Resarch On Underwater Target Localization Algorithm Considering The Uncertainty Of Anchor Position

Underwater acoustic sensor networks(UASNs)play an important role in the fields of marine resource exploration,underwater environmental monitoring,and military and national defense.Accurate acquisition of sensor node position information is the foundation and key for UASNs to perform complex tasks.Most existing localization algorithms assume that the anchor position is precisely known and focus on how to overcome underwater weak communication constraints to locate target node,neglecting the impact of anchor position uncertainty on algorithm performance.To address the above issues,considering the weak communication constraints of underwater acoustic stratification and clock asynchrony,this paper studies the underwater target localization algorithm with anchor position uncertainty.The main work is as follows:Aiming at the problem of underwater target localization with uncertain position of all anchor nodes in the network originating from random topological disturbances such as ocean currents,a localization algorithm based on reinforcement learning(RL)is proposed.A disk error model is introduced to describe the influence of the anchor position uncertainty,and ray tracing method is used to overcome the localization error caused by acoustic stratification.Considering the influence of clock offset,the information transmission mechanism between nodes is designed,and the optimization problem is established based on acoustic time-of-flight(ToF)model.Furthermore,a RL-based localization algorithm is proposed to solve the problem.The Cramér-Rao lower bound(CRLB)is analyzed and the effectiveness of the algorithm is verified by simulation.Aiming at the problem of underwater target localization under the condition that the accuracy of anchor node position estimation method is limited,which leads to the uncertainty of the position of some anchor nodes in the network,and the measurement data acquisition is disturbed by unknown noise,a localization algorithm based on deep reinforcement learning(DRL)is proposed.Further considering the impact of clock skew,the asynchronous communication protocol between nodes is designed and the localization optimization problem is established.The priority experience replay(PER)mechanism is added to the framework of the soft actor-critic(SAC)algorithm to improve the learning ability and convergence speed of the algorithm,and a DRL-based localization algorithm is designed to solve the problem.While accurately locating the target,the position coordinates of the uncertain anchor are corrected to reduce the uncertainty of the network topology.In addition,in order to overcome the influence of internal covariate shift during the training process of deep neural networks(DNNs),layer normalization(LN)is adopted to improve the network structure and ensure the stability of the network output.The validity of the algorithm is verified by simulation and experiment.