View article

While transportation researchers have long recognized that travellers’ decisions are interdependent, actor interdependencies are rarely incorporated in our field. Good reasons for this are that the common statistical tools requiring the assumption of independent decision makers have worked well for many applications. Meanwhile, real maps of social, work, and business interactions have not been made at the scope required for transportation studies. As discretionary travel constitutes an increasing proportion of traffic, and policy makers seek ways to reduce low-occupancy vehicle travel, understanding these interdependencies will be more important for modeling and policy impacts. Agent simulation is available in lieu of comprehensive empirical studies to construct the interdependencies between travellers and activity-travel choices. This paper introduces a general spatial social interaction model, based on the Multi-Agent Transportation Simulation Toolbox (MatSim-T, Rieser et al 2007), to serve as a laboratory for scenario-and hypothesis testing. The model has two features that suit it to general study of social networks and activity spaces: modular structure and the speed to simulate large numbers of agents in the travel simulation. The basic tools of transportation science: utility maximization, activity plans, and generalized travel cost, are used to construct social networks for a geographically distributed population of agents. This network is then used in a further step to modify travel demand (location, activity, and/or timing). This paper summarizes the model, presents some preliminary results, illustrating some of the issues of investigating complex …