• autonomous underwater vehicle
• AUV
• control allocation
• control system
• Gazebo
• linear programming
• ROS
• underwater
• uuv_simulator

The thesis aim was to design a control system for an over-actuated AUV, with particular focus on the control allocation system. Of the latter problem various resolution algorithms were implemented and compared: Pseudoinverse, Linear Programming and Mixed Integer Linear Programming. The comparison between the various optimization methods was made by defining and evaluating some performance indicators, such as computational time and allocation error.
A request from the project leader of "A2I2" was to evaluate the performance of the robot for two different inspection tasks. Thus, a motion planner emulator was implemented for the two scenarios: the survey of a wall and the survey of a monopile. For each task various mission specifications were provided, such as: image stability, maintaining of a certain viewing angle and a desired distance from the inspected structure.
A 3D simulation environment, in ROS / GAZEBO, using AUV dynamics as well as models for thrusters and for navigation sensors, has been developed, starting from the open-source package uuv_simulator, in order to validate the software and acquire the results. The simulator is designed to be as generic as possible, in order to facilitate future developments and adapt to user's needs.
From the results of the simulations it was possible to carry out a more in-depth analysis of the allocation algorithms and a comparison of the vehicle performance with three different thruster configurations. From these verifications the LP was chosen as the allocation method, since it represents a trade off between allocation error and computational time.
As far as the comparison on the three configurations is concerned, it is chosen to evaluate the performance, in terms of compliance with mission specifications, energy used during the mission and mission time.
Once the software is tested, it is integrated with the On-board Control System, a vehicle-independent software developed by NOC engineers, respecting certain new constraints.
In order to make simulations, as realistic as possible, external disturbances (e.g. sea current) and faults in the actuators were added. A new vehicle performance analysis, evaluating the new results, was given.
Finally a fault reaction system was studied, in a theoretical way. This system should allow the robot to decide, in case of identified fault, whether or not to abort the mission. To avoid faulty thrusters some on-line change in thruster algorithm method are implemented, and a comparison in case of fault between the "standard" algorithm and the updated one was given.