• autonomous underwater vehicle
• control allocation
• optimal control
• underwater optical wireless communication

This thesis has the ultimate goal of contributing to increase the reliability and robustness of Autonomous Underwater Vehicle (AUV) operations. It focuses specifically in inter-vehicle communication reliability and in robust thrust allocation, offering developments and contributions as detailed in the following sub-sections.
There are several ways to contribute in the improvement of the communication reliability underwater, but as a matter of fact the acoustic technology reached a mature level and the last improvements are related to increase the data rate at short distances. To achieve higher speed and a more reliable communication link, it is reasonable to adopt an alternative technology based on optical transmission that can provide such features in short communication ranges. The proposed solution based on visible light is provided in the following where the optical communication is employed in a harsh operative scenario with high disturbances such sunlight and turbidity.
The thrust allocation is another subject where the robustness of AUVs can be improved. The proposed work exploits the redundancy in the actuation system of over-actuated AUVs in order to improve the performance and robustness of the vehicles. The goal is achieved by solving optimal constrained problem also formulated on the mixed logical dynamical system concept.
Optical Modem
The first part of this thesis deals with the development of a novel Underwater Optical Wireless Communication (UOWC) device to exploit the potential of the Light Emitting Diode (LED) technology in an underwater scenario. In particular, the optical modem is designed to operate in shallow harbour water which is the worst scenario for optical communications. To be usable in operative condition, the performance must address a reasonable communication range and high data rate, so the modem is designed to achieve at least a maximum range of 10 m at 10 Mbps. The high speed allows the transfer of big data in a reasonable time without the need of the AUV resurfacing, and the distance of 10 m is an acceptable minimum range. The success of the project is the first step to validate the visible optical communication as additional technology to improve the existing underwater transmission methods. The optical modem is designed and funded within the European Project 7th Framework Programme for Research and Technological Development (FP7) SUNRISE as sub-project that applied at the first open call for introducing a novel technology in one of the SUNRISE infrastructures. The detail of the project and the development of the optical device are organized in the first module of the thesis as following:
• The OptoCOMM Project: the project is described in details with a comprehensive introductive state-of-the-art, the motivation behind the project, and the description of the scenario where the modems are deployed.
• Design of the modem: the development of the modem is described in all its fundamental parts: the employed electronics for the optical part and the hardware, the mechanical structure with the watertight container, and the software inclusive of the user-interface as well as the communication protocol.
• Experimentations: this chapter presents the laboratory tests done to validate each part of the modem as well as integration between all the components. At last, sea trials were performed to test and validate the adopted design in the proposed scenario.
Control Allocation
The second module of the thesis exploits the features of Mixed Integer Linear Programming (MILP) and Mixed Integer Quadratic Programming (MIQP) in the control allocation problem to improve the performance of the over-actuated AUVs. The control allocation problem can be defined as finding a set of input commands to the actuators such that the forces and torques exerted on manoeuvrable Degree Of Freedoms (DOFs) of the system are equal to the desired ones computed by the control module. The control allocation problem maybe is underestimated in the Guidance Navigation and Control (GNC) scheme adopted in all autonomous vehicles, but especially in the over-actuated vehicles plays a crucial role in the control effectiveness. The classic method adopted is to map the controls on the actuator by means of the pseudoinverse of the geometric allocation matrix. This method has some limitation, especially in the case where the requested forces go beyond the limits of the actuators. The actuators saturation and the redundancy in the actuation make the problem constrained and with multiple solutions, and more effective methods are used to solve the problem, i.e. Quadratic Programming (QP) and Linear Programming (LP). The proposed method improves the existing methods based on QP and LP by integrating in the control allocation problem the characteristic of the actuators and consequently the problem is transformed in the mixed-integer version. A comprehensive comparison between the existing methods and the pro- posed ones is presented to validate the real improvement. Moreover, the over-actuation allows the partial recovery if a fault occur on a thruster, and the control allocation must be capable to re-allocate the requested forces on the healthy thrusters. The Fault Detection & Isolation (FDI) with fault recovery is an other interesting sector where the control allocation is crucial for the effectiveness of a vehicle recovery or conversely the lost of it. Thus, in the thesis is presented a FDI technique with fault recovery by means of the proposed techniques. In the thesis, the developed techniques are tested in simulation on the over-actuated AUV V-Fides and in some cases verified on the real vehicle during the sea trials. The V-Fides vehicle was developed within the V-Fides project funded by Tuscany Region where Finmeccanica was the leader. The control allocation module is presented as following:
• The Control Allocation: in this chapter is presented an introduction with a comprehensive state-of-the-art, and then the control allocation techniques are presented. The investigated resolution method are: the classic pseudoinverse, the LP, QP, MILP, and MIQP.
• Case Study: The V-Fides Vehicle: this chapter describes in details the vehicle and the model adopted in the developed simulator implemented in MATLAB Simulink. The comparison between all the control allocation techniques is simulated on the V-Fides vehicle. In addition, the control allocation system is tested in FDI and Fault Recovery scheme to reallocate the requested control on the healthy actuators.