The main objective of this thesis is to investigate and present the most relevant techniques and topics within the field of robot modeling and control. The studies will then be used to develop a working control system for the 'Raptor' manipulator stationed on the ROV 'SubFighter 30K'. Due to insufficient information, a simplified model is made to resemble the actual manipulator. This model forms the foundation of all subsequent actions, including the model based control design. The dynamic model is developed by the well known method of Euler-Lagrange. Since this is an energy based method, both the kinetic and the potential energy of the system must be calculated. Systematic procedures are given to clarify the process of these calculations. In this thesis, a sliding-mode controller is derived and proposed as a suitable controller for the given manipulator. The control objective is to force the manipulator to track a time dependent, desired path in the joint space. However, since it is inconvenient for the operator to specify joint space trajectories, several inverse kinematics algorithms are suggested. Due to the kinematic structure of the manipulator, no closed-form solutions can be obtained. The focus is thus directed towards numerical Jacobian based methods. A full-scale implementation requires a working interface between the developed control system and the manipulator system. For that reason, the main concepts of digital communication are presented. Although no communication data is logged from the Raptor, this presentation will pose an advantage if the work is continued. When no control forces are applied to the dynamic model, the manipulator model is expected to behave like a multi joint, three dimensional pendulum. The simulations corresponds to this assumptions, thus the model is assumed to be correct and valid. Simulations of the complete system shows that the sliding-mode controller works as intended. Two chosen IK algorithms are then implemented and compared through simulations. The DLS method proves to be superior to the simple inverse Jacobian method. Finally, the control system is implemented in LabVIEW and thus prepared for full-scale testing.

The main contribution of this thesis is an investigation of model predictive control(MPC) for marine diesel electric power plants. Recommendations and new ideasfor further development are emphasized.The motivation of the thesis is to develop a controller for diesel electric power plantsthat can control the plant in a more efficient way. This includes reducing wear andtear, fuel consumption, and emissions. However, the safety aspect is always themost important factor and must be handled with care.The case plant to be studied is a diesel electrical power plant consisting of severaldiesel driven generators (genset). These gensets produce electrical power to servethe electrical demands on a marine vessel. The consumers can be propulsion units,heave compensators, drilling equipment, and hotel loads. These highly dynamicconsumers are large compared with the producers. This gives unwanted fluctuationof frequency. In some vessels this effect is so large that more gensets are requiredfor transients than for peak demands. This can be avoided with better controlstrategies.The controller developed in this thesis adjusts the local controllers on the dieselengines. The objective is to keep the genset at a given load sharing, while keepingthe frequency within rules and regulations. In addition is the plant controlled to astate where a single point failure does not lead to blackout.Blackout is prevented by calculating a failure case in addition to the normal case.The failure case may be a disconnection of the largest genset on the power bussegment. The case is calculated in the controller to make sure that if the caseoccurs the plant is able to handle the failure without a blackout. A normal case,where everything goes as normal, is calculated to optimize the current operation.The controller is verified by simulation done in MATLAB/Simulink. Theimplemented controller performs well during all of the simulated cases. However,the predictions made by the controller are in some cases conservative. This is dueto the choice of the fuel rate constraints. Lastly, suggestions for how to improvethe performance of the controller are included. The most important suggestionsare to include a model of the turbocharger in the control plant model and toinclude more failure cases.

Installation of deep water Xmas trees for subsea oil production is sometimes done by lowering the tree using one wire. Xmas trees are interfacing with other equip- ment on the seafloor and will therefore need to be positioned, and oriented correct. Today aligning the Xmas tree to existing interfacing structures on the seabed dur- ing installation is done by manual control of an ROV. In this thesis it is proposed to automate this process. The benefits of doing this is to gain a faster and more precise control of the position of the tree, as well as being less prone to human errors. This will make the operation faster, safer and less expensive.This thesis is a feasibility study of this new solution for aligning the Xmas tree in the installation process. In order to investigate the feasibility of the solution a mathematical model of an ROV and a Xmas tree that is hanging in a wire has been developed. Interaction forces between these objects are then defined. A model of the environment and sensor output from the ROV is also developed.A controller algorithm has been defined that uses the position of the Xmas tree to calculate the control forces the ROV need to control the motions of the Xmas tree. This controller algorithm is a nonlinear PID-controller where the output is translated from the body-centre of the Xmas tree to the body centre of the ROV in order to use some of the existing control system in the ROV. In addition to the controller an extended kalman filter has been implemented in order to handle the sensor feedback, and a reference model has been made to generate smooth and feasible trajectories as input to the control system.The mathematical model is used to simulate the behavior of the system, when the control system is connected. The simulations shows very promising results.An experimental setup has been made in order to test the control system in real life. The experimental setup consists of a downscaled Xmas tree model, a docking frame, and a small scale ROV. This setup was used to test the control system in the Trondheim fjord using R/V Gunnerus. Due to an error in the programming the testing was not able to fully verify the simulation results. The results obtained during the test did however seem promising when this error is taken into account, and thus proof of concept was established.

This thesis consists of three main parts. The first part of the master thesis looks at the identification of thruster dynamics and low speed ship dynamics. The relevant parameters identified are time constants and time delays in the system. Simple step tests are used for the identification. Different models for identification are suggested, both for uncoupled surge, sway, and yaw dynamics. Other test results, such as agility plots, DP 4 corner tests, and pure DP tests (stationkeeping) are reported. All the results are to be compared to similar tests performed after R/V Gunnerus has a retrofit of the thruster system.The second part discusses another problem, and that is the topic of numerically estimat- ing the body frame position of the GNSS and MRU sensors. For the GNSS position an Luenberger observer design and an adaptive scheme are proposed and analyzed. The es- timation designs are tested using numerical simulations and experimental data from the Gunnerus sea trials. A similar Luenberger observer is proposed for the MRU positions, and experimental data from the sea trials are used to test the observer.The third part discusses a hybrid augmentation of integral action. The motivation is a DP system, where typically the integral action is tuned very low to avoid oscillations due to the integral action. When there is a sudden load change, such as a ice load that hits the vessel, or if a mooring wire snaps, then a hybrid update augmentation could be useful, to speed up the convergence of the integral action. The update law is a linear update law based on the error in the states (the velocity for the DP system). The augmentation can significantly improve performance, especially for very large disturbance changes.