农业植保无人机变结构线性滚动时域估计器研究

Variable Structure Linear Moving Horizon Estimator for Plant Protection UAV

植保无人机的惯导数据输出频率最高可达8 k Hz,而常用GPS输出频率最高只有10 Hz,会导致位置估计器的数据源之间存在延时问题。为此设计了一种处理延时测量值的变结构线性滚动时域估计器(Variable structure linear-Moving horizon estimator,VSL-MHE)。首先将多传感器测量数据进行序列化;然后通过线性索引来确定排列项中无数据的坐标,根据该坐标对代价函数中权重矩阵进行更新,从而改变估计器结构;最后通过VSL-MHE估计植保无人机位置。在室内测试中,通过Opti Track向植保无人机发送模拟GPS信号,输出频率范围为2~20 Hz。将VSL-MHE估计的位置、MHE估计的位置和循环迭代扩展卡尔曼滤波器(Circular iterated extended Kalman filter,CIEKF)估计的位置分别与Opti Track获取的精确位置做对比,结果表明,VSL-MHE的位置最大偏差小于CIEKF和MHE的位置最大偏差。在室外40 m×30 m范围的飞行测试中,VSL-MHE的航线位置最大偏差小于CIEKF和MHE的位置最大偏差,验证了该算法的有效性。

In order to solve the delay caused by different sampling frequencies of multi-sensors in position estimation of plant protection unmanned aerial vehicle (UAV), a variable structure linear moving horizon estimator (VSL MHE) was designed to process delay measurement values. Firstly, the measurement data of multiple sensor was serialized. Then the linear index method was used to determine the coordinates of the zero elements in the arrangement term, and the weighting matrix in the cost function was updated with the coordinates of the zero elements to change the structure of the estimator. Finally, the plant protection UAV position was estimated by VSL MHE. The accuracy of the position estimation was verified from the actual flight tests. In the indoor experiment, the GPS output was simulated by the output of OptiTrack motion capture system, the frequency could be controlled in the range of 2~20 Hz. Through the fixed-point hover experiment, the positions estimated from VSL MHE, MHE and the circular iterated extended Kalman filter (CIEKF) were compared with that obtained from OptiTrack, respectively. The maximum positional offset of VSL MHE was smaller than the positional offset of the CIEKF and MHE. In the outdoor experiment, the range of path planning was set to be 40 m×30 m, and the maximum positional deviation from the VSL MHE was less than that estimated from CIEKF and MHE. The experiment results showed that VSL MHE can effectively reduce the adverse effects on the accuracy of position estimation.