中文摘要：

拖拉机作业环境恶劣,测量信号容易受到噪声干扰,其滑转率的计算过程对于输入信号的相对误差有极强的放大作用,因而造成其滑转率难以精确测量。该文提出带噪声观测器的变结构并行自适应数据融合算法,对轮速传感器、角加速传感器、车身加速度计和全球定位系统的信号进行融合,在不需要先验误差统计规律的前提下实现了对拖拉机驱动轮滑转率的在线精确估计。仿真测试结果证明:采用信息融合方法求得的驱动轮滑转率信号几乎与理论值曲线重合且鲁棒性好,平均误差为中值滤波的1/10左右,为卡尔曼滤波的1/5;算法的噪声观测器能够实时估算测量信号的白噪声方差,求得的稳态平均方差与已知精确先验误差的卡尔曼数据融合算法无明显差异;在从动轮速度信号受到有色随机噪声干扰的特殊工况下,算法的信息融合机制能够补偿大部分由有色噪声干扰造成的误差。实测试验证明:在拖拉机稳定工作工况,在线求得的测量信号噪声方差均值在5%的范围内波动,采用数据融合算法求得的驱动轮滑转率误差均值为0.012,误差绝对值最大值为0.027,与离线拟合得到的参考值非常接近。该研究为拖拉机实现精确控制提供了参考,其在线测量信号方差统计方法为拖拉机总线网络的传感器信息共享提供了技术基础。

英文摘要：

Calculation process strongly amplifies the relative error of input signal, which makes it difficult to measure the sliding rate. The key to obtain the accurate value of the sliding rate lies in the real-time and accurate measurement of the tractor speed and the theoretical speed of the driving wheel. Multi sensor information fusion algorithm based on Kalman filtering can effectively improve the measurement accuracy. Due to the development of controller area network, sliding rate measuring node can share the information of other sensors on the bus based on ISO11783 protocol, which provides a convenient condition for realizing multi sensor fusion. However, the measurement noise variance of the sensor signals of the tractor is unpredictable, which is the problem that the algorithm must solve. Aimed to this problem, the adaptive data fusion algorithm with noise observer is proposed in this paper. Multi sensor signals from wheel speed sensor, angular acceleration sensor, vehicle body accelerometer and global positioning system are integrated by this algorithm, and at the same time, the noise variance of the sensor signal is calculated online, so as to accurately estimate the tractor driving wheel’s sliding rate online without the prior noise measurement signal variances. The simulation results show that the sliding rate estimated by the proposed algorithm is almost coincident with the theoretical value. The average error of the sliding rate estimated by adaptive data fusion algorithm is about 1/10 of that by the median filtering method, and 1/5 of that by the Kalman filtering method. The algorithm has good robustness. The noise observer of the algorithm can estimate the white noise variance of the measured signals in real time, and there is no significant difference between the average estimated variance of steady state and the Kalman data fusion algorithm with exact prior error. The algorithm has the mechanism of the fusion according to the weight of the signal covariance, and the distortion signal can be modified