| dc.description.abstract | The contribution of this disseration is threefold. First, it provides insight into the use of a certain type of nonlinear observers (NLO) for platform independent position, velocity and attitude estimation, exemplified by an unmanned aerial vehicle (UAV) and an anchor handling tug supply (AHTS) vessel. Second, it considers methods for fault detection and isolation for various sensors: Accelerometers, angular rate sensors, gyrocompasses and position reference systems. Third, it discusses application of the topics above in dynamic positioning systems. The NLOs are based on the principle of reference and measurement vectors for attitude estimation, and may or may not be aided by a separate translational motion observer in feedback interconnection. The work is presented through the findings of six separate chapters, each based on a single journal or conference article.
The first chapter considers global navigation satellite system (GNSS) and inertial navigation system (INS) integration in general, presenting methods for loose and tight integration, GNSS error models, and experimental results using a UAV. The experimental results using NLOs are compared to a real time kinematic GNSS solution, and they show unsurprisingly that a tightly coupled integration solution is superior.
The second chapter covers the use of NLO-based INS on an AHTS vessel. The first topic is attitude and heave estimation, where two distinct versions of the aforementioned NLOs are compared to each other, and to an extended Kalman filter (EKF)-based solution. Also, two MEMS-based inertial measurement units (IMUs) from different manufacturers are used in the testing: the ADIS16485 and the STIM300. The difference between the NLO versions is whether or not they are aided by position measurements. The EKF represents the classic way of attitude estimation in navigation, and the NLOs handle themselves well in comparison, with the aided NLO coming out on top. No tangible variation is found between the two IMUs for attitude estimation, while for heave estimation the ADIS16485 is preferable, owing to a better accelerometer. The second topic of the chapter is dead reckoning, where the NLO-based INS has to calculate heading and position without aiding in two separate tests. For heading the results are very promising for the STIM300, drifting only one degree in the heading estimate after one hour. For position, it is found that both MEMS IMUs cannot provide an adequate and independent position estimate performance wise, while they might well be used to increase navigation fault tolerance.
The third chapter deals with a redundant IMU configuration, presenting two alternatives for combining IMUs together for performance and fault detection and isolation (FDI). Alternative 1 is based on the classic parity space method, while alternative 2 uses quaternion-based weighting. Considering performance, they are basically equal, both improving upon the results from the previous chapter, especially for heave estimation and dead reckoning. However, in a fault case for an angular rate sensor in the IMU, alternative 2 is evidently the best solution.
The fourth and fifth chapter both investigate the same topic, namely FDI in the sensors and systems aiding theNLOs. Aselection ofNLOvariants are evaluated in an FDI context, and it is revealed that in order to successfully detect position reference faults, the attitude estimation should preferably be completely independent from the position measurements. A new attitude estimator with virtual aiding is proposed. Algorithms for FDI are also presented, where one is found in the literature and applied in a novel way to detect slow drifts. Faults in both position reference systems and gyrocompasses are sufficiently dealt with. Note, on the contrary to the previous chapters, these chapters encompass mainly simulation results.
The sixth and final chapter based on an article, attempts to bring the newfound knowledge of previous results together, to provide a novel approach of assembling a dynamic positioning (DP) system. New sensor structures are considered, and a case is made to revise today’s DP classification and testing, and move towards new ideas in system- and risk-based thinking allowing integrated INS a place in DP.
The conclusion of the thesis is that using NLOs with MEMS IMUs has great potential for navigation, in DP and otherwise. With the IMU’s small size and the NLO’s low computational demands and model-less approach, the combination of the two is truly platform independent. Using the methods described in this work may improve the performance and fault tolerence of any navigation system. | nb_NO |
