COSYO – COmplex SYstem Optimization

Principal investigator: Abraham Duarte Funding entities: URJC and Comunidad de Madrid (URJC-CM-2008-CET-3731) Duration: 01/01/2009 - 31/12/2009

Abstract:

There is a type of optimization problems especially difficult to solve in which only partial information is available, called Complex Systems. In these problems there is no explicit description of the problem since some of its characteristic elements, such as the objective function or the constraints, are obtained indirectly. As a consequence, they are treated as a black box.

The research project will focus on the design of a generic solver (Context-Independent Solver) for the optimization of complex systems using metaheuristic techniques. The Solver developed will generate solutions as input to the black box and then analyze the returned result, extracting information about the solutions, so that iteratively generate higher quality solutions.

To design the Solver, first the problems will be categorized according to whether they are described by integer variables, permutations of elements or continuous variables. Subsequently, a method based on metaheuristics will be designed to solve each type of problem. The last step in the design of the generic Solver will consist of the integration of the three methods into a single general scheme that will select the most suitable one for solving each problem. The Solver will be complemented with an implementation of the Solver in a tool called COSYO.

COSYO will be a generic Solver for the optimization of complex systems modeled as a black box. Two user profiles of the tool will be considered. On the one hand, researchers or professionals with optimization knowledge (using it as a programming library) and, on the other hand, professionals who do not have advanced knowledge in optimization (using it from the OpenOffice.org spreadsheet).

The methods proposed in the development of the project will be compared with the best existing methods for such problems in both academic and commercial fields. This will result in both an application providing high quality solutions and scientific publications of international impact.