Underwater glider navigation techniques with ensemble-based estimation and streamline-based planning
- Publication Type:
- Thesis
- Issue Date:
- 2023
Open Access
Copyright Clearance Process
- Recently Added
- In Progress
- Open Access
This item is open access.
Underwater gliders are appealing for oceanic applications due to their endurance, allowing operation over hundreds of kilometres for several months. However, their limited thrust poses significant challenges for autonomous navigation under ocean currents and unreliable long-term forecasts. This thesis addresses three critical problems in estimation and path planning to reduce reliance on remote supervision.
The first problem focuses on flow field estimation with limited computational resources. We propose a novel algorithmic framework leveraging environmental estimation systems like those from the Australian Bureau of Meteorology (BOM) to extract recurring flow patterns from ensemble outputs. These patterns help reconstruct estimates of environmental behaviour, ensuring constant computational time complexities for updates from vehicle-measured flow velocities and providing estimates at given positions, unlike Gaussian process-based approaches that require cubic computational time.
The second problem, not well-studied in literature, involves determining fixed controls for underwater gliders to reach another position under ocean currents, solving the broader path planning issue. We develop streamline-based control theory foundations, inspired by oceanographic concepts, to constrain the search for fixed vehicle velocities without disregarding feasible solutions. Through simulated environments using synthetic flow patterns and the Eastern Australian Current (EAC) based on BOM data, we show that this constrained technique significantly improves path quality by over 35% compared to unconstrained searches with similar computation budgets.
The third problem explores new methods to improve upon the standard approach of generating feasible paths for vehicles in strong ocean currents, typically involving extending a search tree using random controls. We propose measures of reachability and a filtered approach to connect to random positions more effectively, accounting for kinematic feasibility with streamline-based control theory. Simulations in environments with vortices show our methods achieve higher connection rates than traditional Euclidean distance-based quantifications of reachability. Additionally, in scenarios using real forecast data from the BOM, our methods produce initial solutions earlier than the standard random controls approach.
This work provides an automated alternative to traditional underwater glider navigation workflows, transforming them into autonomous vehicles rather than tools. A robust automated navigation system for underwater gliders can revolutionise marine biology and oceanography research by simplifying multi-vehicle operations. Furthermore, streamline-based control theory offers promising applications beyond oceanography, providing innovative solutions for disciplines dealing with similar flow fields, such as electromagnetic and gravitational fields.
Please use this identifier to cite or link to this item: