Ralf Bachmayer, Peter Kampmann, Hermann Pleteit, Matthias Busse, Frank Kirchner
Editors: Frank Kirchner, Sirko Straube, Daniel Kühn, Nina Hoyer
In AI Technology for Underwater Robots, Springer International Publishing, volume 96, pages 71-82, 2020. ISBN: e-book: 978-3-030-30683-0 hardcover: 978-3-030-30682-3.
Free floating underwater robotic vehicles are free to move in all six degrees of freedom. While active pitch and roll is typically limited by design, i.e. hydrostatic stability, the robots attitude, position and speed control is based on thrusters possibly in combination with control surfaces, moving masses or variable buoyancy systems. Current systems often lack self-diagnostic capabilities and redundancy, leaving the high level mission control in the dark about the state of the thruster. This lack of information can lead to uncertain binary decisions about aborting or continuing missions. Better information possibly taking system redundancy into account will make it possible for the high level mission controller to scale the fault or system performance response accordingly, increasing the likelihood of at least partial mission success including system and data recovery compared to loss of data and possibly total system loss.
In this chapter we propose to approach the topic of propulsion from different perspectives like motor design and control, systems engineering as well as optimization through machine learning and adaptive identification and control.
The driving motivation is the research towards a propulsion solution, that suffices the requirements for a long-term autonomous underwater robot with respect to high system efficiency, reliability, and self-diagnostic capabilities. This will be achieved through an integrated systems approach between the electric machine, the propeller and possibly a nozzle. Furthermore research is going to focus on the real-time system performance using machine learning techniques in combination with more deterministic model based approaches for performance prediction and monitoring for failure detection of soft and hard errors.