Ehsan Ramezani Khansari - PhD Candidate, Amirkabir University of Technology
Masoud Tabibi - Assistant Professor, Amirkabir University of Technology
Fereidoon Moghadas Nejad - Associate Professor, Amirkabir University of Technology
Atiye-Sadat Hashemiyan - MSc, Azad University of Zanjan

Many studies have tried to estimate time headway distribution in a traffic flow. But investigations on differences between congested and uncongested time headway distributions is rare, more over it should be estimated for each different country. This paper tries to propose time headway distribution for native drivers in different speed ranges. The results showed although Weibull 3-parameter distribution is suitable for congested and uncongested flows, parameters are different. Contrary to the idea increasing speed will increase time headway, smaller time headway was monitored in higher speeds. Genetic algorithm (GA) was used rather than conventional classic methods such as maximum likelihood estimation (MLE) for parameter estimation. Results showed better fit to the natural data. Proposed GA improved distribution error about 50 and 12% by estimating better parameters

Car following, Time headway. Genetic algorithm, Congested, Uncongested