| dc.description.abstract | This master thesis dealt with the development of a fault-tolerant model-based observer for a low-cost ROV developed by BluEye Robotitcs, P2-Alpha. The observer was developed for 4 degrees of freedom (DOFs) using the installed sensor suite on P2-Alpha consisting of an Inertial Measurement Unit (IMU), a pressure sensor and a compass. In addition to this a hydro-acoustic positioning system (HPR) was added to the sensor suite by simulation. With this available sensor suite, a model-based Extended Kalman filter was used for sensor fusion towards fault-tolerance. The model parameters used in the observer was determined through experimental testing in the Marine Cybernetics Lab (MC-lab), as well as the thruster characteristics. Statistical methods for signal fault modelling with detection and diagnosis by the observer algorithms for state estimation was investigated. A signal processing module was developed for preprocessing the measurements entering the Kalman Filter, and additional Kalman Filters were designed for fault-detection and isolation in conjunction with the cumulative sum test(CUSUM). The observer was tested and verified through simulations, showing good performance and fault-tolerance towards outliers, signal freeze and high variance. Sensor bias detection using the CUSUM test was conceptually verified. Lastly the observer was tested experimentally in the MC-lab, showing satisfying results for state-estimation and a degree of fault-tolerance towards outliers, signal freeze and high variance on all sensor measurements. | |
