Tools and Methods for Autonomous Operations on Seabed and Water Column using Underwater Vehicles

Abstract

This thesis concerns several aspects of the automatic system that controls the behavior of Remotely Operated Vehicles (ROVs), and elaborates the data collected during the operations. It presents the algorithms used to develop the motion control system implemented in the NTNU AUR-Lab’s fleet of ROVs, and to accomplish sea operations in collaboration with academic and industrial partners. The required improvements of such system, and the need for a higher level of autonomy in the field of marine robotics, have been the main drivers of this thesis. The participation in field work and practical missions, the collaborations with scientists with different backgrounds, and the current vision of the marine robotics community and of the centre for Autonomous Marine Operations and Systems (NTNU AMOS), have contributed to direct this research towards the nodes which the author considered crucial. This thesis presents contributions to the different modules of a control system for ROVs. A control system takes as input reference and sensor signals, using them to produce a proper control action in order to obtain the desired vehicle motion. The characteristics of the input signals are therefore important for the behavior of the whole control system feedback chain. For such reason, a signal processing module is developed and implemented, with the purpose of filtering out the outliers often produced by acoustic positioning systems. The sensor signals need to be merged and filtered, in order to produce a complete and reliable estimation of the vehicle. This problem, known as the navigation problem, can be solved by model-based or kinematic observers (estimators). Moreover, underwater vehicles are utilized in various use-modes and with different environmental conditions. These aspects can be taken into account in the design of the observer module. In fact, a bank of observers can be designed so that it is possible to switch between the single algorithms to improve the accuracy of the estimation. A so-called multi-objective observer, based on the nonlinear passive filter, is designed for this purpose. The reference signal, which is often a destination point in space or a line to be followed, cannot always be fed into the control system. A guidance module is often used to produce intermediate desired states that converge to the ultimate goal. This module is designed so that environmental disturbances, as well as kinematic and dynamic constraints, can be automatically taken into account during the process. In this context, an Integral Line Of Sight (ILOS) algorithm has been implemented and tested on an ROV system. The users are also part of the control chain, although they are often not formally included in the system architectures. This is important to have in mind, since the behavior of the human operators influences the performance of the vehicle. An important task for those who design control systems is to provide the operators with both necessary and sufficient information, and a way to efficiently steer and control the vehicle. In this context, an innovative Human-Machine Interface (HMI), which uses the Head-Mounted-Display (HMD) technology, has been developed and implemented, with the purpose of improving the state-of-art of the ROV interfaces. The camera technology mounted on the AUR-Lab fleet of vehicles is also included in this work, not only to produce data of valuable importance for the end users, but also to produce useful information for the control system itself. Underwater mapping is presented as an application of this technology, together with the development of underwater off-line and real-time mosaics. Underwater photomosaics and camera information can be used to give important information on the morphology of the terrain. With such information, the vehicle can be automatically steered giving priority to those areas with a greater density of Object Of Interest (OOIs). An attempt to understand the best direction of the vehicle, based on camera information, is proposed. A path-planning system is necessary to define the mission of a vehicle. In a context of increased autonomy, underwater vehicles have the necessity of being equipped with a fast and safe autonomous planning and replanning systems. In this work, a system to automatically plan and re-plan the path of a marine vehicle (possibly underactuated) is proposed and developed. The system has been first developed to plan surface vessel paths. Then, the system has been improved including a fast, automatic re-planning procedure, and finally extended to be used in the underwater three-dimensional environments. This work also presents some examples of practical operations, which have been performed along the Norwegian coastline by AUR-Lab engineers, technicians and scientists, in collaboration with different end-users commissioning the work (companies, other researchers, institutions). Some details of such operations, their critical aspects and the goals of the key people involved are described in this thesis, as well as the operations of AUR-Lab, which successfully merges the academic need for scientific development and the end user’s needs.