a study on artificial neural network application to automatic freeway ramp metering control

• Main Authors: wu, kun-yu, 吳耿毓
• Other Authors: wei chain hong
• Format: Others
• Language: zh-TW
• Published: 1996
• Online Access: http://ndltd.ncl.edu.tw/handle/94028689253318316194

碩士 === 國立成功大學 === 交通管理(科學)學系 === 84 === In most freeway control system, the optimal metering rate is calculated by complex arithmetic methods. However, if we solve the problem of human behavior by biological perception and reaction, it will be more human nature. Artificial intelligence, instead of arithmetic methods, is used to provide optimal control strategy. We treat freeway as a biological system which can learn control strategy and make suitable decisions by neural network. An expert system is also needed to evaluate deriation and adjust direction. This research uses southloowd of Sun Yet-Sen freeway as an example. First, a neural network metering rate estimating model is constructed with time-space feature for handle dynamic freeway traffic flow. Second, a metering rate self-adjustment expert system with time- space feature is developed for modifying inappropriate metering rates. While the system is running, the neural network model calculates metering rates to implement, and the self-adjustment model evaluates performance of metering control and modification strategy. Then, relevant information is sent back to neural network metering rate estimating model for on-line retraining. It is expected that the freeway control system will perform better and better by this type of feedback learning. This research develops a freeway traffic simulation program as a tool for observeing metering control system performance. We compare FREQ control mode with several neural network control modes. From the simulation results, we find the neural network system is able to adjust behavior gradually according to preset target, and improve performance. Implementing self- adjustment model leads the system toward more appropriate control strategy.