OD Estimation Algorithm Based On Ordered Weighted Averaging Operator

As the rapid economy development, technological progress and fast urbanization, urban traffic congestion is becoming the bottleneck for further development of urban areas. A lot of traffic managerial personnel and researchers are dedicating themselves to researches of optimizing the limited road network conditions and enhancing the efficiency of road usage for maximum utilization of the limited traffic resource. Traffic designers and managerial personnel are trying to meet the travelers'demands and improve the service quality.OD matrix is the important basic data for the transportation planning and design, traffic control and management. In general, there are two ways to obtain OD matrix:one is the OD surveys:this is a direct access to get the OD matrix by questionnaires and other similar methods; the other is OD estimation by analyzing the observed data. This estimation method is increasingly attracting the concern of traffic managerial personnel due to its high efficiency, convenience and preciseness.This paper commits to figuring out a method to obtain much more accurate OD matrix based on the comparative analysis of the traditional algorithm methods.At first, this paper analyzes the derivation, advantages and disadvantages of different traditional OD matrix estimation methods, compares their application fields and applicable scopes, and analyzes the Ordered Weighted Averaging (OWA) operator weight calculation methods, their features and other relevant theories, which lays the foundation for the further study.Based on the OWA theoretical knowledge and the characteristics of the OD estimation model of bus transit routes, this paper propose a new model, which considers bus passenger travel behavior characteristics, the Markov chain properties of transit route, affinity weight coefficients between the stops and other features to avoid the limitations of a single estimation method. Further more, it also analyzes the weight coefficients and the weighted OD matrix under different orness levels of subjective decision making factors.For the OD estimation model of urban roads, this paper further compares the advantages and disadvantages of the maximum entropy model, generalized least squares model, maximum likelihood model, information minimizing model and bi-level programming model, and establish a combined estimation model with applying the OWA operator theories to the urban road OD matrix estimation study. The paper then verifies the validity and effectiveness of such combined model through the case. The paper also analyzes the result trends of the orness level of different subjective decision factors when the combined model is applied, and identifies the reasonable orness level in such case, which improves the accuracy of the estimated OD matrix to some extent.