重庆理工大学学报(自然科学) ›› 2023, Vol. 37 ›› Issue (4): 85-94.

• 车辆工程 • 上一篇    下一篇

改进支持向量机的车辆定位导航精度提升方法

岳钰隽,邱 娜,金志扬   

  1. 1.海南大学 机电工程学院,海口 570228; 2.清华大学 苏州汽车研究院,江苏 苏州 215134; 3.清华大学 车辆与运载学院,北京 100084; 4.香港理工大学 土木及环境工程学系,香港
  • 出版日期:2023-05-06 发布日期:2023-05-06
  • 作者简介:岳钰隽,男,硕士研究生,主要从事车辆导航研究,Email:18009065050@163.com;通信作者 金志扬,男,博士,教 授,主要从事智能驾驶研究,Email:jinzhiyang94@163.com。

Research on accuracy improvement of vehicle positioning and navigation with an improved support vector machine

  • Online:2023-05-06 Published:2023-05-06

摘要: 车辆定位导航是实现智能车辆环境感知的基础,为解决智能车辆在 SINS/GPS组合 导航下误差问题,提出一种基于蚁群算法改进支持向量机的车辆定位导航精度提升方法。首 先,使用状态变换扩展卡尔曼滤波对组合导航进行初步降噪;其次,运用支持向量机及神经网络 辅助导航,解决组合导航中位置误差较大、对导航效果产生影响的问题;然后,通过蚁群算法改 进支持向量机,对支持向量机核函数参数进行迭代优化;最后,在实车采集数据集下,与神经网 络辅助进行对比。结果表明:在东北天 3个方向上,神经网络降低误差均方根值的效果达到了 72.88%、68.66%、63.87%,而改进支持向量机的效果达到了 82.09%、79.62%、90.14%。改进 支持向量机能够辅助优化组合导航位置误差,提升车辆定位导航精度。

关键词: 支持向量机, 组合导航, 误差优化, 机器学习, 蚁群算法

Abstract: Vehicle positioning and navigation is the basis of realizing environment perception of intelligent vehicles. To solve the error problem of intelligent vehicles under SINS/GPS integrated navigation, this paper proposes a method of improving vehicle positioning and navigation accuracy based on an improved support vector machine with ant colony algorithm. Firstly, an extended Kalman filter with state transformation is proposed to reduce noise of the integrated navigation system. Secondly, the support vector machine and the neural network aided navigation are proposed to solve the problem of large position error and the influence on navigation effect in the integrated navigation. Then, the support vector machine is improved by ant colony algorithm, and the kernel function parameters of the support vector machine are optimized iteratively. Finally, it is compared with the neural network assistance in the real vehicle collection data set. The results show that the neural network can reduce the root mean square value of error by 72.88%, 68.66% and 63.87% in the three directions of east, north and up (ENU), while the improved support vector machine can achieve 82.09%, 79.62% and 90.14%. The improved support vector machine can help optimize the position error of the integrated navigation and improve the accuracy of vehicle positioning and navigation.

中图分类号: 

  • P228.4