Ed Manley

Ed Manley

Modelling Driver Behaviour to Predict Urban Road Traffic Dynamics


Urban road traffic patterns are a product of the behaviours, choices, and movements of many individual drivers. Realistically capturing and modelling the behaviour of these individuals is, as such, central to the accurate simulation of traffic dynamics. However, many conventional simulation approaches tend to represent driver behaviour in only a simplistic fashion. Individuals are assumed to consistently select the least travel time path with consideration of a multitude of possibilities. Research into traveller behaviour would suggest that the human route selection process, in choosing a route from an origin to a destination, is not so straightforward. In this thesis, advances are made to improve the representation of individual behaviour within traffic simulation, and in doing so improve the predictive power of traffic simulation more widely. Central to this work is the development of a range of new models that incorporate human spatial cognition and bounded rationality into a driver behavioural framework. In the first instance, a large dataset of observed route selections is processed and analysed, extracting a range of novel observations with respect to route choice behaviour. Following this stage, new approaches towards the modelling of route choice are developed, building both on these data analyses and integrating wider observations drawn from the literature. The influence of heterogeneity in individual perception of space is also considered, with a model describing variations in spatial knowledge introduced. These elements are integrated within a framework reflecting realistic heterogeneity in motorist behaviour, and then implemented within a largescale agent-based model of road transportation around central London, United Kingdom. Traffic speed and flow datasets are extracted from the agent-based model and compared against observed traffic data. It is shown how this new representation of driver behaviour enables a marked improvement in the prediction of real-world traffic dynamics than offered by conventional approaches.

Contact: ed.manley@ucl.ac.uk

Page last modified on 06 jun 16 11:39