Autonomous Docking: System identification, Model Predictive Control and control allocation of marine vessels

Abstract

This thesis investigates autonomous docking for monohull marine vessels, with the objective of developing a generalized maneuvering framework applicable across multiple monohull vessels. In today’s application regarding autonomous docking, the solutions are often highly vessel-specific, which limits the ability to adapt. To address this limitation, this thesis proposes a general maneuvering script capable of identifying different vessel dynamics and supporting the vessel in critical low-speed docking scenarios while maintaining safe and accurate behavior. The proposed approach integrates vessel dynamic modeling together with advanced control and optimization techniques, including genetic algorithm, Model Predictive Control, and control allocation methods. A mathematical vessel model is used to approximate real-world behavior, which limits the controller and optimizer to realistic results. The genetic optimization process is used to tune the dynamic model based on real vessel behavior using recorded data, while Model Predictive Control ensures optimal decision-making over a finite horizon. The control allocator is applied to distribute control inputs among the available actuators in a feasible and efficient manner. The results show that the proposed framework performs well in simulation environments and on a simulation rig, successfully executing autonomous docking maneuvers while adapting to various vessel characteristics. However, performance in real-world open-water experiments was limited due to model inaccuracies, hardware constraints, and time limitations, indicating that further development and testing are required for reliable real-world application.