A robust return-home strategy for a long-range AUV operating beneath ice
Autonomous Underwater Vehicles (AUVs) are free-swimming robots that can explore regions of the ocean beyond the capabilities of ships or tethered vehicles. Their independence al?lows them to work deep, over long distances, and beneath the cover of floating ice. This independence comes at a cost as t...
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | unknown |
| Published: |
2024
|
| Subjects: | |
| Online Access: | https://doi.org/10.25959/26087335.v1 https://figshare.com/articles/thesis/A_robust_return-home_strategy_for_a_long-range_AUV_operating_beneath_ice/26087335 |
| Summary: | Autonomous Underwater Vehicles (AUVs) are free-swimming robots that can explore regions of the ocean beyond the capabilities of ships or tethered vehicles. Their independence al?lows them to work deep, over long distances, and beneath the cover of floating ice. This independence comes at a cost as they need to navigate and position themselves with no external aids or reference. This work aims to improve the ability of an AUV to conduct ex?ploratory missions in un-surveyed and high-risk areas, such as beneath the ice in Antarctica, and return to a known safe recovery location. AUVs suffer from navigational drift making it dangerous to conduct these missions in their current state since the presence of ice can hinder the use of external navigational aiding. A solution to this problem is to allow the AUV to venture into an area, while accumulating error, then re-trace the inward path on the way out; if path was safe to get in then it is safe to get out. Navigation is driven by the registration of sensor readings to a repository of data collected on the path in. To enable this a methodology known as teach-and-repeat, taken from terrestrial robotics with application in extraterrestrial exploration, was developed and tested. AUVs rely on acoustics. A methodology to generate image views from sonar data was developed, based on existing work in terrestrial robotics utilising visual cameras. An adaption of teach-and-repeat (TR) was developed to navigate an AUV along a previously traversed path through image registration and tested successfully in the field. The AUV navigated solely on the sonar information and its mapping to the initial path. The localisation was topological and relative only to the taught path, and thus was not affected by the inherent errors in global positioning. Further development applied this concept to an AUV collecting 3D bathymetry data, where subsets of data collected at different times could be compared to derive a likelihood that they were of the same area of the seafloor. This capability ... |
|---|