Research And Application Of DOA Eatimation Algorithms Based On Ultrasound Sensor Array

Different from traditional fixed robots,mobile robots can realize autonomous perception,planning and operation through artificial intelligence and other technologies under unknown environment.Therefore,mobile robots must have positioning function,and the specific positioning mode is directly related to the type of sensors used by the robot.At present,monocular and binocular video sensor units are the hot spot in the research of visual navigation and positioning for mobile robots.They are widely used in visual navigation and positioning because they have the advantages of low cost of sensor acquisition system and can simulate human vision.However,the visual navigation and positioning technology based on video sensor unit has some limitations,such as the inability to recognize transparent objects,which seriously affects the positioning and recognition effect of subsequent processing systems.Therefore,how to effectively remove such limitations in the application of vision sensor system in mobile robots has become the first starting points of this paper.In order to solve the above problems,this paper uses the ultrasonic sensor array to solve the problem of depth data blind area in the video sensor unit acquisition image.Although ultrasonic array has the ability of active positioningand recognition of targets,it has been widely used in underwater target detection and positioning,vehicle obstacle avoidance and so on.But at the same time of forming high-precision array,the complexity of data acquisition increases,and the cost of array becomes very expensive,which greatly limits its practicability.Therefore,how to achieve a certain balance between technology and application is the second starting point of this paper.1.A two-dimensional DOA(Direction Of Arrival)estimation algorithm for L array is proposed.Its core idea is to transform the output matrix of the array into a centrosymmetric matrix by matrix reconstruction,and then it is transformed from a complex matrix to a real matrix by unitary transformation matrix.This method can directly get the target parameters without conversion,which greatly reduces the amount of operation.Compared with the matrix enhancement and matrix pencil(MEMP)algorithm for L array,this algorithm can estimate more targets' DOA and obtain higher resolution target images.2.A joint distance-angle estimation algorithm(JD-AE)is proposed.An arc array structure is designed and the flow pattern of the array is deduced.The multiple signal classification algorithm is used to simulate one-dimensional and two-dimensional DOA estimation,and a joint distance-angle estimation algorithm is proposed based on the principle of ultrasonic sensor ranging.The algorithm can simultaneously estimate azimuth and source distance.The traditional multiple signal classification algorithm can get better estimation results only under the condition of high SNR.In this paper,when the outputcovariance matrix of the array is denoised,the joint estimation of the azimuth and distance of the signal can also be obtained with high precision under the low SNR.When the SNR is [-5dB-40dB],the probability of success of JD-AE algorithm is always higher than that of traditional multiple signal classification algorithm,which effectively improves the estimation ability of source parameters.3.Aiming at the practical application of mobile machine platform,this paper studies the scheme of sparse arc-shaped ultrasound sensor array to compensate the blind area of visual image data.In order to verify the effectiveness of this method,an experimental platform of ultrasonic sensor array is built in combination with the laboratory environment.Finally,through a number of experimental tests,the corresponding processing and analysis of the obstacle data results show that when the obstacle distance is less than 80 CM,and the azimuth angle is greater than 30 degrees and less than 150 degrees,the error of the estimated results is less than 10%,which is more accurate,thus verifying the effectiveness and superiority of the algorithm in this paper.