Double-Fuzzy Kalman Filter Based on GPS/IMU/MV Sensor Fusion for Tractor Autonomous Guidance

Sensor fusion technique has been commonly used for improving the navigation of autonomous agricultural vehicles by means of combining sensors mounted on such vehicles for the position and attitude angle measurements. In this research, a real-time tractor position estimation system, which is consisted with the global positioning system (GPS), the six-axis inertial measurement unit (IMU) and the machine vision (MV) was discussed. A double-fuzzy Kalman filter (DFKF) was used to fuse the information from these sensors so that the noise in the GPS and the machine vision signals was filtered, the redundant information was fused and a higher update rate of output signals was obtained. The drift error of the IMU was also compensated. One of the double-fuzzy logic controller was designed to modify the filter gain matrix K and the measurement noise covariance R on line based on dead reckoning algorithm, and the other fuzzy logic controller was designed to modify the process noise covariance Q on line based on the variety of the innovation vector. Through trials with simulated data the procedure's effectiveness is shown to be quite robust at a variety of noise levels and relative sample rates for this practical problem.