In order to increase accuracy and robustness of their state and pose estimations, autonomous underwater vehicules (AUV) use a growing number of sensors. All these sensors must be calibrated before being used. This calibration step is essential in order to receive the most reliable data possible: Without a good calibration, one AUV can get untrustworthy data and, as worst-case consequences, make decision that leads to a fatal collision or to a loss of the robot.
Currently, several AUVs at the DFKI RIC use hard-coded poses or tuning parameters for sensors. These parameters can be taken, for example, from the sensor manufacturer or from the CAD model. This means firstly that one cannot be sure that these poses/parameters are accurate and on the other hand, these values must be manually adapted whenever there are changes in the AUV-hardware. On the top of that, there is currently no framework at the DFKI that automatically calibrates sensors AUV-independently. This leads to the fact that some calibration functions might be reimplemented from scratch, even if there are already existing functions written as part of other projects.
In this thesis, a framework for underwater robot calibration is presented as well as two calibration methods. The first calibration allows one to find the pose of multiple sensors in an AUV by doing a predefined set of movements with the robot. The second presented method is an optimization for an underwater motion model parameters, which can be used, for example, to predict the acceleration made by an AUV, given the forces/torques of its thrusters.