论文部分内容阅读
针对车辆运动的非线性特性,利用比支持向量机(SVM)测试时间短、多样本时具有计算量小的相关向量机(RVM)对车辆行驶状态进行估计。为了能够较为准确地估计车辆行驶状态,采集实车试验数据,利用Kalman滤波器对采集到的车速和横摆角速度数据进行滤波,将滤波后的数据作为RVM的输入。依据贝叶斯理论建立最大似然函数,考虑到横摆角速度和车速变化的差异性,依据不同迭代次数下最大似然估计值、伽马值以及值的差异性确定最佳的迭代次数,保证模型具有较短的测试时间和较高的击中概率。有效性验证结果表明:该模型能够较为准确地逼近待估计样本的真值,其中波动性较大的横摆角速度所需要的迭代次数更多,伽马值和值的变化更为迅速,收敛速度较快。
According to the non-linearity of vehicle motion, the vehicle running state is estimated by using a correlation vector machine (RVM) which has a shorter test time than SVM and a small amount of computation in multiple samples. In order to estimate the driving status of the vehicle accurately, real vehicle test data were collected, the collected vehicle speed and yaw rate data were filtered by Kalman filter, and the filtered data was used as the input of RVM. Based on the Bayesian theory, the maximum likelihood function is established. Taking into account the difference of yaw rate and vehicle speed, the optimal iteration number is determined according to the maximum likelihood estimation value, gamma value and value under different iteration times, The model has a shorter test time and a higher probability of hit. Validation results show that the model can approximate the true value of the sample to be estimated more accurately, in which the more fluctuating yaw rate requires more iterations, the gamma value and value change more rapidly, and the convergence rate Faster.