Motion Model Parameters Identification Of USV Based On Muti-innovation Theory

USV(Unmanned Surface Vessel),as a development direction of automatic and intelligent ship,is widespread concerned by scholars.USV needs to have good maneuvering characteristics to achieve unmanned operation under complex navigation conditions.The maneuverability forecasting of USV is an effective method for evaluating the maneuvering characteristics and can provide a reasonable model reference for an adaptive control of USV.Since the accuracy of the USV motion model is very important both in the maneuverability forecasting and motion control of USV,determination of the model parameters becomes the key issue.Nowadays,the development of modern control theory makes parameter identification become an important method for model research.At present,the method of identification is widely divided into two categories: the least squares method and Kalman filter based classical identification methods and the neural network and support vector machine based statistical identification methods.In ship motion model parameter identification field,the application of classical identification methods is more mature.In this paper,Abkowiz hydrodynamic model and response model of USV are used for model parameter identification based on multi-innovation method.The classical methods including the least squares method,the kalman filter method and the least squares support vector machine are used as a contrast.The main work of this paper is as follows:1)With application research of multi-innovation least squares method,multiinnovation calman filtering and least squares support vector machine,parameter identification methods of USV motion model based on these three methods are put forward.2)With improve research of multi-innovation calman filtering by adding a forgetting factor,an improved algorithm and the convergenceis proof are put forward;With improve research of least squares support vector machine by combination with the multi-innovation theory,a multi innovation least squares support vector machine is put forward for identification.3)The data acquisition experimental platform and parameter identification simulation platform of USV model ship were constructed.The Z experimental data of USV model ship is used to carry out parameter identification test and maneuverability forecasting,and the comparisons of multi-innovation method and traditional identification method are carried out from two aspects of identification accuracy and parameter convergence speed.Finally,the experimental results show that the multi-innovation methods and the improved methods are more accurate and identification converges are faster compared with the classical identification methods,which can provide reference for the online parameters identification method of USV.