Key Technologies Study On Inertial Integrated Navigation And Positioning For Underwater Gliders

With the development of technology for autonomous underwater vehicles (AUVs), the underwater glider as a novel and important kind of AUV is paid more and more attention. The glider plays a significant role in ocean engineering applications and it becomes the research focus currently because of its characteristics of low consumption, long endurance and small volume, et al. The accurate attitude and position information is essential for the glider to work underwater for a long time, therefore, the high-precision navigation and positioning with long endurance is one of key technologies for the research on gliders. The glide is dependent on the current and self-adjustment system to glide in the sea. The glider differs from other AUVs because it nearly doesn’t rely on any energy except the current. Therefore, the glider is important and valuable in practical applications. The low consumption limitation makes the glider’s volume smaller and simpler in design, moreover, the quantity navigation sensors are as less as possible. The Global Positioning System (GPS), as the mature technology in the land application, is not available underwater, the inertial navigation system (INS) is a better substitute for GPS due to its autonomous navigation calculation ability. INS is able to achieve the real-time the rotating angular velocity and linear velocity of the carrier through gyroscopes and accelerometers, the data measured is used to compute the attitude, velocity and position by integrating. However, the sensor errors drift with time and tend to grow unbounded, the navigation precision is decreased seriously by only using INS for long time navigation. The MEMS (Micro-Electro-Mechanical System) inertial measurement unit (IMU) is chosen considering the limitation such as cost and volume for the navigation system applied to underwater gliders, and the error and random drift are more serious especially for the MEMS inertial sensor. With the limitation of low consumption, long endurance, low cost and small quantity for low accuracy sensors, the underwater navigation with high accuracy and high reliability is key well as the difficulty of the current research. On the basis of a large number of literatures on achieving high accuracy navigation with long endurance by using low accuracy navigation sensors, this paper analyzes the model of underwater glider in detail, derives the dead reckoning (DR) model which is integrated with INS to aid and correct the output of INS. The navigation system for underwater glider is designed and the detailed error model of inertial measurement unit (IMU) is analyzed. The corresponding methods are used to correct and compensate errors for different navigation sensors. The data fusion algorithms with high accuracy and high ability are proposed for the underwater circumstance to make the navigation system and algorithms more available in the practical application. The main content and innovations of this work are as follows:(1) The underwater glider dynamic model is analyzed, and the DR is modelled to integrate with INS, and the design of navigation system for underwater gliders is completed. Considering the underwater circumstance and the characteristics of glider, the dynamic model of glider is analyzed in detail. The DR is derived and integrated with INS to aid and correct INS. The navigation system for gliders, relying on MEMS INS as the main navigation component, is designed to meet the requirements of long endurance and low consumption. It can provide high accuracy and high ability navigation and positioning information. Moreover, this system can intelligently receive the GPS or other satellite signals to update the navigation information when the glider rises to the surface.(2) The errors of MEMS inertial sensors and magnetometer are modelled, and the different methods are proposed to correct and compensate the errors of different sensors. The zero bias, scale factor and installation error are very important paremeters to accurately construct error model of MEMS IMU. The static eight-position experiment is designed, and the gyroscope and accelerometer zero biases are obtained according to multi-position symmetrical measurement theory. The scale factors and installation errors in three axes are derived through the rate experiment for gyroscopes. The dynamic experiment proposed in this paper for accelerometers can determine in real time scale factors and installation errors of accelerometers more effectively and more conveniently than the traditional static multi-position method. The error compensation is derived from the system error model and the errors of gyroscopes and accelerometers are compensated and corrected by using the relevant error parameters. The processing of the magnetic field information measured by magnetometers would fit the three dimension ellipsoidal magnetic field after denoising for original magnetic field information containing interfere information such as (soft) magnetic interference. The fitting parameters are derived to correct and compensate the outputs of magnetometers so measurement accuracy of the magnetic field is improved.(3) The extended Kalman filter and Runge-Kutta (EKF/RK4) data fusion algorithm is proposed for the practical nonlinear system and the smooth variable unscented Kalman filter (SVUKF) is proposed to further improve the accuracy in the wider applications. The glider glides at certain depth in the sea and the circumstance is relatively stable, the nonlinear model can be devided into some piecewise linear models. Based on this idea above, the model is simplied to improve the accuracy without increasing the complexity and calculation burden too much. The proposed EKF/RK4 data fusion algorithm integrates EKF with RK4 for the practical underwater navigation system. The performance of EKF is better when the system nonlinearity is not very serious, moreover, RK4 is fused. The experiment results show that the proposed EKF/RK4 data fusion algorithm can reduce errors effectively and improve the estimation accuracy of attitude and position greatly compared with the traditional EKF and UKF. In order to widen the scope of the algorithm, the SVUKF is proposed considering the better performance of UKF in the wider nonlinear models. The smooth process based on UKF achieves the better improvement of optimal estimation for the nonlinear system. It is demonstrated from the experiments that the estimate accuracy, the stability the robustness for the proposed algorithm are improved effectively.(4) The improved Gaussian mixture particle filter (IGMPF) algorithm is proposed in the more complicated system model and the performance is evaluated by experiments. For the practical problems, comprehensively considering all kinds of performance criterions and comparing with all algorithms mentioned in this paper, the back decoupling and adaptive extended Kalman filter algorithm (BD-AEKF) is proposed. There are some possibilities including non-Gaussian noises in certain moments or regions and the performance of UKF for optimal estimates may degrade. The mixture Gaussian model is built and the improved Gaussian mixture unscented Kalman filter is proposed based on the particle filter (PF). It is concluded that the attitude and position estimation accuracy for IGMPF is improved compared with other algorithms, but the real-time becomes poor and the calculation speed obviously decreases. It is not widely applied for the improvement of estimation accuracy at the expense of the lower effectiveness. Moreover, it is inevitable for the misalignment errors between installing axis and corresponding reference axis in the reference frame for inertial measurement unit. This inherent errors lead to the cross-coupling among three attitude angles (heading angle, pitch angle and roll angle) to make the attitude calculation incorrect and erroneous. This cross-coupling appears more seriously when the pitch or roll motion occurs. The pitch and roll motions are common when the glider glides in the sea and the attitude calculation errors caused by cross-coupling among attitudes appear and accumulate all the time. The BD-AEKF algorithm is proposed for this problem. The moment is judged where the calculation is wrong due to cross-coupling, and then the cross-coupling is eliminated by using backing decoupling. Subsequently, the parameters are adjusted in real time by AEKF to smooth the filtering output. The coupling is eliminated, and the estimation accuracy and output stability are improved. For practical application, one must choose a method which has the best trade-off for various properties such as estimation accuracy, algorithm stability, computational burden, numerical robustness and ease of implementation. Compared BD-AEKF with other algorithms, the accuracy is higher but the effectiveness is not low. The BD-AEKF is a high accuracy and effective algorithm in the practical application.