Abstract
Traffic systems are ill-defined structures and so complicated that theoretical general-purpose design methods are difficult to use. In this paper, a flexible design method in which control strategies are represented as a set of rules or a look-up table is proposed. This method is based on a genetic algorithm-like framework and automatically generates feasible control rules by efficient trials of simulations such as those done using the Monte Carlo method. If the evaluation method for each simulation result is clear, the method is easily applied to various types of traffic control systems. A prototype system for offset control table of traffic signal is evaluated by means of simulation experiments.