Online System Identification of an Autonomous Underwater Vehicle Via In-Field Experiments

摘要: The dynamic characteristic of an autonomous underwater vehicle (AUV) is affected when it reconfigured with different payloads. It desirable to have updated model, such that the control and guidance law can be redesigned obtain better performance. Hence, we develop a method enable online identification AUV dynamics via in-field experiments, where commanded execute compact set maneuvers under doublet excitation. process has two stages. In training stage, state variable filter recursive least square (SVF-RLS) estimator used estimate unknown parameters. validation prediction capability model checked using fresh data set. parameters converged within 12 s in experiments five thrusts. Validation results show identified models are able explain 78% 92% output variation. Next, compare SVF-RLS conventional offline method. comparison shows terms accuracy, computational cost time. usefulness highlighted applications. We use turning radius at speeds, design gain-scheduled controller.