Electrical Engineering Department, Universidad de La Frontera, Temuco 4811230, Chile.
Magister en Ciencias de la Ingeniería, Universidad de La Frontera, Temuco 4811230, Chile.
Sensors (Basel). 2024 Mar 13;24(6):1849. doi: 10.3390/s24061849.
This work is focused on developing a self-calibration algorithm for an orientation estimation of cattle movements based on a quaternion Kalman filter. The accelerometer signals in the earth's frame provide more information to confirm that the cow is performing a jump to mount another cow. To obtain the measurements in the earth's frame, we propose a self-calibration method based on a strapdown inertial navigation system (SINS), which does not require intervention by the user once deployed in the field. The self-calibration algorithm uses a quaternion-based Kalman filter to predict the angular orientation with bias correction, and update it based on the measurements of accelerometers and magnetometers. The paper also depicts an alternate update to adjust the inclination using only the accelerometer measurements. We conducted experiments to compare the accuracy of the orientation estimation when the body moves similarly to cow mount movements. The comparison is between the proposed self-calibration algorithm with the IvenSense and Bosch and the quaternion attitude estimation provided in the . The auto-calibrating algorithm presents a mean error of 0.149 rads with a mean consumption of 308.5 mW, and the Bosch algorithm shows an average error of 0.139 rads with a mean consumption of 307.5 mW. When we executed this algorithm in an , the average error was 0.077 rads, and the mean consumption was 277.7 mW.
这项工作专注于开发一种基于四元数卡尔曼滤波器的牛运动方向估计的自校准算法。地球框架中的加速度计信号提供了更多信息，可以确认奶牛正在进行跳跃以爬上另一头奶牛。为了在地球框架中获得测量值，我们提出了一种基于捷联惯性导航系统 (SINS) 的自校准方法，一旦在现场部署，就不需要用户干预。自校准算法使用基于四元数的卡尔曼滤波器进行带有偏差校正的角方向预测，并根据加速度计和磁力计的测量值进行更新。本文还描述了一种仅使用加速度计测量值来调整倾斜度的替代更新。我们进行了实验，比较了当身体运动类似于奶牛攀爬运动时，方向估计的准确性。比较对象是提出的自校准算法与 IvenSense 和 Bosch 以及. 中提供的四元数姿态估计。自校准算法的平均误差为 0.149 弧度，平均功耗为 308.5 mW，而 Bosch 算法的平均误差为 0.139 弧度，平均功耗为 307.5 mW。当我们在 An 中执行此算法时，平均误差为 0.077 弧度，平均功耗为 277.7 mW。
