Abstract: In the study of the attitude estimating optimization of the teaching arm, aiming at the low accuracy and poor stability of the traditional method that use single set of sensors for data fusion and attitude estimation, a method based on multi sets of MEMS sensors is designed. Six sets of sensors are mounted on the three axes of the carrier coordinate system, measuring two sensor data separately. The vector product of the sensor data measured by the combination and estimated by quaternion method is used as the input of the complementary filter instead of the attitude error. Fuzzy controller and PI controller adjusted the gyroscopes outputs separately according to the principle of complementary filter. Attitude is estimated through extended Kalman filter and quaternion with higher accuracy is achieved. Then it is transformed to attitude angulars. Simulation results show that, instatic and dynamic situations, the accuracy and stability of attitude data with combined adjustment based on multi sensors are improved more than that with PI adjustment based on single sensor.

Key words: MEMS sensors, extended Kalman filters, quaternion, attitude estimation

摘要： 在示教机械臂姿态解算精度优化的研究中，针对使用单组传感器进行数据融合，姿态解算的传统方法中存在的精度低，稳定性差的问题，设计了一种组合MEMS传感器的姿态解算方法。将六组传感器安装于载体坐标系三个轴上，分别测量两组传感器数据。以传感器量测数据与四元数估计数据的向量积代替姿态角误差作为互补滤波器的输入量，分别利用模糊控制器和PI控制器，根据互补滤波原理调节陀螺仪输出量。通过拓展卡尔曼滤波器进行姿态估计，得到更精确的四元数，进而转化为姿态角。仿真结果表明，在静态和动态情况下，多组传感器组合调节后的姿态角数据相比单组传感器PI调节在姿态角精度和系统稳定性上有进一步提高。

关键词: MEMS传感器, 拓展卡尔曼滤波器, 四元数, 姿态解算