This article describes principles of motion sensors data fusion. We analyze various methods of inertial sensors fusion to ensure more accurate and reliable measurements by using the benefits of multiple sensors to eliminate their individual weaknesses. An Attitude Heading Reference System (AHRS) provides 3D orientation (roll, pitch, and yaw) of a moving object (car, aircraft, drone, etc.) with position and heading information. For implementation of a low cost AHRS system Micro-Electrical-Mechanical System (MEMS) based sensors were used (accelerometer, gyroscope and magnetometer). Accelerometers suffer from errors caused by external accelerations that add up to gravity and, such errors make accelerometers-based rotation inaccurate. Gyroscopes can compensate for such errors, but they also create drifting problems. The direction of the measured Earth magnetic field (magnetometer) is used as a (3D) compass to determine the direction of the North (heading or yaw). A locally disturbed (warped) magnetic field causes an error in orientation that can be quite substantial. So, for the precise data additionally three very common and well-known filters were introduced to the system: moving average filter, complementary and Kalman filters. In this paper a comparison of system performance is shown separately, so that the filter with the best performance can be chosen for a specific system. In the practical part we use Xsens Mti-20 inertial measurement unit. Three sensors' data processing and filtering methods are presented: Kalman filter, moving average filter and complimentary filter. The moving average filter provides simple data filtering by averaging within a window of specified length, while the complimentary filter performs fusion of accelerometer and gyroscope data by calculating pitch and roll angles. In order to evaluate the performance of each filter, measurements were made in real time for a moving vehicle.