| dc.description.abstract | After two decades of dedicated research and development, autonomous underwater vehicles (AUVs) are today becoming accepted by an increasing number of users in both military and civilian institutions. Despite recent progress, underwater navigation remains a substantial challenge to all submersibles. The actual autonomy of the vehicles in existence today is also limited in many ways. Advances in both navigation and autonomy will enable new operations which earlier have been considered infeasible, or at best difficult. Examples of emerging applications include fully autonomous naval operations, polar deployments and under ice surveys, pipeline inspection, and oceanographic research in the mid-water zone.
This thesis is primarily concerned with inertial navigation of AUVs and the development of unconventional aiding tools in order to improve the performance and sustainability of the integrated navigation system. The investigated topics are all related to the same main objective – to enhance the level of navigation autonomy possessed by an AUV. A vehicle possessing a high level of navigation autonomy will be able to navigate for considerable time without operator supervision or intervention, and possibly with sparse external positioning and subject to velocity sensor failures or dropouts. Three problems arise in this thesis.
The first and main problem involves the employment of a kinetic vehicle model in order to provide velocity aiding to the inertial navigation system (INS). Velocity aiding is crucial for limiting the INS error drift in between position measurement updates. The performance of the model-aided INS is evaluated on experimental data from a field-deployed AUV. In combination with embedded sea current estimation, model aiding is demonstrated to be an effective approach and contribution toward solving the above mentioned challenges, including improved mid-water zone navigation, environmental estimation, and increased level of navigation autonomy. The methodology does not require any additional instrumentation, and since it is merely an addition of software, the time between failure is extensive. This thesis is the first report in the attainable literature on the experimental evaluation of a complete model-aided INS, and its practical application to underwater vehicle navigation
The kinetic vehicle model utilized for aiding the INS is a typical grey-box model where the vehicle motion is described by a set of parameterized ODEs. Prior to carrying out the parameter identification it was necessary to estimate and compensate for ocean currents. For this purpose, a novel framework has been developed which may be used for obtaining steady-state maneuvering characteristics of a wide class of underwater vehicles. The framework may be of broader interest because of its applicability to surface vessels. As a standard step in parameter identification, the final model was cross-validated against independent data. The model development in this thesis is one of the few published reports which experimentally validate and compare kinetic vehicle models for AUVs.
The second problem considered in this thesis is the development of a Doppler velocity log (DVL) water-track aided INS. As for earlier work the literature is scarce, and most of the available reports involve pure black-box systems. This thesis gives, to the best of the author’s knowledge, the first in-depth discussion, derivation, and experimental evaluation of DVL water-track aided INS. As when applying model aiding, the sea current estimation is done in real-time. Besides being used for navigation, a current estimate may be of interest for applications such as oceanography and marine research, and autonomous mission planning and decision making. The DVL water-track aided INS is evaluated on experimental data from a field-deployed AUV. Both horizontal and vertical (i.e. descent and ascent) navigation performance are examined. The inclusion of the water-track data in the navigation system is for both cases shown to significantly enhance the precision, and more importantly, the robustness and sustainability of the integrated navigation system. While not considered in this thesis, the methodology can easily be extended to under-ice operations using an upward facing DVL for measuring the ice-relative velocity. Similarly, the movement of the (slowly or non-rotating) ice may be estimated and taken into account.
The third and last topic examined in this thesis is the verification of underwater transponder positioning (UTP) and tight integration of acoustic range measurements with INS. While the two velocity aiding techniques mentioned above significantly enhance the navigation performance and level of autonomy, they cannot bound the INS error drift indefinitely. The use of UTP on the other hand allows truly autonomous operations once moored on the seabed, and the AUV may navigate with bounded error by visiting the transponders occasionally. Any number of transponders may be integrated with the INS. Compared to conventional long base-line (LBL) positioning, UTP provides increased accuracy due to tight coupling with the INS, increased operating area, better robustness, and significantly less deployment cost. Both in-situ and post-processed navigation results verify that UTP aiding is a feasible and accurate INS aiding technique which improves underwater navigation capabilities for systems where the need for flexibility, redundancy and navigation autonomy is important. This thesis is one of the few published reports showing experimental results of two-way travel time range aided INS, and to the author’s best knowledge, the first report documenting in-situ large-scale operations with multiple independent seabed transponders.
In summary, this dissertation advances the current state-of-the-art in underwater inertial navigation and autonomy by experimentally demonstrating the feasibility and performance enhancement achieved by aiding the INS with several unconventional aiding tools. | nb_NO |
