CFD in the Loop: Ensemble Kalman Filtering With Underwater Mobile Sensor Networks

In marine environments, sparse in-situ measurements can be used for the estimation of the fluid dynamic field. To make best use of a mobile sensor network in an environment whose dynamics can be described by the Navier-Stokes equations, we developed a framework for data assimilation with motion-constrained underwater vehicles, that takes the physical field properties into account while sampling. Our algorithm uses an ensemble Kalman filter that propagates hundreds of slightly varied coarse fluid dynamic simulations through time. Flow and scalar measurements from the mobile sensors are integrated into all ensemble members. We implemented a model predictive controller to calculate covariance minimizing paths from the estimated flow field and motion primitives of the vehicles, which are affected by a strong current. Thereby, we were able to indirectly track dynamically changing wall temperatures through measurements of flow field variables.