抄録
Traffic jams are getting more attention due to their environmental impacts in addition to resultant economic losses recently. The most popular method of traffic signal control is to select an appropriate signal pattern for the current traffic flow out of those pre-calculated. This selection process depends on experiences of traffic engineers and. social experiments currently. Furthermore, a signal control to improve a traffic flow at a specific cross section does not necessarily improve the flows of the whole area; i.e. it may cause a traffic jam in another area. To resolve these issues, we study a traffic signal control using optimization methods. We calculate the parameters of every single signal inclusively and optimize a flow as a whole. The goal of this work is to find optimal sets of control parameters of traffic signals to improve the traffic flows globally. We developed an optimization system using the genetic algorithm in conjunction with our multi-agent traffic simulator. We applied the system on traffic flows in an area in Okayama prefecture. The results shows that our approach is reasonable, but needs improvements in accuracy.
