| Summary: | The Seahog, which is an underwater Remotely-Operated Vehicle (ROV) was initiated to meet objectives set in conjunction with the marine biology group at the University of Cape Town for observation and light sample material collection from shallow sea floors. Prior to the commencement of this project, the Seahog was in a state where the basic functional units such as the thrusters, cameras, lights and power and communication systems were tested and installed. A crucial unit was however missing: the localization sensing unit. Localization comprises of linear and angular position sensing. The attitude of an underwater vehicle is its angular position relative to an inertial frame. Using the standard format for underwater vehicles, these angles are typically known as roll (about an axis along the surge direction of the vehicle), pitch (about an axis along the sway direction of the vehicle) and yaw (about an axis along the heave direction of the vehicle). Attitude sensing can be achieved using inertial sensors. Due to the shortcoming associated with the gyroscope’s measurements in the long term and noise sensitivities associated with the accelerometer and magnetometer, the optimal attitude estimation of the Seahog required sensor fusion. The Seahog’s Inertial Measurement Unit (IMU) is a 9 degree of freedom (DOF) unit fitted with an embedded processor collectively referred to as the iNEMO. Due to the time-variant nature of the magnetic field around the Seahog owing to variations in the magnetic coupling-based thrusters, the magnetometer was excluded from usage in the sensor fusion. The performances of several sensor fusion filters for estimating attitude were observed through simulation results. Steady state conditions were assumed for all but one test scenario. This was deemed sufficient as the Seahog is a slow moving vehicle and an assumption of mostly steady state was applicable. The filters were the Extended Kalman Filter (EKF), the standard Unscented Kalman Filter (UKF), Spherical Simplex Unscented Kalman Filter (SS-UKF) and a quaternionbased Complementary Filter (CF). The basis of performance assessment was the Root Mean Square Error (RMSE) from each filter’s simulation result. In almost all test scenarios, the UKF produced the least RMSE, i.e. the most accurate attitude estimates over the test period, though the results were not significantly better than the EKF’s. Under non-accelerating conditions, the filters were found to estimate the tilt angles relatively accurately but drift was present in the yaw estimation especially when no rotation occurred about the yaw axis. Ultimately, the EKF was selected as the attitude filter for implementation on the Seahog as its performance was comparable to the UKF’s at the tested sampling frequency and its computational cost was significantly less than that of the UKF. The EKF was implemented on the iNEMO’s embedded processor and put through some tests to observe its performance. For yaw angle estimation the filter ultimately yielded a minimized drift rate of about 0.87 °/hr which was deemed sufficient. The filter performed exceedingly well for the roll and pitch angle estimation, providing accuracies well within the typical tilt angle (roll and pitch) accuracy of 1 ° [1]. Ultimately, under steady state conditions, while the iNEMO provided accurate estimates for the tilt angles, a gyrocompass was suggested as the best solution for yaw estimation as it is a compass unaffected by magnetic field variations. When a vehicle is undergoing an unsteady motion underwater, extra resistance is experienced by the vehicle. This resistance differs to the drag forces as it is independent of the vehicle’s relative velocity. This resistive force is known as the added mass. Added mass covers both the translational mass and the rotational moment of inertia. This can be viewed as the volume of fluid that a vehicle has to move aside while in motion. A Simulink-based simulation model of the Seahog had been previously developed. Its hydrodynamic models were based on results from SolidWorks’ simulations and existing empirical formulation [2]. The added mass values implemented in the simulation were estimated using available empirical data, though for the formulas to be applicable, an oversimplification of the Seahog’s geometry was performed. For this reason, a different approach to added mass estimation was considered. Added mass are typically estimated computationally using software programs like WAMIT or experimentally through captive measurement at a tow tank facility fitted with Planar-Motion Mechanism (PMM) equipment. There was no access to such a facility nor a budget to purchase a CFD package licence that could serve this purpose (these are usually expensive, where a typical licence could range in the US$ 1000s). Therefore another approach had to be considered for estimating the added masses. A simple free-decaying pendulumbased experimental procedure was proposed by Chin et al. [3]. This approach was firstly verified by comparing the experimental test results to the analytical results of simple cuboids. Then the experiment was performed on a scaled version of the Seahog called the mini-ROV. Least square regression was employed in the MATLABbased algorithm used in processing the angular position data, which ultimately outputs the translational added mass. Finally, a tilt unit controller for setting the angular position of the front camera and light tray was designed, tested and installed on the Seahog. Its motor-to-tray torque transmission was magnetic coupling-based as this solution offered a safer and more durable option when compared with the dynamic sealing solution.
|
|---|
