Evaluating motion capture as a means of system identification of a quadcopter

This thesis describes the method of identifying unknown parameters that affect the dynamics of a given quadcopter, also known as grey box system identification. This was primarily done utilising an inertial measurement unit and a motion capture camera system. A system of equations describes the dynamics of the quadcopter, and was later coupled with data gathered while flying, in order to use different methods of system identification. As quadcopters are unstable, the first task was to design a stabilising regulator, making stable flight possible, and thus gathering flight data. A few parameters regarding motor dynamics were evaluated via simple experiments with tools including a tachometer, a scale and a microphone. When it comes to flight dynamics, the first method of identification was to use a prediction error method which, given data regarding input signals, output signals and a mathematical model, tries to evaluate unknown parameters by minimising the error between state measurements and estimated states based on the earlier mentioned model, in each timestep. This method proved to be unsuccessful, for reasons partly unknown, and was later changed for a method utilising an extended Kalman filter, which gave more reliable results. Possible explanations to this phenomena may include that the Kalman filter implemented beforehand in the camera system may need to be retuned and that the aforementioned mathematical model needs to be reevaluated. Estimated parameter values works well with the model, but that is not so say that there is not room for improvement