| Summary: | 碩士 === 國立臺灣科技大學 === 電機工程系 === 92 === Generally, traffic signal can be controlled by two kinds of approaches, timer signal traffic control and dynamic calculation traffic control. Timer signal traffic control is a timer-schedule model, which is to pre-define signal schedules for each day. In this kind of approaches, there are usually several pre-set schedules which are used in different time slots in a day. Dynamic calculation traffic control is to dynamically tune the traffic control strategy based on some observed information, such as traffic flow or traffic density. Usually this kind of approaches needs to collect information every 10~15 minutes so that it can predicts traffic in the future. In timer signal traffic control, many scholars such as Webster, Akcelik, etc., have proposed various approaches. In dynamic calculation traffic control, there also exist several mature models as the OPAC mode, the MOVA mode, etc. In this thesis, we model traffic signal control based on Akcelik’s delay model. We use a simple genetic algorithm to find the optimal valid period. We then analyze the relationship between flow and the obtained valid period to discover any possible ways of dynamic calculation traffic control. We also study the effects of different powers of the delay time used in the fitness function value. It is found that it can have a smaller delay time with a better safe velocity when the power of the delay time is 2. We also analyze the relationship between traffic flow and the valid signal control period. It is concluded from our simulation that the dynamic control rules can be modeled as a curve as the difference between the maximal x-coordinate flow and the maximal y-coordinate flow vs. the difference between the maximal x-coordinate valid period and the maximal y-coordinate valid period. It can be verified that the total delay time based on the obtained curve is smaller than that of using the original timer signal traffic control.
|
|---|
