Motion Control Of Unmanned Surface Vehicles Based On Model Predictive Control

Unmanned surface vehicles(USV),which are navigated without persons onboard,are obtaining more and more attention.Autonomous navigation is a key chanllenge for USV application.Considering that a motion controller with the ability of precise tracking control is the basis of realizing autonomous navigation,the two patterns of motion control,i.e.,trajectory tracking and path following,are focused and discussed in this thesis.Model predictive control(MPC)combined with line-of-sight(LOS),least squares support vector machines(LS-SVM),extended state observer(ESO)and other methods is used for USV motion control studies to improve the precision and reliabilities for the motion control when there exist environmental disturbances,model parameter change,unmeasurable states and disturbances.A USV motion control experimental platform is built based on a model ship to verify the effectiveness of proposed methods.The main works and contributions in this thesis are summarized as follows:(1)For the USV trajectory tracking control under uncertainties,a linear MPC(LMPC)and a nonlinear MPC(NMPC)method based on dynamics models are proposed.The modelling,model linearization,model discretization,constraints setting,objective function designing are elaborated specifically.The trajectory tracking control experiments with LMPC and NMPC are realized on the Matlab simulation platform,respectively.The simulation experiment results show that the two methods both track the trajectory well even under disturbances.However,LMPC has higher computational efficiency than NMPC because the optimization problem for LMPC is a quadratic programming problem.In contrast,NMPC has higher tracking precision because there is no linearization process to decrease the model accuracy.In comparison with the sliding-mode control(SMC)based method,the MPC method has better performance of trajectory tracking.(2)For the USV path following control,a path following method based on a 2~ndd order Nomoto model is proposed to avoid the difficulties of parameter identification for the dynamic models.Firstly,an adaptive LOS navigation algorithm is proposed to improve the tracking precision by setting the different acceptance circle automatically according to the angle between two adjacent path segments.Secondly,a path following control method is proposed based on the adaptive LOS navigation algorithm and the MPC method.Lastly,the path following performance by the use of the proposed method is realized on the simulation platform under both disturbances and no disturbances.The results show that the MPC path following control method combined with the adaptive LOS navigation algorithm has higher tracking precison in comparison with the MPC method combined with the traditional LOS navigation algorithm,and it performs well with the nonzero current speed and disturbances.(3)For the problems that the USV path following performance may turn worse when the model parameters change because of aging,draught change,etc.,aλ-LS-SVM based MPC adaptive control method for path following is proposed.Theλ-LS-SVM based on the weighted LS-SVM is proposed to use for the model online parameter identification.To be specific,an index is designed to detect the possible model changes,which speeds up the rate of parameter convergence;a sliding data window strategy is used to realize the online identification;a persistent input excitation scheme is introduced.The effectiveness of online parameter identification withλ-LS-SVM is verified by setting different simulation scenarios.The proposed adaptive control method can track the path well when the model parameters change during the path following.(4)Considering that some parameters or disturbances cannot be measured directly,an MPC method based on compensation extended state observer(CESO)is proposed.The CESO unifies nonlinear terms of the system model and disturbances as lumped disturbances,and makes the nonlinear model change to a linear model with the lumped disturbances compensation.The stability of CESO is analyzed in theory.For the MPC,as previously mentioned,a linear model can make the computational efficiency higner.Two cases,i.e.,numerical calculation case and USV heading tracking case,have been used to verify the effectiveness of the control and state estimation performance.(5)A novel path following control method,namely LEM(LOS-ESO-MPC),based on the adaptive LOS navigation algorithm,the CESO,the MPC for USV is proposed.The simulation results show that a precision path following control performance can be achieved when there exist external disturbances and unknown or unmeasurable states by the LEM method.The LEM method has higher tracking precision in comparison with PID(proportional–integral–derivative)method.(6)A USV motion control platform based on a model ship is built to verify the effectiveness of main proposed methods in this thesis.To solve the locating and heading sensing problems of the model ship for the indoor environment,a monocular vison based locating algorithm is designed.The course and speed keeping stabilities for the model ship are tested.Based on this platform,the MPC methods for course tracking and path following,and the adaptive LOS method and the LEM method are all tested.The results show that the motion related methods proposed in this thesis perform well in real environment by comparing with other methods.