Research And Design On Load Balancing Algorithm Of Microscopic Traffic Simulation

The objective of this research was to investigate, develop and evaluate static load-balancing strategies for distributed execution of microscopic road traffic simulations. With the huge demand of computational resource as well as the requirement of the high speed-up of simulation, there are urgent needs of research on load-balancing strategy of distributed microscopic traffic simulation. The load-balancing strategy can be divided into two categories: static load-balancing strategy and dynamic load-balancing strategy. Due to dynamic load-balancing strategy need to re-allocate the task frequently in the simulation process and slow down the speed of simulation, our research focus on static load-balancing strategy. In this thesis, in order to address the needs of actual traffic simulation project, an model of simualotion time cost and an algorithm to optimize the configuration of Terminals in distributed microscopic traffic simulation system were proposed, and based on this model, the algorithm were compared with the traditional recursive bisection algorithm in the experiment. The experiment result show that the above algorithm we proposed have some advantages. The main work of this thesis was summarized as follows:An algorithm to optimize the configuration of Terminals were proposed. According to the project’s actual need of speed-up ratio, an model of simualotion time cost was built, based on the idea of greedy growing, the number of simulating terminals can be calculated and an partition of the road network can be loaded into these terminals.We designed and completed an experiment, compared with the traditional recursive bisection algorithm, the outstanding performance of our model and algorithm were proved. We designed and implemented the load balance module of DTMS. Our test results show the load balance module of DTMS could meet the functional requirements.The achievement of this research can be used to solve the load-balancing problem of distributed microscopic traffic simulator. On the other hand, the model and the algorithm proposed above can be extended to other distributed simulation system.