Section: Software
Metis
Participants : Olivier Teytaud [correspondent] , Adrien Couëtoux, Jérémie Decock, Jean-Joseph Christophe.
Energy, Optimization, Planning
Many works in Energy Optimization, in particular in the case of high-scale sequential decision making, are based on one software per application, because optimizing the software eventually implies losing generality. Our goal is to develop with Artelys a platform, Metis, which can be used for several applications. In 2012 we interfaced existing codes in Artelys and codes developed in the Tao team; experiments have been performed and test cases have been designed. A main further work is the introduction of generic tools for stochastic dynamic programming into the platform, for comparison and hybridization with other tools from the UCT-SIG.
Our favorite challenge is the hybridization of “classical” tools (based on constraint satisfaction problems, or mixed integer linear programming or mixed integer quadratic programming), which are fast and accurate, with non-linear solvers which can take care of a sophisticated (non-linear) model.