Attitude Estimation Of ROV By Using Complementary Filters And Inertial-SLAM Algorithms

Attitude estimation has broad applications such as aerial,underwater,robotics,navigation systems,games,industrial,augmented reality systems and so on.More recently,extensive research has powerful estimators Kalman based algorithms and complementary filters which are complex an d simple,respectively.Generally,Conventional attitude or orientation estimating filters are computationally complex.This initiates the necessity for a computationally simple yet satisfactorily precise algorithm for applications.Computationally complex is of our prime interest.Thus,the growth of complementary filters(CFs)are due to the need for a relatively robust,simple and equally efficient method for several applications.First,the latest development in CFs which are applicable to low cost and low power Micro Electro mechanical Systems(MEMS)on based inertial measurement units(IMUs)employing quaternion are Fixed Gain Complementary Filter(FGCF)and Gradient Descent based Complementary Filter(GDCF).These fixed gain estimators use gyroscope a nd accelerometer triads for high and low frequency attitude estimations,respectively.The performance compar isons of GDCF and FGCF are validated using simulation and experiment by e mploying MEMS based MPU6050 IMU for different practical implementation.The performance of both filters are restricted to attitude estimation in terms of Euler pitch and roll angles since only IMU is employed without assisted sens ory system.The valuation of the filters results in identical outcome,neverth eless,FGCF is a little bit superior to GDCF partly due to the higher accuracy and partly due to the two adjustable gains.Furthermore,FGCF is less sensitive to deviation in filter gain relative to GDCF.The computational complexity of both algorithms is almost alike.Second,a comparative study of a computationally simple algorithms FGCF and GDCF together with a complex extended Kalman filter have done for att itude estimation based on MEMS IMU.Performance of the estimators is a ppraised for Euler angle estimation by making use of simulation and experimental data from MPU6050 IMU.The evaluat ion is made based on the RMSE(Root Mean Square Error)computation.Furthermore,the algorithms are mod ifiable parameters exploited for a range of values in the search for perfection.In the case of computation burden are disadvantageous,Kalman filter and its var iants are the standard for the problem of position and attitude estimation,hence FGCF and GDCF are effective approaches in such condition.The assessment of the filters outcomes the best for EKF,nevertheless,the execution time was co nsiderably larger compared to CFs.In contrast,FGCF may have a little bit sup erior than GDCF partly due to the two adjustable gains resulting in extra sele ctions.Third,the performance of FGCF,Variab le Gain Complementary Filter(VGCF)and EKF which are computationally simple,medium and complex,r espectively,which are efficient solution in many applications with cheap fixed gain.The accuracy of complementary filters coupled with MEMS IMU can be enhanced by varying filter gain with a little computation cost.Both methods can be powerfully employed in aided INS system where less computational burden is of prime interest.Fourth,GDCF employed for attitude estimation have fixed filters gain which make them impassive to the dynamic condition.In such case the system may result in erroneous estimations.Most of the complex algorithms have been employed with the cost of computational complexity but are not suitable for most applications based on simple appr oach and resources.We propose Fuzzy Tuned Complementary Filter(FTCF)to eliminate the errors with the benefit of least computational burden.The proposed algorithm is evaluated and certified in conjunction with the well-developed Kalman filter.It has been demonstrated that FTCF has considerably minimized errors in attitude estimation compared to GDCF.The results validate th at tuning of the filter gain for the dynamic condition plays a crucial role in minimizing attitude estimation errors.Furthermore,FTCF has a little computational cost but its performance is better than GDCF and equivalent to Kalman filter.At last,the Inertial-SLAM algorithm that we develop in this dissertation uses the output data of IMU and the features observed by sonar to e stimate the underwater vehicles velocity and pose,without using other position system such as GPS.For underwater vehicles,the Inertial-SLAM algorithm is a combination of INS and the SLAM algorithm.The time complexity is low for Ine rtial-SLAM compared to EKF-SLAM.The accuracy of particle filter with a small quantity of particles can reach that of EKF-SLAM,but it is faster.