A UAV ice tracking framework for autonomous sea ice management

This paper describes an unmanned aerial vehicle (UAV) ice tracking framework for use in sea ice management applications. The framework is intended to be used in an ice management scenario where the UAV should detect and track the movement of icebergs and ice floes in an Arctic environment, and seeks to enable the UAV to do so autonomously. This is achieved by using an occupancy grid map algorithm and a locations of interest generator coupled with a Model Predictive Control (MPC) UAV path planner. The main contribution of this paper is interfacing the occupancy grid map algorithm with a machine vision object detection module in order to enable the UAV to generate an occupancy grid map of a pre-defined search area in real-time using on-board processing of UAV sensor data. Further, the paper presents a locations of interest generator module which generates locations that the UAV should investigate based on the generated occupancy grid map. These locations of interest are then used by an MPC path planner in order to make the UAV autonomously investigate and track ice features at said locations. Furthermore, the paper verifies the use of the developed ice tracking framework for autonomously detecting and tracking ice features based on thermal images captured with a UAV, as well as verifying the usefulness and role of UAVs in ice management scenarios by conducting two flight experiments.