Position tracking of underwater vehicle using extended Kalman Filter

Abstract

Position tracking is essential for mobile robots for autonomous functionalities and navigation especially for robots that are deployed in underwater conditions. Hence, this thesis proposes the usage of the Extended Kalman Filter (EKF) for position tracking of an underwater vehicle. Underwater vehicles cannot use conventional GPS for position tracking due to radio signals being damped by the body of water surrounding it. Underwater GPS(UGPS) is used for predicting the position of underwater vehicles, but it suffers from latency issues. Therefore, estimation algorithms like Kalman Filter (KF) and EKF are applied to provide a consistent position value from the UGPS. The main advantage of the EKF estimation algorithm is it can estimate the state of a non-linear system without an observable model. It is a nonlinear extension of KF, and it is a popular method used in estimating robot position due to its simplicity and consistency. The main objective of this research is to implement EKF in underwater conditions using UGPS relative position and Inertial Measurement Unit (IMU) orientation. The secondary objective of this research is improving EKF positioning estimation by implementing of outlier filters. Overall, the proposed system allows accurate position tracking of underwater vehicles. Before EKF is applied, the dead reckoning model of the ROV was developed as the vehicle odometry. In addition, an experiment is conducted by evaluating the odometry of the robot where the transmitter of the UGPS is attached to the Remotely Operated Vehicle (ROV) and need to travel a pre-measured distance and compare the odometry output of the ROV with the measured distance. To test the effectiveness of the proposed method, the EKF was implemented offline with recorded data consisting of Underwater GPS (UGPS) and Inertial Measurement Unit (IMU). The filtered EKF output is evaluated by using MSE and RMSE to ensure the distinct features of the output signals are retained. The MSE and RMSE of median mean filter are less than 0.1 meter which signifies the filtered output of EKF retains the distinct features of the raw output of EKF. The proposed method can overcome the UGPS latency issues and accurately estimate the underwater vehicle’s pose.