Autonomous USV Navigation and Control for AUV Tracking via Optimal, Collision-Free Trajectory Generation

The increasing reliance on Autonomous Underwater Vehicles (AUVs) for complex subsea operations is often hampered by the severe limitations of underwater communication. This thesis addresses this challenge by developing a comprehensive autonomy stack for an Unmanned Surface Vehicle (USV) designed to act as a mobile communication gateway for an AUV. The core objective is to enable the USV to autonomously track the AUV's trajectory while ensuring safe navigation in cluttered and dynamic marine environments. This work, developed in collaboration with Saipem, presents the design, implementation and rigorous comparative validation of multiple hierarchical planning and control architectures. The methodology involves the development and integration of several state-of-the-art algorithms within a ROS 2 framework, validated in a high-fidelity Gazebo simulation environment featuring realistic hydrodynamics and wave disturbances. At the global planning layer, a metaheuristic Improved Particle Swarm Optimization (IPSO) with Simulated Annealing (SA) is compared against a sampling-based Real-Time RRT* (RT-RRT*). The local planning layer evaluates the performance of the reactive Dynamic Window Approach (DWA) and Vector Field Histogram (VHF) against an optimization-based Model Predictive Controller (MPC). To complete the control stack, these high-level planners are coupled with two distinct low-level motion controllers: a classic Proportional-Integral-Derivative (PID) and a model-based Sliding Mode Control (SMC). A systematic experimental analysis, based on a comprehensive set of performance metrics, is conducted to evaluate this matrix of architectures. The results reveal critical trade-offs between mission speed, path efficiency, control smoothness, and safety. The analysis demonstrates that the SMC consistently enables faster mission completion than the PID, albeit at the cost of higher control effort. The findings identify two top-performing hybrid architectures: RT-RRT + DWA (PID), which excels in energy efficiency and control smoothness, and RT-RRT + VHF (SMC), which achieves the fastest mission times. The standalone MPC is confirmed as the safest but most computationally demanding solution. Ultimately, this thesis provides a quantitative framework for selecting an optimal mission-specific autonomy architecture. It concludes that for robust and all-around performance, a hierarchical approach combining a sampling-based global planner with a reactive local planner offers the best balance of efficiency, safety, and real-time feasibility for cooperative USV-AUV operations.