欢迎访问《汽车安全与节能学报》，
汽车安全与节能学报

汽车安全与节能学报 ›› 2023, Vol. 14 ›› Issue (4): 457-462.DOI: 10.3969/j.issn.1674-8484.2023.04.007

• 智能驾驶与智慧交通 • 上一篇    下一篇

在自动驾驶车辆测试中计算加速度的变步长差分算法

魏文明1(), 刘伟平1(), 范志鹏2, 白创1   

  1. 1.国汽（北京）检测技术有限公司，北京 100176，中国
    2.宁波吉利汽车研究开发有限公司，宁波 315000，中国
  • 收稿日期:2022-09-03 修回日期:2023-04-21 出版日期:2023-08-31 发布日期:2023-08-31
  • 作者简介:魏文明（1991—），男（汉），江西，硕士研究生。Email：1253483846@qq.com。
    刘伟平（1985—），男（汉），江西，高级工程师。Email：liuweiping@china-icv.cn。

Differential algorithm with variable step-sizes to calculate accelerations in autonomous vehicle tests

WEI Wenming1(), LIU Weiping1(), FAN Zhipeng2, BAI Chuang1   

  1. 1. National Innovation Center of Intelligent and Connected Vehicles, Beijing 100176, China
    2. Ningbo GEELY Automotive Research and Development Institute, Ningbo 315000, China
  • Received:2022-09-03 Revised:2023-04-21 Online:2023-08-31 Published:2023-08-31

RichHTML

6

PDF

52

摘要：

为降低自动驾驶测试的成本并保证精度，不用惯性测量单元，而用了对速度进行差分的方法来计算加速度。对差分计算误差分析。对比了等步长差分、均值等步长差分及变步长差分等3种算法。与实车测量的结果对比。结果表明：合适的时间步长可提高计算精度。当步长取0.2 s或0.3 s时，等步长差分计算结果较符合实际；当步长相同时，均值等步长差分结果的波动小于等步长差分。等步长差分、均值等步长差分、变步长差分的最大绝对误差分别为0.77、0.64、0.46 m/s2，绝对误差平均值分别为93、82、65 mm/s2。因此，变步长差分算法的精度是3种方法其中最高的。

关键词: 自动驾驶测试, 加速度, 变步长, 差分算法, 计算精度

Abstract:

A differential algorithm was developed without using any inertial measurement unit, to calculate the acceleration of autonomous vehicles, and to reduce the test cost with a rational calculation accuracy. The calculation error was analyzed to contrast three algorithms, which had the fixed step-size, or the mean fixed step-size, or the variable step-size. The calculation results were compared with the measured accelerations in real car. The result shows that a suitable step-size improves the calculation accuracy; The calculation accuracies of the algorithms with 0.2-s or 0.3-s step-size are better than that of the algorithms with other step-sizes when using fixed step-size; The fluctuation of the calculation results by using the mean fixed step-size is smaller than that with the fixed step-size when the step-size is the same; The maximum absolute errors is 0.77 m/s2 for the fixed step-size algorithm, with being 0.64 m/s2 for the mean fixed step-size algorithm, and with being 0.46 m/s2 for the variable step-size algorithm; The average absolute errors are 93, 82, and 65 mm/s2 respectively for the fixed step-size algorithm, the mean fixed step-size algorithm and the variable step-size algorithm. Therefore, the calculation accuracy by using the variable step-size algorithm is the best among the three algorithms.

Key words: autonomous vehicle tests, accelerations, variable step-sizes, differential algorithms, calculation accuracy

中图分类号: