Collision avoidance and motion control for autonomous surface vehicles

This thesis presents topics related to collision avoidance (COLAV) and motion control for autonomous surface vehicles (ASVs). The thesis contains a collection of nine publications, of which six are peer-reviewed conference articles, two are journal articles and one is a book chapter. In addition, it contains an introduction providing context to the topic of COLAV at sea as well as explaining the relationship between the publications. When designing a control system for an ASV, one can approach the COLAV problem in two distinct ways: using a single algorithm for handling all aspects of the problem, or employing a hybrid architecture with multiple algorithms to exploit their complementary strengths at different layers in the architecture. We have chosen the latter approach, resulting in a modular system that can be tailored and extended with new functionalities and algorithms. In particular, our COLAV system has three layers, distributing the task of COLAV, and its inherent motion planning tasks, at three distinct responsiveness and completeness layers. The top layer is responsible for performing energy-optimized path planning in a global setting, finding a trajectory throughout an environment populated by static obstacles such as land, islands, shallow waters and navigational marks. The planning algorithm uses a vessel model together with ocean current information to produce an energy-optimized trajectory, which minimizes the required energy to reach the goal. The middle layer follows the trajectory specified by the top layer, while avoiding moving obstacles in a predictable fashion in accordance with the International Regulations for Preventing Collisions at Sea (COLREGs). The COLREGs contains a set of rules on how vessels shall maneuver in situations where there is a risk of collision. Specifically, the developed algorithm used at this layer considers COLREGs rules 8, 13–16 and parts of Rule 17. The bottom layer follows the trajectory specified by the middle layer, while ensuring that commands specified for the vessel controller are feasible with respect to the vessel’s capabilities. This layer also handles emergency situations, such as obstacles detected late or behaving dangerously, and situations where the middle layer fails to find a solution. In particular, this thesis includes two algorithms for the bottom layer, which are tested in several full-scale experiments using a radar-based system for detection and tracking of obstacles. The first algorithm, based on the dynamic window (DW) algorithm, have problems with the amount of noise present in the obstacle estimates provided by the tracking system, as well as having other limitations. The second algorithm, named the branching-course model predictive control (BC-MPC) algorithm, addresses the weaknesses of the DW algorithm, and is the algorithm found to be most suitable for the bottom layer. The BC-MPC algorithm ensures compliance to the remaining parts of COLREGs Rule 17, while also considering rules 8 and 13–15, making our COLAV system compliant with COLREGs rules 8 and 13–17. A COLAV system is highly dependent on a well-performing vessel controller. Relevant work to that end in this thesis has been directed towards high-speed ASVs, which commonly operate both in the displacement, semidisplacement and planing regions. Existing methods for modeling such vessels are complex, and not well suited for controller design. Therefore, a method for modeling and identification of high-speed ASVs is also proposed. This method is used to create a model of a high-speed ASV, shown to be valid for the displacement, semi-displacement and planing regions. This model is used to design two vessel controllers that utilize model-based feedforward terms, shown to outperform traditional controllers in full-scale experiments. Summing up, this thesis contributes at all the three described COLAV layers, in addition to modeling, identification and control of high-speed ASVs, and represents a step on the road to autonomy at sea