. Furthermore, the effects of changes in infrastructure outside a study area are also unknown. It is therefore necessary to model transportation networks at a larger scale than required before. This thesis presents a large scale traffic modelling framework, potentially capable of modelling the impact of public transport and infrastructure investment for the South African context. The framework builds on the basic procedure of the Four-Step Modelling methodology. To address the shortcomings of static trip distribution models, a proposed, temporally adjusted, doubly constrained gravity model is formulated. Kernel density functions are fitted from survey data to dene unique impedance of travel functions for each travel analysis zone. Route selection is determined by a heuristic approach to dynamic traffic assignment which is implemented in a mesoscopic traffic simulator. The proposed gravity model is iteratively calibrated using the output of the traffic simulator in a positive feedback loop to produce a trip distribution that approaches an equilibrium assignment. The gravity model calibration and validation shows that the proposed gravity model is more accurate than the single impedance of travel gravity model. The framework produces OD trip matrices, inter- and intra-zonal routes, quarter-hourly traffic flows and a measure of congestion, all which can be visualised in a GIS environment.

AFRIKAANSE OPSOMMING : Verkeersopeenhopings is ‘n immertoenemende probleem in Suid-Afrika. Verskeie bestaande tegnieke wat gebruik word om verkeer te modelleer, is ondoeltreffend wanneer dit op ‘n groot skaal geimplementeer word. Aannames word gemaak oor die kwantiteit en lokaliteit van verkeer wat vanuit, na en deur die studie-area vloei. Infrastruktuurontwikkeling gebaseer op onakkurate verkeersmodellering kan die intensiteit van verkeersopeenhopings vererger. Die impak wat so ‘n verandering van infrastruktuur sal he buite die studie-area kan ook nie geevalueer word nie. Dit is daarom nodig om vervoerinfrastruktuur op ‘n groter skaal as vantevore te modelleer. Hierdie tesis bied ‘n grootskaalse verkeersmodelleringsraamwerk aan wat potensieel die impak van publieke vervoer en infrastruktuurontwikkelings kan modelleer vir die Suid-Afrikaanse konteks. Die raamwerk bou voort op die beginsels van die Vier-Stap Modelleringsmetodologie. ‘n Dubbele beperkte gravitasie model wat aangepas is, word voorgestel en geformuleer om die tekortkominge van statiese ritverspreiding modelle aan te spreek. Kerndigtheidsfunksies word gepas aan opname-data om unieke reisimpedansiefunksies te denieer vir elke verkeersanalisesone. Ritroetes word bepaal deur middel van dinamiese verkeerstoedeling wat geimplementeer is deur ‘n mesoskopiese verkeerssimulasie. Die gravitasiemodel word iteratiewelik gekalibreer deur die uitvoer van die verkeersimulasie te gebruik in ‘n positiewe terugvoer lus om konvergerende ritverdelings te produseer. Die kalibrering en validering van die voorgestelde gravitasiemodel toon aan dat dit meer akkuraat is as die enkele ritimpedansie gravitasiemodel. Die raamwerk produseer oorsprong-bestemming matrikse, inter- en intrasone roetes, kwartuurlikse verkeersvloeie en ‘n maatstaf van verkeersopeenhopings wat in ‘n GIS omgewing visualiseer kan word.

Please refer to this item in SUNScholar by using the following persistent URL: http://hdl.handle.net/10019.1/100856
This item appears in the following collections: