Simulation And Optimization Of Largescale Traffic Network

Because of worsening traffic pressure, dynamic traffic theory has been studied intensively and extensively in the past 30 years. But its application on large-scale network is unsuccessful, due to the network size, some key algorithms, such as shortest path algorithm and DTA solution algorithm, can't meet the demands of practical application. This paper gives the author's results of research on real-time simulation, some key algorithms, DTA algorithm and ITS GIS system for large-scale traffic network.This paper, on basis of cell automaton model and CS model, with the help of multi-thread, realizes high-speed updating of the whole simulation system. This paper also designs a distributed simulation system on RMI, experiments shows that on a traffic network, which consists of three subnetworks with 10x10 traffic sections and 5000 vehicles, the whole network can update within 0.4 second.This paper, based upon community analysis, proposes an algorithm to simplify the large-scale network and a brand-new algorithm, Shortest Path Algorithm based upon Community Analysis(CA-SPA), for shortest path calculation on large-scale network, the time-complexity and validity of this new algorithm are discussed. On an experiment network with 285 nodes, CA-SPA with Dijstra Algorithm as the core shortest path algorithm is obviously superior to the original Dijstra Algorithm and LC Parallel Algorithm in efficiency.This paper, by analyzing the factors that affect the efficiency of parallel computing, the correspondence between various network partition results and these factors, gives a definition of partition index, for scaling the effect of different network partition on the efficiency of parallel computing. The definition is validated by experiment with LC parallel shortest path computing. Based upon community analysis a brand-new partition algorithm is constructed. Theoretical proof shows that this partition algorithm is NPI optimal.On constructing simplified network, this paper decomposes the original DTA problem on large-scale network into two parts: the DTA problem on Simplified Network and the DTA problems on subnetworks, and proposes a parallel GP algorithm for DTA problem. Complexity of this parallel GP algorithm is calculated, it is shown that, comparing with the original GP algorithm, this parallel version is superior on large-scale network.This paper uses deegree, an open source project, as the GIS framework, and integrates with traffic simulation and controlling system, quickly realizes the whole ITS system.