基于并行遗传算法的公路线形优化  

Optimization of Road Alignment Using Parallel Genetic Algorithms

作　　者：陈国军[1] 刘岩 

机构地区：[1]中国石油大学(华东)计算机与通信工程学院,青岛266580

出　　处：《系统仿真学报》2013年第10期2332-2336,共5页Journal of System Simulation

基　　金：北京航空航天大学虚拟现实与可视化技术国家重点实验室开放课题(BUAA-VR-10KF-04)

摘　　要：传统的公路线形优化方法效率低下,不能满足实时交互选线。提出预先筛选和并行计算的优化效率算法。首先根据公路路线的垂直和水平拟合的要求,建立计算曲线的几何约束关系。在此基础上通过剔除违反水平和垂直公路线形设计约束减少种群数量,再将遗传算法中的关键步骤种群迁移、遗传操作等分解为多个并行执行过程,给出各自的并行计算模型,并在GPU上利用CUDA实现。实验表明,提出的方法在保证选线较优的情况下,预先筛选提高了算法执行时间,而并行执行算法执行速度提高显著。Traditional highway alignment optimization methods are inefficient in interactive highway selection, which can＇t meet requirement of real-time. A parallel prescreened highway alignment optimization method was proposed. First, highway curve geometric constraints relationship was established according to the vertical and horizontal highway alignment fitting requirements. Then population size was decreased by prescreening non-feasible highway alignments which violate design constraints. The key steps of genetic algorithm, such as population migration, genetic manipulation etc, were decomposed into several independent processes, and the parallel model of each process was implemented with CUDA on GPU accordingly. The results indicate that the proposed method can get a fair good highway alignment with good efficiency.

关 键 词：公路线形优化 并行遗传算法 GPU CUDA 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]