This paper presents an agent-based method for modelling route choice behaviour under the influence of real time traffic information. The data for model development was obtained from a field survey of driver behaviour which was conducted on a congested commuting corridor in Brisbane, Australia. Fuzzy artificial neural networks (ANNs) were used to describe driving behavioural rules and analyse the impacts of socioeconomic, context and information variables on individual behaviour and propensity to change route and adjust travel patterns. A number of ANN architectures were examined. The results showed that learning vector quantization (LVQ) neural network models outperformed the other architectures for this application. The models developed in this study were also interfaced to a microscopic traffic simulation tool. A number of methods to calibrate the membership functions, fuzzification, and defuzzification are reported in this study. The fuzzy-neural models are also compared against binary probit and logit models and ANN models. The results showed that the fuzzy-neural models outperformed the other models tested. (Author/publisher) For the covering entry of this conference, please see ITRD abstract no. E214133.
Abstract