Find in Library

Highways serve as vital connectors between cities, yet they often suffer from traffic congestion as the population continues to grow. Various intelligent frameworks or models for traffic status prediction have been employed in the Intelligent Transport System (ITS) to provide services for convenient and safe traveling, effective traffic management, and smart signal control. Most of these frameworks typically involve learning processes that utilize learning algorithms and requires training data. For highway traffic, the Greenshields model offers a practical relationship among vehicle speeds, traffic flows, and traffic density, which can serve as fundamental knowledge for developing intelligent traffic management systems. This paper proposes a fuzzy logic system based on the Greenshields model as the knowledge base for quickly predicting highway traffic congestion without extensive preparing data. Our system operates in two modes: jam and non-jam modes. In each model, the two inputs of vehicle speed and traffic flow are processed respectively with specified membership functions for effective fuzzification. The set of rules and conditions guided by the Greenshields theory is governed by the inference mechanism, which makes decisions according to the input field. Subsequently, the defuzzification process converts the fuzzy sets obtained by the inference engine into a congestion level as the output. To validate the accuracy of our system, a polynomial regression model utilizing realistic data from roadside equipment on the Sun Yat-Sen Highway in Taiwan is established for comparison. Comparing the observed data points from the polynomial regression model with the outputs obtained from our system using the same inputs, both predicting outputs are found to be consistent, affirming the practical feasibility of the proposed system. Moreover, our proposed scheme is adaptable to suit diverse road conditions without extensive training data and possesses a short memory to perform tasks. Integrating several systems by cascading them across different segments of the highway enables rapid congestion prediction for long-distance traffic. These advantages make the proposed system much more convenient for performing congestion prediction for traffic management and control in ITS.