Research On The Computational Method Based On Markov Chain For Road Weight Of Traffic Network

Urban traffic system consist of large quantities of vehicles,roads,crossings and traffic engineering facilities, all of them link tightly to make up the traffic network. And the characteristics of traffic flow are self-organization, adaptive and self-driven. Non-linear interaction exists between the vehicles of the road network, thus urban traffic system are complex network system. For this reason, how to study the complicated network characteristic of the traffic system and access to relevant features information, have the very important significance regarding the design of driving cycle of the city.On the basis of study of complex networks theory and theoretical research of Markov chain referring to complex network, in this paper, we construct network travel dynamics model taking road section travel as basic unit, obtain the data of traffic network by floating car technology, do research on regularity in the operation, analyze its characteristics of dynamics statistics, and calculate road weight of traffic network, so as to get reference information of driving cycle which represent the properties of urban traffic networks. The main works are list as follow:1. With the help of the software Google earth, we queried geographic information (latitude and longitude primarily) of 703 roads and 1270 major road intersection of urban district. This information forms the embryonic form of the electronic map by programming and builds the topological matrix of road connecting. In order to improve efficiency of road match in the future, we carried the grid division on the initial electronic map, built grid-based digital map of Changchun City and the road geographic information Database.2. We carried out a one-year period taxi test installed GPRS traveling data recorder in Changchun City. During this period, the experiments record driving data of taxi day and night continuously. On the basis of the collected data, we established the time-based database of Changchun City taxi test, preparing for the later work.3. We make improvement on the 6-8 algorithm about road match of navigation system, get the road information of the taxi test data and also the taxi driving data of these roads.4. The course of road change while a car's moving in the traffic networks is a random process endowed with the Markov property, which is elaborated and explained. We confirm that this course is time-homogeneous Markov chain with a finite state space, and propose a Markov chain model of road transferring of vehicles traveling.5. On the basis of former work, we make statistics of vehicles turning at intersections so as to get the probabilities of road transition matrix. According to the features of stationary distribution of Markov chain, we calculate road weight through the probabilities of road transition matrix.6. To verify the rightness of the method calculating road weight using Markov chain theory, we work out the road weight by traffic flow model. First, we take the taxi test data which has the same period as the radar test as original data, then build the traffic flow model by means of experimental calibration, thus average density of each road could be obtained according to mean speed of road database. At last, we calculate road weight of every road on the basis of road length and mean density. We comparing the two results, the coefficient correlation is high and the result is approximate. The method calculating road weight using Markov chain model is verified.7. We extracted VA distribution of each road through road match-based taxi database. Later, we make synthesis of VA distribution of the whole city according to road weight, and compare the one synthesized with out road weight.In a conclusion, this paper introduces a new method calculating road weight. The course of road transferring in traffic network is a time-time-homogeneous Markov chain, and we calculate road weight making use of relevant properties of Markov chain. To validate rightness of this method, we work out the road weight using traffic flow model made by means of spot-fixed radar detection speed experimental calibration.