Hide/Show Apps

Land vehicle navigation with gps/ins sensor fusion using kalman filter

Download
2009
Akçay, Mustafa Emre
Inertial Measurement Unit (IMU) and Global Positioning System (GPS) receivers are sensors that are widely used for land vehicle navigation. GPS receivers provide position and/or velocity data to any user on the Earth’s surface independent of his position. Yet, there are some conditions that the receiver encounters difficulties, such as weather conditions and some blockage problems due to buildings, trees etc. Due to these difficulties, GPS receivers’ errors increase. On the other hand, IMU works with respect to Newton’s laws. Thus, in stark contrast with other navigation sensors (i.e. radar, ultrasonic sensors etc.), it is not corrupted by external signals. Owing to this feature, IMU is used in almost all navigation applications. However, it has some disadvantages such as possible alignment errors, computational errors and instrumentation errors (e.g., bias, scale factor, random noise, nonlinearity etc.). Therefore, a fusion or integration of GPS and IMU provides a more accurate navigation data compared to only GPS or only IMU navigation data. v In this thesis, loosely coupled GPS/IMU integration systems are implemented using feed forward and feedback configurations. The mechanization equations, which convert the IMU navigation data (i.e. acceleration and angular velocity components) with respect to an inertial reference frame to position, velocity and orientation data with respect to any desired frame, are derived for the geographical frame. In other words, the mechanization equations convert the IMU data to the Inertial Navigation System (INS) data. Concerning this conversion, error model of INS is developed using the perturbation of the mechanization equations and adding the IMU’s sensor’s error model to the perturbed mechanization equation. Based on this error model, a Kalman filter is constructed. Finally, current navigation data is calculated using IMU data with the help of the mechanization equations. GPS receiver supplies external measurement data to Kalman filter. Kalman filter estimates the error of INS using the error mathematical model and current navigation data is updated using Kalman filter error estimates. Within the scope of this study, some real experimental tests are carried out using the software developed as a part of this study. The test results verify that feedback GPS/INS integration is more accurate and reliable than feed forward GPS/INS. In addition, some tests are carried out to observe the results when the GPS receiver’s data lost. In these tests also, the feedback GPS/INS integration is observed to have better performance than the feed forward GPS/INS integration.