Research On UUV Underwater Maneuvering Target Tracking Considering Model Error And Short-term Detection Failure

Ensuring maritime security is an important cornerstone of the development of the national economy.In recent years,underwater escort using Unmanned Underwater Vehicle(UUV)as a carrier has become a major trend,and the ability to track underwater maneuverable targets will affect the key performance of the entire system.Underwater maneuvering target tracking is different from air maneuvering target tracking,its motion form is complex and changeable,and the measurement information is detected by sonar sensors.However,the effectiveness,range and accuracy of sonar sensor detection are very limited under the influence of complex marine environment.Therefore,it is of great significance to study the UUV underwater maneuvering target tracking method based on the sonar detection information.The main purpose of this paper is to solve the two major reasons that affect the tracking accuracy of underwater maneuvering targets,namely model uncertainty and short-term sensor failure.The specific research is as follows:(1)This article studies the CV model,CA model,CT model,Singer model and CS model,and gradually discusses the Kalman Filter(KF)algorithm,the Extended Kalman Filter(EKF)algorithm and the Unscented Kalman Filter(UKF)algorithm.Finally,we analyze the advantages and disadvantages of each target tracking algorithm.The above research lays the foundation for the follow-up single-model tracking algorithm and multi-model tracking algorithm;(2)This paper designs a single model tracking algorithm in a three-dimensional coordinate system for underwater non-maneuvering,low-speed maneuvering and high-speed maneuvering targets,and compares the tracking performance of the single-model algorithm based on the above five different motion models;(3)Aiming at the single model algorithm that is prone to produce larger model errors when the target is maneuvering,this paper studies the Interacting Multiple Model(IMM)target tracking algorithm.Due to the nonlinearity of the underwater target tracking system,this paper further designs the IMM-EKF and IMM-UKF algorithms for underwater maneuvering target tracking,and completes the simulation verification.In view of the inaccuracy of model switching in IMM,which leads to large model errors,this paper starts with model parameter adaptation and noise covariance matrix adaptation,and further analyzes the reasons affecting target tracking performance,and designs UKF based on dual adaptive CS model Target tracking algorithm,and complete simulation verification experiment;(4)Aiming at the short-term failure of the sensor,this paper starts with the Open-Loop Kalman Filter(OLKF)algorithm,and designs the Open-Loop Unscented Kalman Filter(OLUKF)algorithm and complete the simulation experiment under the short-term failure of the sensor.Secondly,this paper uses the Long Short-Term Memory(LSTM)neural network with non-linear time series prediction ability to design an LSTM neural network structure suitable for tracking underwater maneuvering targets under short-term sensor failure.Thirdly,in order to verify the prediction effect under asynchronous length,this paper carried out a non-linear data compensation simulation experiment based on historical measurement information under asynchronous length.Subsequently,in view of the spike error caused by the open-loop unscented Kalman filter,this paper designs the UKF algorithm based on LSTM neural network online compensation,and completes the simulation verification work.Finally,comprehensively considering the problems of model errors and sensor short-term detection failure in the UUV underwater maneuvering target tracking system,this paper further designs the UKF algorithm of the dual adaptive CS model based on the LSTM neural network online compensation,and passes the simulation.