+ ARTIFICIAL INTELLIGENCE FOR ADAPTIVE FLUID CONTROL
Inverse design approaches that combine artificial intelligence and simulations are powerful tools for the characterization of fluid-mechanic phenomena in biology as well as in engineering. In this context numerical solvers face many challanges. Algorithmically they must exhibit controllable accuracy, flexibility and adaptivity, while software-wise they must be able to achieve high performance on a multitude of hardwares.
We develop numerics that allow for the seamless integration between artificial intelligence techniques such as evolutionary optimization or reinforcement learning and large scale simulations. Applications range from the characterization of biolocomotion strategies to the group dynamics of energy harnessing devices.
Gazzola M, Tchieu AA, Alexeev D, de Brauer A, Koumoutsakos P, Learning to school in the presence of hydrodynamic interactions, Journal of Fluid Mechanics, 2016.
Rossinelli D, Hejazialhosseini B, van Rees WM, Gazzola M, Bergdorf M, Koumoutsakos P, MRAG-I2D: Multi-resolution adapted grids for remeshed vortex methods on multicore architectures, Journal of Computational Physics, 2015.
Gazzola M, Hejazialhosseini B, Koumoutsakos P, Reinforcement learning and wavelet adapted vortex methods for simulations of self-propelled swimmers, SIAM Journal on Scientific Computing, 2014.
Gazzola M, Mimeau C, Tchieu AA, Koumoutsakos P, Flow mediated interactions between two cylinders at finite Re numbers, Physics of Fluids, 2012.
Gazzola M, Chatelain P, van Re es WM, Koumoutsakos P, Simulations of single and multiple swimmers with non-divergence free deforming geometries, Journal of Computational Physics, 2011.
Gazzola M, Vasilyev OV, Koumoutsakos P, Shape optimization for drag reduction in linked bodies using evolution strategies,
Computers and Structures, 2011.
Chatelain P, Gazzola M, Kern S, Koumoutsakos P, Optimization of aircraft wake alleviation schemes through an evolution strategy, Lecture Notes in Computer Science, 2011.