State Estimation For Underwater Robots Based On Fusion Of Multi-sensor

Centimeter level high precision operation of micro underwater robot has important application value.In order to meet the needs of high-precision control,it is necessary to obtain high-precision and low noise position,attitude and speed information,and these state information also needs to have the ability of long-term stability.At present,in the traditional state estimation method,the DR(dead Reckling)scheme which combines DVL(Doppler velocimeter)and ins(inertial navigation system)is the most mature one,but it has the problems of high cost,dead zone near the bottom of the water and excessive noise in shallow water.The existing problems of state estimation greatly restrict the intelligent development of the underwater robot,which has become the bottleneck of the development of the underwater robot industry.The introduction of visual sensor information into the state estimation is a common method of robot navigation.It is a good scheme to apply computer vision to the state estimation of underwater robot.It has the advantages of low price and high precision,but it still has the problems of low robustness and a little noise.The research focus of this paper is to provide a robust and high-precision state estimation method of underwater robot combined with vision.The main work of this paper is as follows:Firstly,the sensors used in the state estimation of underwater vehicle are introduced,including binocular camera,IMU,magnetic compass,depth meter and SBL positioning system.Then the distortion model of the camera,the monocular calibration method based on the Zhang Zhengyou calibration method,the binocular calibration and correction method,the calibration and correction methods of IMU bias,calibration factor and axial error,and the error model and calibration and correction method of the magnetic compass are studied.Secondly,in view of the problem that the measurement noise of SBL increases with the increase of slant distance,sage Husa adaptive Kalman filter is used to fuse the data of VO,SBL,magnetic compass and depth sensor,and the results of state estimation of underwater vehicle are compared with those of traditional Kalman filter.The results show that the performance of sage-Husa adaptive Kalman filter is obvious It is better than the method based on Kalman filter.Finally,the adaptive Kalman filter is optimized by the state estimation method assisted by the dynamic model.Firstly,the dynamic model of the underwater vehicle system is modeled by the experimental method,and then the real-time speed calculated by the dynamic model of the underwater vehicle is added to the measurement vector of the filter as the measurement.The simulation results of virtual environment show that the improved method greatly improves the accuracy of state estimation,which makes the system meet the requirements of high-precision operation of underwater robot.