基于交互式多模型滤波的车辆状态可重构估计算法

Vehicle State Reconfigurable Estimation Algorithm Based on Interacting Multiple Model Filtering

摘要:

为了解决车辆状态估计精度低和移植性差的问题，提出了一种车辆状态可重构估计算法。设计了一套可重构估计算法架构，能兼容多种传感器信号输入；结合车辆运动特性和多传感信号特征，开发了一套交互式多模型卡尔曼滤波融合估计器，对不同传感器输入均可实现车辆状态参数的精确估计，并可以通过可重构矩阵实现算法高效复用；搭建了多传感器输入的实车测试环境，在典型工况下进行了算法验证。实验结果表明，所提出的可重构估计算法具有较高的准确性，纵向速度和横向速度的估计平均误差小于0.6 m?s^-1，横摆角速度时延更短、精度更高。

Abstract:

To solve the issues of low accuracy and poor portability in industrial vehicle state estimation， this paper proposes a vehicle state reconfigurable estimation algorithm. First， a reconfigurable estimation architecture for vehicle state was designed， compatible with multiple sensor inputs. Then， an interacting multiple model （IMM） Kalman filter fusion estimator was developed based on vehicle motion characteristics and multi-sensor signal features， enabling accurate estimation of vehicle state parameters for different sensor configurations and efficient algorithm reuse through a reconfigurable matrix. A real vehicle testing environment with multi-sensor inputs was then established， and the algorithm was validated under typical driving conditions. The experimental results show that the proposed reconfigurable estimation algorithm achieves high accuracy， with mean estimation errors below 0.6 m?s^-1 for longitudinal and lateral velocities， and reduced time delay along with higher accuracy for yaw rate.