FACULTY OF SCIENCE UNIVERSITY OF COPENHAGEN
PHD DEFENCE BERNHARD SNIZEK Department of Geosciences and Natural Resource Management Faculty of Science University of Copenhagen
MAPPING CYCLISTS’ EXPERIENCES AND AGENTBASED MODELLING OF THEIR WAYFINDING BEHAVIOUR WEDNESDAY, OCTOBER 28TH, 2015 13:00 AUD. A1-04.01, GROUND FLOOR, GRØNNEGÅRDSVEJ 7, FREDERIKSBERG C 55.6798718,12.5381017
Assessment Committee: Prof. Dr. Andy Hudson-Smith, Head of Department, Centre for Advanced Spatial Analysis, Faculty of the Built Environment, University College London, UK Dr. Seraphim Alvanides, Reader in GI Science, Department: Architecture and Built Environment, Northumbria University Newcastle, UK
Academic main supervisor: Hans Skov-Petersen, PhD, Department of Geosciences and Natural Resource Management, University of Copenhagen, Denmark Academic co-supervisor: Thomas Alexander Sick-Nielsen, PhD, Department of Transport, Technical University of Denmark, Kgs. Lyngby, Denmark Academic co-supervisor: Prof. dr. Nico Van de Weghe, CartoGIS, Ghent University, Ghent, Belgium Academic co-supervisor: dr. Tijs Neutens, PhD, CartoGIS, Ghent University, Ghent, Belgium
ABSTRACT This dissertation is about modelling cycling transport behaviour. It is partly about urban experiences seen by the cyclist and about modelling, more specifically the agent-based modelling of cyclists' wayfinding behaviour. The dissertation consists of three papers. The first deals with the development and application of a method for collecting experiential data via an internet-based questionnaire and statistically relating them to physical features of the city as well as the characteristics of their routes. The other two papers explain methods for building, calibrating and validating an agent-based model of cycling transport behaviour using geodata, data from the Danish travel survey as well as behavioural data extracted from trajectories recorded utilising GPS units.
Mapping Bicyclists’ Experiences in Copenhagen This paper presents an approach to the collection, mapping and analysing of cyclists’ experiences. By relating spatial experiences to urban indicators such as land-use, street characteristics, cycle infrastructure, centrality and other aspects of the urban environment, their influence on cyclists’ experiences were analysed. 398 cyclists responded and plotted their most recent cycle route and a total of 890 points for locations along the route where they had had positive and negative cycling experiences. The survey was implemented as an online questionnaire built on top of Google Maps, and allowed up to three positive and three negative experiences to be plotted on a map and classified. By relating the characteristics of the experience points and the routes to the traversed urban area in general, the significance of the preconditions for obtaining positive or negative experiences could be evaluated. Urban spaces can, thereby, be mapped according to whether they potentially promote positive or negative experiences. Additionally, the method may be applied to measure the effect of proposed changes to the urban design in terms of cyclists’ experiences. Statistical analysis of the location attributes, traffic environments and conflicts, cycle facilities, urban density, centrality, and environmental amenities indicate that positive experiences or the absence of negative experiences are clearly related to the presence of en-route cycling facilities and attractive natural environments within a short distance of large water bodies or green edges along the route.
Modelling cyclists' GPS trajectories with spatial agents and model calibration data creation This paper has two objectives, which are to develop and present a method for simulating single GPS-based trajectories by applying an agent-based model and to acquire parameter values for CopenhagenABM, an agent-based model of cyclists’ behaviour. The core of the model, the Behavioural Edge Choice Matrix (BECM), which is responsible for the agent's network edge choice in any node of the road network, was designed and expressed to comprise both local parameters and a global one. The local parameters, which represent subjective weighted preferences of the immediate vicinity of the decision point, i.e. a node of the network such as the greenness of outgoing edges and the availability of cycle infrastructure, the traffic environment are set by values retrieved from the analysis of GPS trackings of cyclists on their everyday trips. The global parameter, i.e. the directional deviation was established in order to reflect the agent's knowledge of the direction towards the destination of the journey. In order to analyse the performance of the model, the model was run with a series of different values for the global parameter with and without taking the local parameter into consideration. The resulting routes' overlap with routes taken from the real world was calculated and used as a qualifier for the capacity of the model to explain the real world phenomenon. The analyses and the conclusions from these model results are discussed at the end of the paper.
CopenhagenABM: An Agent-based Model of Bicyclists' Trajectories and Flow CopenhagenABM, an agent-based model of cyclists’ wayfinding behaviour, was designed and implemented and its results were compared to real world counts. The central component of wayfinding, the Behavioural Edge Choice Matrix (BECM), was parameterised building on choice estimates generated from the analysis of GPS tracks recorded from Copenhagen cyclists as a local selection component and the direction towards the trip destination as an overall cognitive component. These rules were implemented into agents and the spatial behaviour was upscaled to the city level. The model was implemented in rePAST, a state-of-the-art agent-based modelling toolkit. The behavioural parameter estimates were generated from GPS tracks. A road network was taken from OpenStreetMap and enriched with information about the traffic environment, public register-based source and destination coordinates and origin/destination data from the Danish traffic model. Counting data provided by the Municipality of Copenhagen were used to compare the results of the model with real world counts.