Robot Localization Method Research Based On Ultrasonic Phased Array

Then,the feature points obtained by the ultrasonic array are sorted out,and a point cloud matching algorithm based on FFT(Fast Fourier Transform)is proposed.The distance from each point to the gravity point is calculated by using the characteristics of point cloud,and the distance is normalized and sorted quickly.The fast sorted sequence is transformed into amplitude information by FF With the development of science and technology,modern robots have been widely used in mechanical manufacturing,military,scientific research and so on.The localization of robots in 3D(three-dimensional)scenes is the current research hotspot and difficulty,which has attracted the attention of researchers.At present,the common sensors can not effective achieve of scene information when they are used in special scenarios such as radiation and darkness.The ultrasonic sensor will not be affected in the above special environment,and the cost is low and the manufacture is simple.Therefore,this paper researched the robot localization method based on ultrasonic phased array.Firstly,this paper completes the 3D modeling and simulation of target scenes.Proe(PTC Creo Elements)software is used to draw 3D scene,and different resolution object models are given according to the distances between the ultrasonic array and the target object.The ultrasonic propagation path in 3D scene is determined by using the ray-tracking method and the law of ultrasonic propagation,and the sound pressure of sound rays is calculated according to the calibration of sound pressure.CUDA(Compute Unified Device Architecture)is used to accelerate the ray-tracking method to improve the speed of sound field simulation and realize the acquisition of ultrasonic echo signals.Secondly,this paper completes the DOA(Direction of arrival)estimation of feature points and the extraction of 3D coordinates.The DOA estimation of reduced-dimension Capon algorithm and Unitary-ESPRIT algorithm is compared and analyzed in terms of angle resolution and time consumption.The results show that the resolution and time-consuming of Unitary-ESPRIT algorithm are better than that of dimension-reduced Capon algorithm.At the same time,the distance between the feature points and the array is calculated by the TOF(Time of flight)method,thus the 3D coordinate information of the feature points is obtained.T transformation,then the effective amplitude information is intercepted.The matching of sparse point clouds is realized by calculating the similarity of the amplitudes of two groups of sparse point clouds.Compared with ICP(Iterative Closest Point)algorithm,FFT-based point cloud matching algorithm is faster and more efficient.Finally,this paper uses scene information matching to complete robot localization.By analyzing the resolvable distance of the ultrasonic array,the appropriate distance and rotation angle for the robot to collect scene information are determined.The 3D scene information base is constructed using the amplitude information of the robot in different positions.The amplitude information obtained by the robot is matched with the information in the information base to complete the robot localization.Through simulation and analysis,when the SNR is less than 20 dB,the localization result of the robot is inaccurate due to the interference of noise.When the SNR is 20 dB,the average error of the proposed algorithm is 0.13 meters in the x-axis and y-axis directions,and the average error of the angle is 11.7 degrees.The positioning error and the calculation speed of the proposed algorithm are better than those of the ICP algorithm.