短时交通流预测系统的效率优化研究

Performance Optimization of Short-term Traffic Prediction System

针对短时交通流预测系统的功能需求特点,提出了一个分布式的系统框架,将系统的计算任务按照功能划分到不同的服务器上,从而降低了单个服务器的负荷,提高了整体的效率。通过对系统在北京市实际路网上运行的性能测试,找出了程序执行效率的瓶颈,分析了执行效率问题的主要原因,并给出了对效率问题的具体优化措施。经过实际验证,该技术方案很好地解决了系统的执行效率问题,满足了系统的实时性要求。

As an important part of Intelligent Transportation System,Short-term traffic prediction system provides information for Advanced Traveler Information System and Advanced Traffic Management System.In order for the traffic prediction system to run in real time,the computational efficiency is critical in practical applications,and this problem is especially important for large scale complex urban road networks.This paper analyzes the function requirements of real-time traffic prediction systems and introduces a distributed system architecture.Different tasks are distributed to different servers according to their functions,which lowers the burden of each single server,and improves the overall performance.Furthermore,based on the performance profiling on a real road network in Beijing,the performance bottleneck of the system is found out.The cause of performance problem is analyzed,and the optimization methods are given.The application of the system in Beijing proves that the presented solution solves the performance problem.The resulting prediction system meets the real-time requirement.