Preview-based control methods for ocean wave disturbance mitigation for underwater robots

Abstract

A major factor in the drive to reduce the costs associated with offshore energy generation has been the development and adoption of robotic systems, which offer cost-efficient and safer alternatives in comparison to traditional human-operated maintenance procedures. With respect to renewable energy devices, challenges arise when attempting to deploy robotic technologies owing to the energetic environments in which devices are located, where large magnitude ocean waves are common-place. Here, shallow operational depth magnifies the complexity and associated risks of inspection and maintenance tasks, meaning traditional feedback control often lacks in performance and fails to guarantee safe operation; this has been a major limiting factor in technology uptake thus far. By investigating a solution to this control problem, the adoption of robotics by the sector will be accelerated, indirectly contributing to reducing operational costs and improving the uptake of clean ocean energy technologies.

In this thesis, the challenges highlighted above are investigated in relation to both rigid and soft robotic systems, aiming to devise a control methodology which can successfully minimize wave-induced disturbances. To do so, firstly a computationally inexpensive and generalised model to estimate the wave disturbances experienced by an underwater vehicle is developed and validated experimentally. Upon verification of this low-order model, an analytical reformulation is performed to establish scaling laws for positional error during feedback-controlled station keeping under wave disturbances. These scaling laws provide a first assessment tool in defining the level of control performance required and define a clear threshold of operative conditions beyond which feedback control becomes unreliable.

This motivates the need for more advanced control strategies. The proposed solution in this thesis is postulated on the inclusion of short-term horizon disturbances within the calculation of future control actions. An ocean wave predictor is implemented based on Deterministic Sea Wave Predictions (DSWP), a technique originally intended for renewable energy harvesters which operate at fixed-points. Here, the accuracy and applicability of DSWP is experimentally verified for short-time horizon conditions, which can be exploited for vehicle control applications. The ocean wave predictor is then implemented in two main control methods: one exploits the disturbance preview knowledge as a purely feed-forward control action, whilst the other embeds this knowledge within a Nonlinear Model Predictive Controller (NMPC) to optimise over a time horizon. A sensitivity analysis is performed with regards to robustness to noise and performance of both control methods, exploring the applicability with regards to real time operations. Lastly, a similar approach is adopted for control of a soft underwater manipulator, which could potentially be mounted on a station keeping vehicle. This explores the use of NMPC for controlling the position of the end-effector in the presence of wave disturbances. This final piece of analysis examines the applicability of the approach in relation to a system with highly nonlinear dynamics, where the body is anticipated to be influenced even further than the rigid-body case.

Upon presentation of the above work, this thesis introduces a solution for exploiting the predictive nature of wave-induced disturbances on submerged bodies to mitigate the experienced loads; in doing so, the effectiveness of this new approach is proven to vastly increase the operational range of existing robots, with minimal external infrastructure modifications required. This realistic solution presents a high potential opportunity to increase the level of automation currently deployed within the offshore sector, facilitating the ability to operate in hazardous scenarios safely and effectively.