Guidance and Control of Marine Robotics for Ocean Mapping and Monitoring

摘要

The ocean is an integral part of life on Earth, both as a source of food and oxygen, a medium of transportation and a major influence on weather and climate. Unfortunately, proper management and preservation of the ocean depend on knowledge we do not necessarily possess, due to what oceanographers refer to as a “century of undersampling”.\nAutonomous marine robotic platforms are showing promising steps towards increasing the quantity and quality of ocean mapping and monitoring. This thesis concerns development of methods for motion control and mapping systems for marine robotic platforms with varying levels of autonomy. The main motivation of the work has been to increase the level of autonomy of the systems, in order to reduce costs and the need for human interaction, while simultaneously increasing recording efficiency and data quality. The research questions guiding the work are focused on how marine robotic platforms with increasing levels of autonomy can improve the quality, quantity and efficiency of ocean mapping and monitoring, both in terms of data collection and interpretation. The thesis presents contributions to different modules of an overarching autonomy architecture.\nThe primary task of a marine robotic platform during a mapping mission is to bring one or more payload sensors to the appropriate locations in time and space. The motion of the platform must be efficient in order to record the desired quantity of data, and accurate in order to achieve the desired quality of data. Sensors such as underwater optical cameras, benefit from a methodical motion pattern and a constant distance to the area of interest (AOI), both for data quality and area coverage. This work presents an automated relative motion control strategy for mapping steep underwater structures using a Remotely Operated Vehicle (ROV). The strategy employs a Doppler Velocity Logger (DVL) oriented in the viewing direction of the cameras to maintain a stable distance to the AOI, ensuring high image quality. Through several full-scale experiments, the strategy is demonstrated to efficiently record high quality datasets of challenging structures.\nA major benefit of autonomous platforms is the ability to operate in scenarios that are deemed too dull, distant or dangerous for human operators. This work presents the results from several mapping missions performed with marine robotic platforms on multiple scientific cruises. Photogrammetry surveys that require highly accurate control over prolonged periods of time demonstrate situations where operator fatigue would have led to a gradual decrease in performance of a manual system. Arctic operations in the polar night that require measurements that are not polluted by artificial light exemplify scenarios which would have been impossible to perform within safety regulations for crewed vessels.\nIn order to approach a truly autonomous system, the system must be able to acquire and analyze the information necessary to make informed decisions. In this work, underwater hyperspectral imaging (UHI) is demonstrated to be a promising tool for high-resolution seafloor exploration and classification. The increased spectral information present in UHI-data compared to regular RGB-imagery is well suited for computer analysis and online classification of objects of interest.\nAutonomous inspection of seafloor structures or other AOIs poses interesting opportunities for both industry and marine science. This work presents experimental results from an ROV implementation of a behavior- and reactive based autonomy architecture. The architecture is able to conduct a realistic mission from surface launch to target area with obstacle avoidance.\nThe work is motivated by the needs of the end-users in the marine sciences, and involves interdisciplinary collaboration, in particular between the fields of marine control systems, marine biology and marine archaeology. The developed methods are applied to real world cases on a variety of