The Research On Mixed Traffic Parallel Simulation Based On Grid

The traffic parallel simulation is becoming the effective means to solve the urban traffic problems. This paper does research on the key technologies in parallel simulation considering the characteristics of the macroscopic traffic simulation. Designs and implements Grid Traffic Parallel Simulation (GTPS) System in MPI environment on Globus platform.Two scheduling algorithms (PMM and BM) for heterogeneous computing environment are given. These two algorithms have the same time complexity as the Min-min, which is always used as benchmark of scheduling algorithm. And experiment data show that the makespan of PMM or BM is less than Min-min. This paper analyzes how to use synchronize strategioes (conservative strategy, optimistic strategy, hybrid strategy, self-suitable strategy) in traffic simulation, and discusses their merits and shortcomings. A new synchronize strategy is presented, which is suitable for the macroscopic traffic simulation. This strategy avoid both overheads of communication in conservative strategy and overheads of rollback in optimistic strategy.This paper builds three hypotheses and a traffic simulation model for the GTPS. The model is made up of six sub-models, which are traffic light model, traffic generation model, traffic driving model, velocity restricting model, map description model, map analysis model. The GTPS is designed by object oriented method, and implemented by C++ and MPI on Globus grid platform. It contains four kinds objects, which are traffic entity object, interactive object, communication object and statistic object.