Research On Localization Method For A Special ROV Using Multi-sensor Data Fusion

This dissertation aims at designing a multi-ranging sonar self-localization system for a special Remotely Operated Vehicle(ROV) which can be used in the towing tank. Meanwhile,in order to improve the localization precision and reduce the uncertainty caused by the measurement noise and other factors,this paper applies the estimation filtering method,which is a part of the multi-sensor data fusion theory,in processing the data from the sensor system.Under most circumstances,the process model and the observation model of actual systems are nonlinear.And Extended Kalman Filter(EKF) is a common method to solve this problem;however,it would induce errors during the linearization process and lead to bad impact on the final estimation result.Therefore, besides EKF,another nonlinear estimation method named Unscented Kalman Filter(UKF) is also studied which can directly use the nonlinear model,avoiding linearization error and involving comparable level of computation complexity as EKF.Firstly,this paper analyzes the underwater vehicle model in terms of kinematics and dynamics and then establishes the special ROV's simulation model on the basis of the special ROV's structure and hydrodynamic parameters and by using the vector modeling method.Further test indicates that this model is necessary for later study of the localization method.Secondly,by comparison and analysis of the land robots' and underwater robots' commonly used localization methods and considering the unique working environment of the special ROV,a multi-ranging sonar based localization method is proposed.Since the special ROV is still at the designing stage,a reverse algorithm of the localization method is introduced to simulate the sonar data so as to verify the viability of the localization method.Thirdly,considering the sensor measurement noise and the targets of the special ROV's localization system to estimate the position and heading,much attention is given to the KF,EKF and UKF principals and algorithms.Then,combining the special ROV's kinematics equations,a comparison is drawn between EKF and UKF in the aspect of the nonlinear transformation's impact on the state variable estimate.Finally,based on the special ROV's kinematics equations and sensor system,the filter model is established and the simulated sensor data are used to test the EKF and UKF based localization algorithms.The result demonstrates that,by applying this method,the localization requirements could be met and a high precision is attained.