In recent years, with the growing traffic congestion and more attentions on time value, travel time reliability has become one of the key issues travelers concerned about. Researchers all over the world give ideas of the definition and calculation formulas of travel time reliability. While most of previous studies assumed that travel time subject to specific forms of distribution, it does not fit the actual operating conditions and urban roads.There exist two main issues relating to the evaluation of path travel time. On one hand, the travel time distribution varies violently across the links, on the other hand, though the average travel time of path can be calculated as the sum of average time on each link, the variance of path travel time cannot be simply overlaid from the travel time distributions on links.The improvement on estimation of path travel time distribution can lead to accurate evaluation of the travel time reliability, the study establishes the evaluation model and algorithm to estimate link and path travel time distribution based on previous studies. One new approach is proposed and verified for evaluating the reliability of urban road links an paths.The main results of this study are listed as follows:Firstly, this paper proposes a new method of travel time data classification by considering the state difference among travel time distribution and judging the boundary line between each group of travelling states. The double lognormal distribution model is employed to fit travel time distribution of each state. The overlaid curve by considering the proportions and travel time distribution for each group of state can be successfully represent the real distribution of road link travel time.Secondly, the correlation of successive links are examined by calculating the transition probability of each group with different states, then path travel time distributions are estimated by applying a Markov chain procedure.Finally, the path travel time reliability can be calculated directed from the fitted distribution curve. Comparison between the proposed method and previous method are conducted. Both network simulation and actual road network are used and verified. |