Research On AUV Cooperative Localization Method Based On Neural Network Processing Abnormal Measurement

Cooperative localization technology,as an important guarantee and key technology to enhance the performance of multi-platform cooperative operation,is playing an increasing role in military and civil fields such as unmanned aerial vehicles,unmanned underwater vehicles,robots and missiles.Underwater cooperative localization method has significant advantages in reducing the configuration cost of autonomous underwater vehicles(AUV)and improving the overall localization accuracy of multi-AUV formations,which has become a research hotspot.Due to the harsh underwater environment,various anomalies will inevitably occur in the sensor measurement of AUVs during long-duration missions and thus it is the technology difficulty to improve the anti-jamming capability and reliability of the underwater cooperative localization system.This dissertation discusses the impact of sensor anomaly measurement problem on the localization system and investigates a series of cooperative localization methods combined with artificial intelligence neural network technology.The main contents are summarized in five sections as follows.1.Scheme design of underwater cooperative localization system based on acoustic ranging.Firstly,cooperative localization schemes based on different configurations are introduced,as well as the underwater acoustic communication and ranging technology.Secondly,a mathematical model of the cooperative localization system is established,which is as the rationale to deal with the sensor abnormal measurement problem in a targeted manner.Finally,for the case that the compass and the doppler velocity log of AUV malfunction,a cooperative localization method based on dual-model is proposed,providing a feasible idea for improving the stability of the system.2.Researches on the cooperative localization method based on RBF neural network dealing with bias measurement error.Underwater temperature variation,communication time delay,ocean current interference and other factors tend to affect the sensor measurement accuracy,thereby reducing the accuracy of the cooperative localization.To address this problem,this paper proposes a method of using neural network compensation filter estimation.Firstly,the cooperative localization performance of different filtering algorithms and different piloting schemes is analyzed,which helps the design and selection of subsequent experimental schemes.Secondly,aiming at the problem that an algorithm relying exclusively on filters may not be able to effectively cope with bias measurement noise,this dissertation proposes a cooperative localization method based on artificial neural network-assisted filters for position estimation.Finally,the effectiveness of the proposed method was verified by trial data.3.Researches on the cooperative localization method based on Adaptive Neuro-Fuzzy Inference System(ANFIS)detecting acoustic abnormal measurement.In the acoustic range based underwater cooperative localization,acoustic communication may be affected by the harsh underwater environment.which results in measurement wild value,data loss,measurement error drift and other problems.To deal with these problems,this paper proposes a method for detecting measurement anomalies based on ANFIS.The basic implemen tation principle of ANFIS is introduced in detail,and a cooperative localization method that can detect and isolate anomalous acoustic ranging information in real time is designed based on ANFIS.Meanwhile,an online data training mechanism is designed to guarantee the reliability of this anomaly detection method under small sample data training conditions.Finally,the stability of this ANFIS-based method for dealing with acoustic ranging anomalies is verified by experim ental data,and it has a broad application prospect in underwater cooperative localization technology.4.Researches on the cooperative localization method based on ANFIS dealing with the continuous m easurement anomalies.If the acoustic measurement information is continuously abnormal,the measure of isolating the abnormal information will make the measurement data insufficient to maintain the stability of the positioning system.Aiming at this situation,a cooperative localization method based on ANFIS-assisted filter is proposed to improve the estimated accuracy of the localization system without measurement information.Furthermore,in order to avoid the influence of underwater environment on the training process of ANFIS model,an improved maximum cross-correlation entropy cubature Kalman filter algorithm is designed to optimize the training data set,which effectively enhances the stability of the cooperative localization system based on ANFIS.5.Researches on the cooperative localization method based on ANFIS-APSO-GA correcting the abnormal sensor measurement.For the problem that the traditional ANFIS using gradient descent method to adjust the affiliation parameters is easy to fall into local minima,this paper first proposes a new hybrid metaheuristic optimization algorithm,called AQPSO-GA,to optimize the ANFIS model parameters,and introduces the specific steps of this method.Then,in order to avoid the influence of isolating too much measurement information on the system localization performance,a method to predict the acoustic range error and correct the distance measurement in real time is designed by using the ANFIS model optimized by AQPSO-GA algorithm.Finally,the effectiveness and stability of the proposed cooperative localization method are verified by trial data.