Location choice modeling for . . .

Transportation and activity location infrastructure is generally lim-ited, which means that users are in competition for time slots to use the infrastructure. Even though the idea of homo oeconomicus clearly needs to be extended by altruism, bounded rationality, satisficing prin-ciples, incomplete information, and choice heuristics, the assumption that people in general do not act against their own good is close to consensus. As a consequence, it can be deduced that choices expected to bring less subjective benefit than others disappear from people’s minds. This can be seen as a coevolutionary mechanism with respect to people’s choices. Although coevolutionary systems exhibit a variety of behaviors, ranging from convergence to a fixed point to periodic or even chaotic behavior, for transport demand modeling, traditionally the attrac-tive fixed points Nash equilibria, the states in which no individual improvement is possible due to unilateral change [Nash (2)] are rel-evant. Even though the individual dynamics of real travelers are not known, building a coevolutionary system model shows promise for potentially capturing the patterns of people’s activity scheduling and participation behavior at a high level of detail in relation to space and time. This coevolutionary approach is adopted in MATSim. COMPUTABILITY OF LARGE-SCALE SCENARIOS Before efforts are made to bring the MATSim simulation results closer to reality by iterative extension, calibration, and validation of the model, two crucial conceptual problems need to be solved. First, practical computability of large-scale scenarios needs to be achieved, and second, further light needs to be shed on the question of the uniqueness of a computed solution (i.e., an equilibrium that is related to the Nash equilibrium). The first part of this paper addresses the computability problem by combining microsimulation and time geography. Being confronted with large location choice sets and large num-bers of agents to be modeled on one hand and knowing that in real-ity people’s choices are in general subject to constraints on the other, researchers recognize that the idea to confine the agent’s location choices through time–space constraints is natural and that it has already been applied both in simultaneous and in sequential opera-tional microsimulations such as PCATS [e.g., Kitamura et al. (3)