An Efficient Technique to Control Road Traffic Using Fuzzy Neural Network System

Conventional road traffic controlling systems are dependent on human operators for most of the decisions. Such human operators have experienced a wide variety of incidents and traffic congestions. But even the most experienced operators fail to control traffic efficiently during non-recurrent situations. Non-recurrent situation is a situation which has not been seen earlier by a human operator prior to its occurrence. Controlling traffic flow in such a situation is a complex task as it demands quick reaction and expert knowledge. This paper proposes a novel and efficient approach to control traffic flow in non-recurrent traffic situations. The proposed approach uses multiple techniques, most of which are borrowed from Soft Computing such as, NN (Neural Network), FL (Fuzzy Logic) and GA (Genetic Algorithm). This approach involves clustering imprecise data, in the form of Gaussian mixtures, into fuzzy sets using Expectation Maximization algorithm. The approach also includes minimizing the initial population of chromosomes in Genetic algorithm using a novel algorithm. The proposed algorithm used in identification of valid rules for fuzzy system reduces space and time complexity of the process. The proposed approach has been validated using METANET.