Agricultural machinery GNSS/IMU-integrated navigation based on fuzzy adaptive finite impulse response Kalman filtering algorithm

Abstract

This study uses global navigation satellite system (GNSS) positioning equipment and inertial measurement unit integrated with accelerometer and gyroscope to improve the accuracy and stability of the current agricultural machinery automatic navigation technology. Considering the actual motion state of agricultural machinery in operation, a fuzzy adaptive finite impulse response Kalman filter (FA-FIR-KF) algorithm was proposed to integrate position information and attitude information, and some necessary auxiliary optimization algorithms were introduced to make innovative improvements. The introduction of quaternion method can suppress the actual nonlinear problem of the agricultural machinery coordinate caused by the attitude angle. A fuzzy inference system was adopted to improve the adaptive adjustment ability to abnormal noise. A forgetting factor was adopted to reduce the system’s excessive dependence on prior statistical information, so that the system can quickly track the abrupt signal. The algorithm simulation program was written on MATLAB, and the performance and effect of the proposed algorithm were verified through simulation and farm experiments. Simulation results of artificially added noise simulation data show that the localization precision in the Xn, Yn, and Zn directions increases by 38.95%, 38.88%, and 32.99%, respectively. This finding indicates that the FA-FIR-KF algorithm can effectively suppress the Gaussian white noise of the GNSS received signal and improve the positioning accuracy of agricultural machinery. The reliability of this algorithm applied to the automatic navigation system was verified through a tractor straight-line navigation experiment. The tractor conducts an automatic navigation test at a speed of 0.8 m/s. Under the GNSS differential state, the average error and root mean square error (RMSE) are 1.074 and 1.396 cm in filtering case and 1.17 and 1.551 cm in nonfiltering case, respectively. Under the GNSS nondifferential state, the average error and RMSE are 2.097 and 2.72 cm in filtering case and 3.663 and 4.633 cm in nonfiltering case, respectively. Compared with the nonfiltering case, the average error and RMSE reduce by 8.21% and 9.99% in the differential state and 42.75% and 41.32% in the nondifferential state, respectively. Test results show that the proposed algorithm can make the agricultural machinery track the desired path more smoothly, stably, and accurately than in the nonfiltered case, and the tracking accuracy is at the centimeter level.