System Identification and Depth Control of a Tactical Diving Vehicle : Foundation for Autonomous Navigation

Sammanfattning

At JFD Sweden AB (JFD) they manufacture Tactical Diving Vehicles (TDV) made formilitary operations. Precise depth holding of TDV’s is important for special operations,yet current manual control places a high cognitive load on operators. This thesis developsand validates black box modeling of the TDV using collected data from test runs, and a con-trol framework to enable automatic depth control of a TDV. First, experimentally recordedbowplane angle, pitch angle and depth measurements were collected under controlled ma-neuvers. These measurements were then preprocessed via detrending and resampling.Discrete time ARX and state space model structures were then identified using predictionerror minimization, yielding models capable of forecasting the TDV’s depth response withapproximately 70% accuracy over a 5 s prediction horizon. Two control schemes were de-signed, a single loop PI controller and a two loop cascade controller. Both satisfied therequirements set for overshoot and hull angle. The cascade controller exhibited a fasterrise time at the cost of added complexity. Due to the requirements set by JFD on perfor-mance, the cascade controller was chosen for the implementation on a Raspberry Pi (RPI).The RPI connected to the TDV showed software timing problems and the controller hadissues with the integrator anti windup.