One of the phases of the installation of Offshore Wind Turbines is the installation of the foundation for the turbine. Herein, a monopile is one of the most frequently used foundation-types. This monopile is a high steel cylinder, over 60 meters high, which supports the wind turbine to the seabed. Currently, monopiles are lifted with large boom cranes on jack-up ships. SeaState5 has developed a more efficient method for monopile installation by using the Grasshopper-T (GH-T), a new crane which skids the monopiles from the deck and places it onto the seabed. The system reduces the weight and footprint for jack-up ships.

This thesis consists of mechanical modeling the GH-T for the purpose of estimating the forces within the lifting system. A non-linear modeling method; Euler-Lagrange
(EL) method with constraints is used to capture the systems dynamic, non-smooth and non-linear behavior over the entire range of motions of the upending process. Secondly, for the purpose of simplification, yet capturing the mentioned behavior, the system is linearized around several areas of operation, analyzing local detectablity and observability. Thirdly, attempts are made to design linear observers and disturbance estimator to both estimate the states of the system and any disturbance applied on the monopile, like hydrodynamic forces. Finally, a non-linear observer is proposed to estimate the states along the entire upending process of the monopile. Experiments for model validation are performed on SeaState5’s 1:25 scale prototype for which a sensor setup has been designed.

Simulations show that linear observers based on the linearized model are able to estimate the states, and compute forces of interest, within a very small area of operations. The designed linear disturbance estimator proves to be ineffective in estimating disturbances and is only valid for the same small neighborhood around the equilibrium point. This motives the need for an observer capable of estimation over a broad range of motions. Therefore, a non-linear observer is designed. For the non-linear observer, the state recovery proves to be efficient and robust enough to cope with sensor noise and the high range of motions of the upending procedure. Experimental results show the model is able to capture the non-linear behavior of the real system and estimate forces of interest within acceptable margins, showing promises for an online state estimation using the non-linear observer.

This project addresses a fundamental problem faced by many reinforcement learning agents. Commonly used reinforcement learning agents can be seen to have deteriorating performances at increasing frequencies, as they are unable to correctly learn the ordering of expected returns for actions that are applied. We call this the disappearing reinforcements problem. Moreover, truly multi-task reinforcement learning is only possible when agents are able to operate across frequencies, as different platforms operate at different frequencies. Most algorithms from control theory working on similar tasks, on the other hand, show improved performances when their operating frequencies are increased. This suggests that addressing disappearing reinforcements should enable reinforcement learning agents to have improved performance and generalization ability across timescales and tasks.

In this project, we show that disappearing reinforcements is an effect seen independent of the function approximator used in reinforcement learning, and is instead of a more fundamental nature. We explore both theoretically and empirically the relationship between agents and their performances at increasing frequencies. We show that two specific types of agents from literature address the problem, and we benchmark the agents' performances with novel benchmarking measures inspired from control theory. Finally, we create a novel agent we call the dueling advantage learner, by combining both approaches from the state-of-the-art. We then benchmark the different agents across frequencies and tasks, and our agent is seen to outperform each of the individual approaches on the majority of the tasks.

Motion compensation for personnel transfer has greatly reduced costs and increased accessibility for offshore assets operation and maintenance. Solutions exist using hexapods or gangways that make a stiff connection by clamping to a customized offshore structure. State of the art gangways allow for a stepless workflow and have a rubber tip which does not require specialized landing structures. However, in these solutions contact force variations are high. This results in a reduction of safety due to the loss of contact or slipping motions. Therefore, a system with force sensors in the tip is proposed. This allows for the tip forces to be used in feedback control. Literature on this subject describes hydraulic actuator force control, impedance control, or force control in free moving robotics. Literature fails to describe suitable solutions to the hydraulic force control problem where motions are dictated and force tracking is required. In this thesis the dynamic problem of force-position control with hydraulics is simplified, allowing linear analysis and controller design. A 3D model has been created and used to show the need for force feedback after which an analysis is performed on the actuator dynamics. The root locus method is used to close the force control loop with a proportional-integral controller. This results in a complete force-position control structure which tracks a force setpoint on a moving object. The controller structure has been implemented during sea trials and showed significant improvement in the force control performance. This approach of force control using hydraulic actuation is expected to significantly increase safety in offshore operations.