Master thesis proposals

System identification for a torpedo-shaped, under-actuated autonomous underwater vehicle

Start : January
Contact : Ignacio Torroba (torroba@kth.se), David Dörner (ddorner@kth.se)
Keywords : System identification, AUV, ML

SAM is a hydrobatic AUV capable of complex maneuvers in the water column despite its shape and limited actuation, thanks to a set of internal actuators that allow it to change its buoyancy and center of rotation. For the design of robust and high-performing control policies for such a vehicle we need a highly-accurate dynamics model. However, some of its hydrodynamic coefficients cannot be analytically parameterized accurately. Instead, we aim to identify them through physics informed machine learning techniques. To collect the necessary datasets an assess the accuracy of the regressed models, we rely on a newly constructed instrumented water tank at KTH with an underwater motion capture system.