电子学报 ›› 2017, Vol. 45 ›› Issue (10): 2473-2483.DOI: 10.3969/j.issn.0372-2112.2017.10.023

• 学术论文 • 上一篇    下一篇

一种基于Kalman滤波和粒子群优化的测试数据生成方法

薛猛1, 姜淑娟1,2, 张争光1, 钱俊彦2, 张艳梅1,3, 曹鹤玲4   

  1. 1. 中国矿业大学计算机科学与技术学院, 江苏徐州 221116;
    2. 桂林电子科技大学广西可信软件重点实验室, 广西桂林 541004;
    3. 南京大学计算机软件新技术国家重点实验室, 江苏南京 210093;
    4. 河南工业大学信息科学与工程学院, 河南郑州 450001
  • 收稿日期:2016-08-07 修回日期:2017-04-23 出版日期:2017-10-25 发布日期:2017-06-27
  • 通讯作者: 姜淑娟
  • 作者简介:薛猛,男.1979年4月出生,江苏邳州人.博士研究生,讲师,CCF会员,主要研究领域为软件测试、测试数据生成.E-mail:mgxue@cumt.edu.cn
  • 基金资助:
    国家自然科学基金(No.61502497,No.61562015,No.61673384,No.61602154);中国博士后科学基金(No.2015M581887);广西可信软件重点实验室研究课题(No.KX201530);南京大学计算机软件新技术国家重点实验室开放课题(No.KFKT2014B19);徐州市科技计划项目(No.KC15SM051);河南省高等学校重点科研项目计划资助(No.16A520005)

A Test Data Generation Method Based on Kalman Filter and Particle Swarm Optimization Algorithm

XUE Meng1, JIANG Shu-juan1,2, ZHANG Zheng-guang1, QIAN Jun-yan2, ZHANG Yan-mei1,3, CAO He-ling4   

  1. 1. School of Computer Science and Technology, China University of Mining and Technology, Xuzhou, Jiangsu 221116, China;
    2. Guangxi Key Laboratory of Trusted Software, Guilin University of Electronic Technology, Guilin, Guangxi 541004, China;
    3. State Key Laboratory for Novel Software Technology, Nanjing University, Nanjing, Jiangsu 210093, China;
    4. College of Information Science and Engineering, Henan University of Technology, Zhengzhou, Henan 450001, China
  • Received:2016-08-07 Revised:2017-04-23 Online:2017-10-25 Published:2017-06-27

摘要: 为减少进化代数,提高路径覆盖成功率,提出了多邻域Kalman滤波PSO测试数据生成方法.在该方法中将粒子固定划分到不同邻域中,各邻域内指定一个粒子向全局最优粒子学习,其余各粒子向所在邻域中最优粒子学习,而全局最优粒子利用无速度项的简化PSO进化.在此过程中,除全局最优粒子外的各粒子利用Kalman滤波方程更新粒子的位置.实验表明,相较于基本PSO和其他PSO方法,即使是覆盖困难的路径,本文方法也具有进化代数少、路径覆盖成功率高及性能稳定的特点.

关键词: 测试数据生成, 粒子群优化, Kalman滤波, 邻域拓扑

Abstract: A test data generation method named multi-neighborhood Kalman filter PSO(MNKFPSO) was proposed to reduce the evolution number and to improve the success rate of path coverage.Particles except the global best one update themselves' positions using Kalman filter.One of them is allotted to a fixed neighborhood.A designated particle learns from the global best particle,others learn from the best in one neighborhood.And the global best particle's position changes by a simple PSO which discards the particle velocity.The experimental results show that it can generate test data covering the specified path in the less evolutionary using MNKFPSO and has high success rate of path coverage even though the paths difficult to cover.The algorithm also exhibits a stable performance.

Key words: test data generation, particle swarm optimization (PSO), Kalman filter, neighborhood topology

中图分类号: