Research On Parallel Traffic Simulation Model And Key Algorithms

Nowadays, Simulation technique is a very popular and effective tool in analyzing many traffic problems, and it can't be displaced by other research methods currently. In all kinds of traffic simulation, microscopic simulation takes individual vehicle as object of study. So it can reflect accurately traffic state and drivers'decision behaviors while they are facing various traffic control measures and incidents. Consequently, it is regarded as the most accurate and valid method in analyzing many traffic problems. Although the traditional sequential simulation can complete simulating traffic on the road network containing several dozen of junctions in a reasonable time, the simulation will become slow dramatically when the scale of road network increase to hundreds of or even thousands of junctions. As a result, the time need for simulation will exceed the real time in most cases, microscopic simulation then become disabled to predict real-time traffic network state online. The theory of parallel computing based on the idea of trading space for time adopts multiple processors to realize parallel simulation, which can provide a feasible approach to speedup traffic simulation within large-scale road network effectively.Although parallel simulation can be performed with dedicated high performance computer to save time, it is too expensive to afford for the common traffic study institutions and traffic management departments, which makes it impossible to be applied widely. So multiple relatively cheaper PCs have been used to establish a distributed cluster through LAN (Local Area Network) in this dissertation, the task of traffic simulation in the whole large-scale road network is then divided and assigned to all the node PCs who collaborate and computer in parallel to accomplish the whole simulation process. By this way the speed and efficiency of simulation can be improved greatly. Furthermore, this is a feasible method of widespread application easily for many traffic study institutions and traffic management departments. Firstly, the necessity of application of parallel computing technique to traffic simulation was analyzed and the relating basic conceptions of parallel computing theory and the popular parallel computing circumstances are introduced in this paper. Moreover, the way and key technique of constructing parallel computing workstation cluster was investigated. The applicability of the two massage passing library, PVM and MPI, was concluded respectively by comparing their merit and weakness. A distributed parallel computing cluster was then established with many PCs, server and exchanger through LAN.Secondly, with combining the current development status of traffic flow theory, the common methods of traffic flows modeling and analyses including macroscopic, mecroscopic and microscopic modeling were summarized. In order to make sure that traffic simulation can reoccur exactly the process of traffic performance, the core models in microscopic traffic simulation model, such as vehicle generating model, car following model and lane changing model, were studied with emphasis. On this basis, the work flow and implementation frame of microscopic traffic simulation were designed. With the analysis deeply of necessity and feasibility of parallel microscopic traffic simulation, parallel microscopic traffic simulation model based on the way of spatial domain decomposition and collaborating subnets strategy were suggested by making advantage of the characteristics of distributed parallel computing technique.Thirdly,key algorithms necessary for implementing distributed parallel microscopic traffic simulation were focused on. The method of orthogonal recursive bisection and multilevel k-way partition were used to divide the whole simulation road networks into multiple independent relatively subnet areas. The synchronization strategy of Conservative Time Windows (CTW) was applied to guarantee the correct performance of parallel traffic simulation among all the road subnets. At the same time, dynamics load balancing mechanism was implemented based on local algorithm in order to reduce the idle time of processors during parallel simulation.Once more, the presentation method of traffic network data was designed, including road network facilities model, the object-oriented abstract of all kinds of network components, and data structures used in parallel microscopic traffic simulation. The way to save and manage traffic simulation projects using database technology was also studied. The hardware and software environment under which the parallel traffic simulation system is developed was established. On the basis of detail analysis of traffic simulation system demand, TPSS (Traffic Parallel System) was designed and implemented successfully using the PVM message passing library and the programming language C++.At last, the validity of TPSS system to simulate network traffic accurately was confirmed through testing a traffic simulation project within a small road network in Changchun city. Grid traffic networks of various scales were built and some key traffic flow parameters were set reasonably. Under the testing environment the changing trend of the benefit of parallel simulation with the increasing of the network scale and the number of the vehicles as well as that of the parallel simulation speedup ratio with the increasing number of node computers in the cluster were analyzed. TPSS was finally applied to the traffic improvement project of Dajing road and Wuma road intersection in Changchun city, and a reasonable traffic control alternation was provided.The application of parallel computing technique to traffic simulation is one of the research hotspots in the field of simulation in recent years. The parallel traffic simulation model based on the method of spatial domain decomposition not only can speed up the traffic simulation in large-scale road network greatly, but also has theoretical reference and practical application value for the development and implementation of traffic status online prediction and inducement system in ITS.