Generating activity-specific multimodal trips from aggregated data sources

Abstract

Mode choice is the result of selecting the most convenient mean(s) of transport among all options available to reach locations where daily activities are performed. Trips made by individuals are enchained, making mode choice an interconnected process, especially considering the spatial connections and temporal constraints to reach the different activity locations. This paper proposes a methodology to estimate ativity and mode-specific dynamic OD trips, based on the economic and behavioural principle of utility maximization. The proposed modelling approach aims at preserving the link between individual trips, while limiting combinatorial explosion of choices when considering individual trip chains. This approach is applied at a zonal scale and relies on observed aggregated OD trips and travel times by mode as main input data. The model parameters are estimated through Markov Chain Monte Carlo sampling and Bayesian updating techniques, hence, instead of calculating only expected values, each parameter is described by a probability distribution, which can represent and incorporate in the model the heterogeneity of population activity-travel behaviour. This model is tested on multi-day travel diary data collected in the city of Ghent, Belgium, containing around 15000 trips classified by four main activity types and four alternative modes. We show that the proposed model can properly estimate and capture the dynamics of modal split and relate these choices to activity participation.