Agent-Based Modeling of Mode Choice with Dynamic Attitudes and Social Influence.

摘要

Models of travel behavior are important tools for understanding how people make choices in the context of everyday activity and travel, and they are used to help predict the consequences of infrastructure and policy decisions. Traditionally, these models were built on a foundation of economic theory, and they focused on representing calculated trade-offs between factors such as time, money, and comfort. However, research on the psychology of human decision-making has made it clear that our decisions may be driven as much by social influence, personal bias, and habit as by reasoned trade-offs. In addition, we often make decisions in concert with others, whether for the purpose of participating in activities together or sharing limited resources. Considering this, it is clear that social and psychological factors could be crucial drivers of travel patterns. Recent studies have shown that these factors do have important effects, and this topic is the focus of increasing interest in the transportation community.;This research is designed to contribute to our understanding of travel behavior by investigating the effects of modeling dynamic attitudes and social interactions in an agent-based travel behavior simulation. There are three valuable insights that this work aims to generate. The first is a fuller understanding of how to simulate these cognitive processes in a way that best reflects our current knowledge of how they work and what their effects are. The second is an understanding of how the specification of model parameters related to these behavioral processes affects mode attitudes and the resulting mode shares. The third is an understanding of how characteristics of the transportation network or the population can affect these two outcomes, and how the inclusion of these additional cognitive processes changes these effects.;An initial study was carried out using a small-scale transportation network and population, implemented in the Netlogo agent-based modeling platform. The goal was to construct a model in which mode choice depends in part on attitudes about travel modes. These attitudes evolve over the course of the simulation in response to experiences on the transportation network and influence from the social network, which also changes dynamically. The advantage of this platform was that it provided a practical framework to explore modeling assumptions without the necessity of collecting empirical data. However, this initial study made clear that there would be additional value in testing such a model in larger-scale, more realistic context.;Thus, the model was implemented with additional improvements in MATSim, a transportation simulation software package that provided the capability to apply the model to the city of Chicago. A series of simulations were conducted to address the second and third research goals described above. The results indicate that the inclusion of dynamic attitudes and social networks in a mode choice model leads to some surprising findings that merit further exploration. For example, in these simulations, social influence plays a much greater role than travel experience memories in the process of dynamically updating mode attitudes. Interestingly, there is potentially a tipping point phenomena occurring in the social network, in which an abrupt change in mode share takes place midway through the simulation. Initial mode attitudes have strong effects on the mode share outcomes, while mode performance levels have surprisingly little impact. Further simulation and analysis would be necessary to verify the hypothesized underlying causes of these observations, but it is clear that the inclusion of dynamic attitudes and social networks in a model of mode choice provides a framework for studying cognitive processes that may have a significant affect on travel behavior.