基于ANFIS的多AUV协同定位系统量测异常检测方法  

ANFIS-based Measurement Information Anomaly Detection Method for Multi-AUV Cooperative Localization System

作　　者：徐博[1] 李盛新 王连钊 王权达 XU Bo;LI Sheng-Xin;WANG Lian-Zhao;WANG Quan-Da(College of Intelligent Systems Science and Engineering,Harbin Engineering University,Harbin 150001)

机构地区：[1]哈尔滨工程大学智能科学与工程学院,哈尔滨150001

出　　处：《自动化学报》2023年第9期1951-1966,共16页Acta Automatica Sinica

基　　金：国家自然科学基金(61203225,61633008);黑龙江省自然科学基金(F2018009);黑龙江省博士后科研启动基金(LBH-Q15032);装备预研重点实验室基金(614221801050717);国家海洋工程重点实验室公开项目(1616);国家装备预研基金(61403110306)资助。

摘　　要：针对异常水声测距信息对多自主水下航行器(Autonomous underwater vehicles,AUV)协同定位系统的不利影响,以及传统故障检测方法在多水声测距信息交替混淆的情况下检测效率低的问题,提出一种基于自适应神经模糊推理系统(Adaptive neuro-fuzzy inference system,ANFIS)的量测异常检测方法.首先,分别建立与各水声测距系统相对应的ANFIS模型;然后,基于自适应容积卡尔曼滤波(Adaptive cubature Kalman filter,ACKF)和马氏距离构造反映量测异常的特征信息作为ANFIS的输入;其次,基于预定义的量测异常信息建立了初始混合数据库以训练ANFIS模型实现对量测异常的在线实时检测与隔离;最后,利用湖水实验数据进行了AUV协同定位仿真验证.实验结果表明该方法可以准确识别异常水声测距信息,与传统故障检测方法相比,误报率(False positive rate,FPR)与漏检率(False negative rate,FNR)均减少70%以上.In this paper,a measurement anomaly detection method based on adaptive neuro-fuzzy inference system(ANFIS)is proposed to address the problem that abnormal underwater acoustic ranging information has adverse effects on the multi autonomous underwater vehicles(AUV)cooperative localization system and the traditional fault detection methods have low detection efficiency when the multi acoustic ranging information is alternately confused.Firstly,the ANFIS model corresponding to each underwater acoustic ranging system is established.Secondly,the characteristic information reflecting the measurement anomaly,which is based on adaptive cubature Kalman filter(ACKF)and Mahalanobis distance,is used as the input of ANFIS.Then we established an initial hybrid database of pre-defined abnormal measurement information to train ANFIS model to realize online real-time detection and isolation of measurement anomalies.Finally,the lake test data are used to verify the AUV cooperative localization simulation.The experimental results show that this method can accurately identify the abnormal situation of measurement information,as the false positive rate(FPR)and the false negative rate(FNR)are reduced by more than 70%compared with the traditional fault detection method.

关 键 词：自主水下航行器 协同定位 自适应神经模糊推理系统 水声测距 量测异常 

分 类 号：U674.941[交通运输工程—船舶及航道工程]