Realtime Offset Estimation of Inertial Sensors

OData support
Kovács Viktor
Department of Automation and Applied Informatics

In our dangerous world there is a great emphasis on security in transportation. Motor vehicles are full of sensors and actuators, which protect our life. Particularly at motorcycles, leaning angle is a key feature. If this angle proceeds a limit angle, the electronic control unit sends signals the brakes to release. The leaning angle cannot be measured directly, we have to derive it from inertial acceleration and angular speed sensors, based on mechanical equations. But unfortunately our sensor measures are charged with offset errors. It means that at 0 input the output is a relatively small non-zero value. This can make our estimation difficult.

In this thesis, at first I am going to create a mechanical model of a rolling wheel in Matlab Simulink. It is a good example to understand the behavior of rolling wheels. Secondly I estimate the leaning angle of a motorcycle using equations and sensor signals. There are several methods which seems to be working, I try to choose the best one. We need the leaning angle only for helping the offset estimation at next point. There are also numerous methods for offset estimation. Some of them are based on my individual ideas. I am going to take tests with them and choose the best one. In the next point I list some opportunities to develop the estimations, for example compile to C code. In the last point I write about most common types of inertial sensors, like capacitive, piezo-electric and MEMS sensors.

The result of experimentation with offset estimation is good. We have found great methods.


Please sign in to download the files of this thesis.