Study On The Optimal Techniques Of Robust Distributed Acoustic Source Localization

With the development of wireless sensor networks,acoustic source localization technology has been widely applied in intelligent environmental monitoring.Compared with a traditional centralized localization system,the distributed system usually has advantages in time,space and performance,thus it can make use of environmental information more comprehensively,and effectively expanding the perception ability of single node.Especially when the detection area is large,and environment is complex and changeable,the distributed localization system can significantly improve localization accuracy and anti-jamming ability.With the improvement of application requirements,the application environment of a distributed localization system presents diversity and complexity,which brings more and more new problems.How to improve the localization performance in complex environment attracts a lot of research attention.In order to solve the problems faced by distributed sound source localization systems,this paper studied the optimal node placement,node selection and environmental adaptability of a localization system,and then the corresponding optimization methods are proposed.These proposed methods are verified by simulations and experiments.The main work of this paper can be summarized as follows:(1)This paper deduces the CRLB(Cramer Rao Lower Bound)of localization error under the condition that the measurement performance of each node is different.On the basis of error analysis,according to the optimal criterion of minimum average CRLB,the adaptive genetic algorithm is used to find the optimal node placement according to the prior distribution of the target in the test area.Simulation results show that the optimal node placement is related to the prior probability distribution of source location,and the obtained optimal node placement has lower localization accuracy for overall detection area.(2)Different performance criteria of a distributed sound source localization system are studied,and multiple criteria are proposed as optimization function of node selection at the same time.The multi-objective optimization method is adopted to find the Pareto solutions.Finally,the multi-attribute decision making method,TOPSIS,is used to determine the selected nodes.The proposed node selection method based on multi-objective optimization takes into account different performance indicators,and the simulation results verify that localization error of the selected nodes is lower than that of the nodes with single performance as the optimization objective.(3)A distributed sound source localization method based on sound velocity correction is proposed.According to the relative angle between wind direction and sound propagation direction,the equivalent sound velocities in different directions are expressed as functions of unknown sound source location,approximating the sound velocity distribution in the wind field,and particle swarm optimization(PSO)algorithm is used to estimate the sound source location.The effectiveness of the proposed method is verified by simulation and experimental research.(4)A new weighted localization method based on node reliability probability is proposed.Mahalanobis distance is used to find outliers from a set of estimated sound source locations,thus the angles with large estimated error are identified.As a result,the reliability probability of each node is quantified.Finally,robust localization is achieved by using the reliability probability weighted method.A robust localization method based on M estimator is also proposed to suppress the effect of outliers.In this method,the relationship between estimation error and threshold value is re-evaluated during the iteration process,and different angle weights are re-assigned to nodes.Then,the iterative re-weighted least square method is used to find source location.To facilitate the use of this method,an approximate relationship between unreliable probability and optimal threshold value is established through simulations.The simulation and outdoor experiments show that the proposed robust localization methods can effectively improve the localization accuracy in the presence of outliers.The research results of this paper have significant theoretical and application value to improve the localization performance and to optimize the system resource allocation for a distributed localization system applied in complex environment.