Underwater iceberg profiling and motion estimation using autonomous underwater vehicles

Zhou, Mingxi (2017) Underwater iceberg profiling and motion estimation using autonomous underwater vehicles. Doctoral (PhD) thesis, Memorial University of Newfoundland.

[English] PDF - Accepted Version Available under License - The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission. Download (42MB)

Abstract

Icebergs originating from high latitude glaciers have drawn much attention from scientists and offshore operators in the North Atlantic. Scientists are curious about the iceberg drift and deterioration, while the offshore industry is concerned about the potential risks and damages on offshore oil platforms and infrastructures. In order to provide information to improve the iceberg drift and deterioration model constructed by scientists, and to assess the threats posed by icebergs to offshore platforms, iceberg shapes need to be measured. For the above water portion, optical instruments such as a camera and a laser scanner/LIDAR can be used. However, measuring the underwater portion of an iceberg is more challenging due to navigational constraints and sensor limitations. One approach, commonly used, is to deploy a horizontal plane scanning sonar from a support vessel at several locations around the iceberg. There are many drawbacks to this method, including the cost, sensing trade-offs in resolution and coverage, as well as constraints because of weather conditions limiting safe operations. The technology of Autonomous Underwater Vehicles (AUVs) has been developing rapidly in the last two decades. AUVs are commonly chosen to carry scientific sensors for various oceanographic applications. Without human intervention, AUVs can accomplish pre-programmed missions autonomously and deliver scientific data upon the users’ request. With these advantages, AUVs are considered as potential candidates in underwater iceberg sensing operations because they can operate close to icebergs to measure shapes and collect environmental data of the surrounding water. Sonar is usually used for underwater mapping applications. Since AUVs are typically quieter acoustically than manned surface vessels, a low noise to signal ratio can be achieved on sonars carried by AUVs. In this research, a technology of AUV-based underwater iceberg-profiling is evaluated. An iceberg-profiling simulator is constructed to analyse underwater iceberg-profiling missions. With the simulator, the accuracy of AUV-based operation is compared with conventional methods of deploying sonar profilers around icebergs. Beyond the simulation, a guidance, navigation, and control (GNC) system is designed with an objective of guiding the vehicle traveling around the iceberg at a standoff distance. The GNC uses measurements from a mechanical scanning sonar to construct a vehicleattached occupancy map (VOM) that the probability of occupancy of the cells in the VOM is updated based on a dynamic inverse-sonar model. Using the occupancy information about the cells in the VOM, the line-of-sight (LOS) guidance law is used to compute the desired heading for the existing heading controller in the AUV. The GNC is first calibrated and validated in a simulated environment. Then, an AUV equipped with a forward side-looking mechanical scanning sonar is deployed in the field. The GNC guides the vehicle circumnavigated an iceberg autonomously, and underwater shape of the target iceberg is represented using the sonar samples. The point cloud may deviate from the original iceberg shape due to the iceberg movement. A motion estimation algorithm is developed to estimate the iceberg motion for converting the point cloud into an iceberg-centered coordinate system. Two point clouds measured at different times, inputs of the motion estimation algorithm, are presumed to be identical in the iceberg-centered coordinate system. Then, the algorithm iteratively updates the motion estimates based on the translational matrix and rotational matrix from an iterative closest point (ICP) algorithm to match the point clouds. The hypothesis that two point clouds are identical in the iceberg-centered coordinate system is valid when the motion estimates are converged in the updating process. Once the iceberg motion is resolved, the point cloud in the inertial coordinate can be converted in to the iceberg-centered coordinate to present the true iceberg shape. The algorithm for estimating iceberg motion is applied to data collected from the simulation environment and the field trials in Newfoundland.

Actions (login required)

View Item