Datum predavanja: petak, 25. 11. 2022.
Vrijeme predavanja: 10h

Ben Wolf, TU Delft
Title: Hydrodynamic Imaging using Artificial Intelligence

Abstract:
We explore the information that is left behind by underwater objects in their wake, how this information can be measured, and how it can be used to identify properties of these objects, such as their relative location and shape.
To this end, a novel type of fluid flow sensor is developed, capable of measuring the fluid flow speed in two dimensions for the first time, which is shown to be beneficial for hydrodynamic imaging: determining the properties of moving objects by measuring their produced flow.
An array of these flow sensors is deployed at different length-scales, from several centimeters to several meters. The measurements resulting from nearby objects in motion are used to show that hydrodynamic imaging can be scaled up considerably from its biological dimensions.
The flow measurements are processed with a variety of artificial neural networks and deep learning methods, including the ELM, MLP, ESN, LSTM, and CNN. These are shown to be well-suited for determining an object’s location, its direction of motion, or its shape solely based on fluid flow data.

Bio:
Ben J. Wolf received the Ph.D degree (cum laude) in Artificial Intelligence from the University of Groningen, the Netherlands in 2020 on the topic of Hydrodynamic Imaging. He is currently a Post-Doctoral Researcher at the Delft Center for Systems and Control, at Delft University of Technology working on image-based under water trash detection in the context of the SeaClear project: a collaboration that includes UNIDU and DUNEA among others. His research interests include machine learning, neural networks, robotics, and hydrodynamic sensing.