汽车工程 ›› 2018, Vol. 40 ›› Issue (7): 770-.doi: 10.19562/j.chinasae.qcgc.2018.07.004

• • 上一篇    下一篇

基于NA-EKF的分布式驱动电动汽车行驶状态估计研究

耿国庆,韦斌源,江浩斌,华一丁,吴镇   

  • 出版日期:2018-07-25 发布日期:2018-07-25

A Research on Driving State Estimation for Distributed Drive Electric Vehicle Based on NAEKF

Geng Guoqing, Wei Binyuan, Jiang Haobin, Hua Yiding & Wu Zhen   

  • Online:2018-07-25 Published:2018-07-25

摘要: 精确的行驶状态估计对分布式驱动电动汽车(DDEV)的纵、横向稳定性控制具有至关重要的意义。本文中应用噪声自适应扩展卡尔曼滤波(NAEKF)算法对DDEV行驶状态进行估计。该算法充分利用观测信号的实时统计信息,通过监测滤波器新息和残差的动态变化,不断修正状态噪声方差和量测噪声方差,从而调整滤波器增益、状态预测值和观测值在滤波后的状态值中的比例,提高状态估计精度。最后利用车辆动力学仿真软件veDYNA对本文应用的算法进行了仿真验证,结果表明:与EKF相比,该算法可有效克服先验统计信息不准确和复杂工况下造成估计不准确的问题,状态量估计的平均误差不超过27%,均方根误差不超过26%,峰值相对误差较小。

关键词: 分布式驱动, 行驶状态, 噪声自适应, 扩展卡尔曼滤波, 分布式驱动, 行驶状态, 噪声自适应扩展卡尔曼滤波

Abstract: It is greatly important to accurately estimate the driving state for the longitudinal and lateral stability control of a distributed drive electric vehicle (DDEV). Noise adaptive extended kalman filtering (NAEKF) algorithm is adopted to estimate the driving state of DDEV in this paper. The algorithm makes full use of the realtime statistical information of observed signal. By monitoring the dynamic change of innovations and residuals of filter, the state noise variance and measurement noise variance are constantly corrected, with filter gain and the proportions of predicted and observed state values in state values filtered adjusted, hence improving state estimation accuracy. Finally the algorithm adopted is verified by simulation with vehicle dynamics software veDYNA. Results show that compared with EKF, the algorithm adopted can effectively overcome the problem of inaccuracy in prior statistical information and the inaccuracy in estimation caused by complex conditions, with a mean error of state estimation not more than 27%, a root mean square error not more than 26%, and a rather small peak relative error.

Key words: distributed driving, running state, noise adaptive, extended kalman filter, distributed drive, driving state, noise adaptiveextended kalman filtering