Link Foundation Ocean Engineering and Instrumentation Fellowship Reports
Underwater vision is usually limited. Object detection and identification is therefore one of the main challenges of underwater navigation. A new sensing modality, specifically developed for underwater environments, would greatly increase the scope of underwater missions. Taking inspiration from the lateral line of fish, I believe that pressure sensing can be a viable alternative to vision in order to detect and identify obstacles. Recent advances in the area of micro-engineering will soon enable to build sensors that match the size and mimic the functions and organization of the lateral line. However, little is known about how the pressure distribution along a fish relates to an obstacle location and shape. Detecting and identifying obstacles from distributed pressure sensors is a complex inverse problem that can be solved using Bayesian inference. For Bayesian inference to be practical, one needs to be able to solve the direct problem in real-time. Therefore, the aim of this project is to provide a tool for fast estimation of the pressure distribution along a vehicle caused by an obstacle. Computational Fluid Dynamics (CFD) is a great tool to investigate flow characteristics because it can give access to information that is not easily accessible experimentally (such as the pressure field). However, the configuration of interest here, which features a vehicle in motion relative to a static obstacle, is very challenging for traditional CFD tools. Such tools (referred to as body-fitted) require the computational grid to conform to the boundaries of the fluid domain. Deforming and re-generating a grid to comply with prescribed motions can eventually become more time consuming that actually calculating the flow properties. Immersed Boundary (IB) methods are much more convenient since they allow the boundaries to move independently of the grid. Unfortunately, currently existing IB methods do not perform satisfactorily for flow over a streamlined body at high Reynolds number (typically between 1000 and 10000 for small fish) because of the thin boundary layer. I have showed that widely used IB methods referred to as direct forcing methods only have a first order treatment of the immersed boundary. I have proposed the addition of a higher order term to an existing method which results in an easy to use flow prediction tool for the configuration and Reynolds number of interest . Though this Navier Stokes solver allows me to accurately calculate the pressure resulting from an obstacle, it is orders of magnitude too costly for real-time use.
Michael S. Triantafyllou
Maertens, Audrey, "Underwater Object Detection And Identification Using Distributed Pressure Sensors" (2012). Link Foundation Ocean Engineering and Instrumentation Fellowship Reports. 20.