Usage of Inertial Measurement Units (IMUs) has become more and more widespread in various areas, because it is applicability to motion characteristics determination. There are applications like navigation systems, games, virtual and augmented reality which utilize IMUs. Further usage of them may increase in user interface developments, image stabilization, and medical or sport related movement analysis.
The measured data of IMUs need to be converted to formats, which are useful for the actual application. This conversation utilizes complex data processing operations like data fusion, whose implementability is limited. Therefore the implementation of them is not an easy task. Data fusion can be splitted into parts and these parts can be realized with different conditions and for different interfaces. In a wireless sensor network there are several aspect like the executive units computing capacity, bandwidth of communication and the consumption of sensor nodes which have to take into consideration.
In my work I introduce the behaviour and error sources of the sensors in the IMUs. I introduce the properties of sensor fusion from the basis and three sensor fusion algorithms. After that I choose one type of IMU (MPU-6050) and attach it to a previously developed sensor network. I compare the three algorithms based on the characteristics of the system, e.g. computing capacity, consumption and the usage of communication bandwidth. Finally, I implement some of them and examine the quality of the orientation data gained from measurements. I have already presented one part of my work in my TDK thesis.