AN ANALYSIS OF ROUTE CHOICE BEHAVIOR BY A REGRET MATCHING MODEL -ATOMIC MODEL WITH IMPERFECT TRAVEL INFORMATION-
Toshihiko Miyagi, Masahiko ISHIGURO · INFRASTRUCTURE PLANNING REVIEW · 2010
In classical traffic assignment procedures, drivers have been treated as a collection of an infinite number of particles that have perfect information and perfect foresight. On the contrary, in this paper, each driver was modeled as an individual decision-maker with imperfect information on his travel environment. We have adopted extensive games as the basic framework to describe the process of route choice behavior of naive drivers, and analyzed the game by using "Counterfactual Regret". Each driver's decision-tree was constructed by K-shortest path algorithm. For link costs, monotone increasing, but non-differentiable functions were assumed. We have shown through numerical experiments that the algorithm developed here achieved Nash equilibrium.