Source code for the Surrogate-Assisted Illumination (SAIL) algorithm, as described in:
"Data-Efficient Design Exploration through Surrogate-Assisted Illumination" published in Evolutionary Computation. https://arxiv.org/abs/1806.05865
"Data-Efficient Exploration, Optimization, and Modeling of Diverse Designs through Surrogate-Assisted Illumination" presented at GECCO 2017. https://arxiv.org/abs/1702.03713
and
"Aerodynamic Design Exploration through Surrogate-Assisted Illumination" presented at AIAA Aviation and Aeronautics Forum 2017. https://hal.inria.fr/hal-01518786/file/aiaa_sail.pdf
Two domains are provided with parameterized and feed forward deformation encodings: 2D airfoils and 3D velomobile shells. To apply SAIL to a new domain only new representation and evaluation functions must be created. More sample domains will be madepublic as their are published. If you are interested in creating a new domain and having trouble, don't hesistate to ask!
Produced using
Matlab R2017b
Required MATLAB Toolboxes:
-
Parallel Computing (for parallel speed ups)
-
Statistics and Machine Learning (for Sobol sequence)
Includes: GMPL Version 4.1, Rasmussen & Nickisch (help gpml)
Required Software:
Airfoil domain:
XFOIL low Rn airfoil design and analysis code
(raphael.mit.edu/xfoil/)
Velo domain:
OpenFOAM computation fluid dynamics simulator (version 2.4.0)
(https://openfoam.org/download/2-4-0-ubuntu/)