In this project the possibility of using a network of magnetometers sensing a permanent magnet for tracking has been investigated. Both the orientation and the position of the magnet have been considered. A dipole approximation of the magnetic field is used to develop two models. One of the models parametrizes the orientation with the magnetic moment vector, while the other parametrizes the orientation with a unit quaternion. An extended Kalman filter have been used to estimate position and orientation.
Several calibration algorithms have been developed to calibrate for sensor errors, differences in sensor coordinate frame orientations and also for the estimation of the magnetic moment norm of a permanent magnet. The models have been tested using an optical reference system for position and orientation estimation. Initial results are ambiguous and further testing is necessary. One conclusion is that the model using the magnetic moment vector as orientation parametrization is less sensitive to the accuracy of the initial guesses of the filter recursions and also less sensitive to possible model errors.
A mathematical result of the possibility of using a non stationary sensor network to track the magnet is also given.