计算机科学
卡尔曼滤波器
理论(学习稳定性)
控制理论(社会学)
超参数
车辆动力学
电动汽车
国家(计算机科学)
特征(语言学)
残余物
算法
荷电状态
贝叶斯概率
递归贝叶斯估计
系列(地层学)
瞬态(计算机编程)
传感器融合
模型预测控制
卷积神经网络
递归(计算机科学)
人工神经网络
滤波器(信号处理)
领域(数学)
弹道
稳健性(进化)
卡西姆
特征提取
时间序列
扩展卡尔曼滤波器
钥匙(锁)
作者
Haoran Li,Haichao Zhou,Guolin Wang,Hai Wang,Rongyun Zhang,Binghao Yu
标识
DOI:10.1088/1361-6501/ae527c
摘要
Abstract To improve the handling and stability of distributed drive electric vehicles, the study proposes a method integrating high accuracy vehicle state estimation and tire longitudinal and lateral forces prediction. First, based on the seventh-order spherical radius criterion, the proposed square root seventh-order cubature Kalman filter (SR7thCKF) algorithm significantly enhances numerical stability and estimation precision, achieving robust fusion of sensor data and high-accuracy vehicle state parameters (VSPs) estimation. Subsequently, based on the time series characteristics of the estimated VSP and sensor information, the temporal convolutional network (TCN) hierarchically expands the receptive field through dilated causal convolutions, capturing multi-scale temporal patterns, while the gated recurrent unit (GRU) further model global temporal dynamics and long-range dependencies via its adaptive memory gating mechanism. The hyperparameters of the TCN-GRU framework are automatically tuned by Bayesian optimization (BO) algorithm to balance prediction accuracy and computational efficiency, thereby constructing the BO-TCN-GRU model is developed for tire longitudinal and lateral forces prediction. Finally, the results show the high-precision vehicle state estimation of SR7thCKF and the advantage of the hierarchical feature extraction capabilities of BO–TCN–GRU for temporal characteristics.
科研通智能强力驱动
Strongly Powered by AbleSci AI