Hardware support for research of the sensor fusion of inertial sensors

Abstract

The aim of this paper was to propose a design of a module that has several inertial sensors of the same type in order to test various approaches of homogeneous sensor fusion. According to the statistics the mean of readings from the same-type sensors should have higher precision than a single sensor. However, this statement is not always correct for real sensors, as identical sensors may not have the same error characteristics. Sensor manufacturers state the typical sensor RMS (root mean square) error, the actual sensor RMS error can differ significantly from piece to piece. When averaging a sensor output from the same manufacturer, we can under certain conditions, obtain a worse value than the output error of the best sensor. This error can be eliminated by fusing the sensors using weighing. To verify this statement, we decided to assemble with as many identical sensors as possible. The IMU (inertial measurement unit) sensor, which measures acceleration, angular acceleration, and magnetic field in three axes, was chosen as the sensor for the variety of measurements. Thanks to this, we can compare up to 9 different outputs at the same time. In the end, we designed a module that has 16 IMUs. As the number of sensors increases, the resulting error decreases on average. However, weighting based on calibration errors did not prove to be the optimal solution because the sensors contain not only stochastic but also systematic errors. The module designed by us will be used mainly for further scientific research in the field of IMU sensor fusion in order to reduce the error.