Autonomous Aerial Ice Observation
Joakim Haugen · BIBSYS Brage (BIBSYS (Norway)) · 2014
This work is concerned with autonomous aerial ice observation. Ice observation is a supporting activity in cold regions marine operations that are disturbed by various ice features. This supporting activity is motivated by the requirement of maintaining an awareness map of the surrounding ice conditions in order to execute an operation in a responsible manner. It is desired that the ice monitoring occurs both efficiently and as autonomously as possible. A part of the ice monitoring is thus to construct frameworks that are capable of executing various monitoring tasks without, or with minimal human intervention. Chapter 2 covers viable instrumentation configurations for remotely sensing different ice features from unmanned aerial vehicles. The chapter also motivates the use of unmanned aerial vehicles together with other sensor platforms, so that the strengths and weaknesses of the various sensor platforms can be exploited when maintaining the ice condition awareness map. The task of monitoring moving surface objects, often called target tracking, is examined in Chapter 3. We make use of nonlinear programming to construct feasible continuous trajectories for mobile sensing agents. The proposed framework uses each object’s Riccati differential equation, which is based on the continuous extended Kalman filter, in feasibly guiding the mobile agents between the objects. The framework is validated by a full-scale hybrid experiment where a singular fixedwing aircraft monitors three simulated objects in a constricted region of operation. We also explore the nonlinear programming approach in solving the dynamic coverage problem in Chapter 4. Here, the task is to remotely monitor a dynamic process in a planar region with mobile sensor agents. As in Chapter 3, the mobile sensor agents have maneuverability constraints that the framework takes into consideration when finding paths that the sensors should follow. The framework employs a simpler model, compared to the Riccati equation in Chapter 3, in describing how the possible information reward changes in space and time. The machinery of control theory and Lyapunov functions is investigated in Chapter 5 as a more computational efficient alternative to the nonlinear programming approach.