Formation Control And Learning Of Multiple Unmanned Surface Vehicles Based On Deterministic Learning Method

With the rapid development of artificial intelligence,information technology,and ocean engineering,unmanned surface vehicle(USV)has attracted a great deal of attentions in the last few decades.Intelligent and unmanned USV is playing an important role in the military and civil fields.In order to overcome the limited capability of single USV,an USV formation system has attracted the increasing attention of many researchers due to their important applications such as cooperative exploration of ocean resources,distributed environmental monitoring,surveillance of territorial waters,and coordinated rescue missions.Considering uncertain dynamics presented in the USV system,this thesis employs radial basis function(RBF)neural networks(NNs)to estimate the unknown nonlinear continuous functions online.Using the deterministic learning theory,furthermore,RBF NN model for each individual vessel is shown to be capable of cooperatively identifying/learning the associated uncertain dynamics along the periodic(or quasi-periodic)state trajectory.The learned knowledge on identified uncertain dynamics can be stored in NN models with converged constant NN weights,which can be effectively exploited to develop knowledge-based control strategy without readapting to NN weight parameters.This thesis studies the formation control and learning problem for multiple USV systems based on deterministic learning.The main contents are given as follows.(1)Chapter 2 presents a design technique of finite-time adaptive NN tracking controller for a fully-actuated USV system.Firstly,we employ an asymmetric function constraint to quantify the transient behaviors of the output tracking error,and take two exponentially decaying functions as the asymmetric boundary functions.Secondly,the tracking error is guaranteed to satisfy the prescribed asymmetric error constraint by the prescribed performance control(PPC)method.Thirdly,the finite time scheme is incorporated into the control design to ensure that the tracking error can converge to a small area near zero in a finite time.Finally,a model-based tracking controller is constructed by combining Lyapunov stability theory with backstepping design procedure.Besides the nominal case where exact knowledge of the vessel model is available,we also deal with scenarios wherein modelling uncertainties are present.Subsequently,we employ NN approximators to estimate uncertain dynamics and design an adaptive NN controller that achieves system stability with prescribed performance.Simulation studies are performed to demonstrate the effectiveness of the proposed design technique.(2)Chapter 3 studies the cooperative learning formation control problem for a group of USVs with prescribed performance under modeling uncertainties and time-varying disturbances.Firstly,the formation tracking error is converted by PPC technology and practical finite-time stable(PFTS)is employed to ensure convergence of the tracking error within a finite time.To mitigate the effects of external time-varying disturbances,disturbance observers are incorporated into the formation control design.The modeling uncertainties including hydrodynamic damping terms and unmodeled dynamics are identified/learned by RBF NNs in a cooperative way.Consequently,a novel cooperative NN-based formation learning controller is proposed by establishing communication topology among NN weight update laws for sharing the online learning knowledge.The proposed formation controller is shown to be capable not only of fulfilling the predefined formation pattern with guaranteed prescribed performance,but also of identifying/learning the associated uncertain dynamics using the cooperative deterministic learning theory.Moreover,the learned knowledge on identified uncertain dynamics can be stored in NN models with converged constant NN weights.Using the stored knowledge,an experience-based formation controller is developed to improve the control performance and reduce the computational burden,meanwhile still preserving the prescribed formation tracking control performance.Simulation results show the effectiveness of the proposed formation controllers.