| This thesis outlines an implementation of Genetic Algorithms to traffic simulation optimization and development of a program called GENOSIM, a Genetic-based Optimizer for Traffic Microscopic simulation Models. GENOSIM is developed as a pilot software that employs the state of the art in combinatorial parametric optimization to automate the tedious task of calibrating traffic simulation models. The employed global search technique, Genetic Algorithms, is integrated with a dynamic traffic microscopic simulation modeler, Paramics, and experimented with Toronto network, Canada. The output of GENOSIM is the near-optimal values of its car-following, lane changing and dynamic routing parameters. Obtained results are promising.;Paramics consists of high performance cross-linked traffic models having multiple user-adjustable parameters. Genetic Algorithms in GENOSIM will manipulate the values of control parameters and search an optimal set of values as starting configuration for these parameters by matching model outcome with observed data. The most of C++ codes shown here have been simplified for clarity. |
