Who gets the key first? Car allocation in activity-based modelling

Abstract

Activity-Based Modelling. International Journal of Urban Sciences, 1-15. Official version can be downloaded at http://dx.Abstract: 296 words Number of words: 4929 words Number of tables: 5 tables Abstract Decisions concerning household car ownership and the corresponding usage by the household members have significant implications on vehicle usage, fuel consumption and vehicle emissions. In this context, long-term and short-term choices which are strongly interrelated with one another play an important role. The long-term aspects involve the number of vehicles and their different types owned by a household as well as the assignment of a main driver, acting as the primary user, to each vehicle. The short-term dimension is represented by the vehicle allocation within a household at a daily level.In order to better understand the vehicle allocation process in the household context, the paper at hand investigates the importance of the short-term and long-term aspects in this process and explores several approaches to model them. For this purpose, four different methods for car allocation within a household, which strongly differ in their complexity, are implemented into a microscopic agent-based travel demand model and subsequently evaluated. The respective approaches are the following: (1) random car allocation, (2) car allocation by age, (3) car allocation by main driver assignment, and (4) car allocation by household optimisation. Given a population of a bigger region that is described by a set of attributes, these various models determine which person of a household uses one of the available cars within the household for his/her daily trips.The simulations show that all four implementations of car allocation result in good representations (with deviations of less than 10%) of observed travel behaviour, their results being closer to each other than initially expected. Model (4), which optimises car allocation for the entire household, shows the best results when compared to realworld data, while model (3) allows for the adaptation of changes in car ownership and/or socio-demographic and socio-economic attributes of the population.