摘要
鉴于常规卡尔曼滤波算法组合导航系统数据融合算法中,存在易于发散的缺陷,尝试将遗传优化人工神经网络引入组合导航系统中。针对传统遗传算法存在的易早熟、算法稳定性差、固定的交叉和变异概率影响收敛效果等缺点,采用浮点式编码方式,两两竞争的选择策略、引入突变操作、重新定义交叉算子和自适应的交叉变异算子等措施进行了遗传算法的改进。仿真结果表明,改进后的算法更为有效,并且精度与常规卡尔曼滤波算法相当。
As the conventional Kalman filter is liable to get divergence in integrated navigation system data fusion, an artificial neural network based on the genetic algorithms was applied in the system. But there are drawbacks of prematurity, bad stability, fixed cross and mutation probability in the conventional genetic algorithms, so an improved genetic algorithm was put forward. The improvements include float coding, competition selection strategy, introduction of mutation opertation, redefined crossover operator and adaptive crossover-mutation operator, etc.. The simulation results indicate that the algorithm is more effective, and its precision is equivalent to that of conventional Kalman filter.
出处
《中国惯性技术学报》
EI
CSCD
2006年第5期24-27,共4页
Journal of Chinese Inertial Technology
基金
国家自然科学基金项目(60374046和50575042)
总装备部国防预研基金项目(514090101035W0609)
关键词
人工神经网络
改进遗传算法
组合导航系统
数据融合
artificial neural network
improved genetic algorithm
integrated navigation
data fusion
