M*: Multiobjective Metaheuristics and Multidisciplinar Aplications

Project Id: TIN2008-06491-C04-04
Ministry of Science and Innovation (Spain)


This project is aimed at innovating in multiple fronts of multiobjective optimization (MO) from the perspective of metaheuristic techniques. For this purpose, we will achieve a relatively ambitious set of contributions at the end of the project. First, we plan to advance in fundamental research by developing new multiobjective models for algorithms such as ant colony, scatter search, cellular genetic algorithms, particle swarm, differential evolution and other procedures capable of solving problems of realistic dimension and complexity. Second, the problems tackled will not be limited to typical instances drawn from standard benchmarks, but instead we will also address a multidisciplinar selection of problems from four hot domains in a coordinated manner to perform an applied research: software engineering, economy, communications and traditional engineering. This way, the benefits will lie in both methodology and real applications. The goal is to unify criteria for experimentation and evaluation of new MO techniques, and to improve their efficiency and effectivity with respect to the present state of the art in the mentioned domains; we aim at showing that the contributed techniques are not only appealing in theory, but also effective and useful for society.

To this, our methodology will include the study of where and when a multiobjective formulation is advantageous compared versus a monoobjective one, as well as the scalability of the techniques developed. We want to advance in the use of new technologies and research lines that are presently hot topics at an international level, but from a multiobjective perspective in our case. For this purpose, we will design MO algorithms that will use parallelism-based technologies (cluster computing, supercomputing, reconfigurable computing) and grid computing on complex problems, and study combinations with other techniques (hybridization) using also problem-aware operations. The studies about parallelism-based technologies, grid computing, hybridization, and a long list of new algorithmic extensions in a unified way inside one project, are only possible thanks to the coordinated work of four teams of researchers. In the same manner, the transfer of this new MO knowledge to the four selected domains is only possible in a joint manner, actually enabling the project to spread the advances to multiple research niches (informatics, operations research, algorithmics, specialists and applications…), thus increasing the international impact.

The main goals in the subproject MSTAR::UC3M are to create a body of knowledge in MO optimization and in applying the resulting techniques to benchmarks and to problems in the context of Economy and Classification:
Adaptation of some methods to Multiobjective problems: Adapt some existing techniques to their use in multiobjective problems. This includes Genetic Algorithms, Evolutionary  Strategies, Ant Colony Optimization, Particle Swarm Optimization and Estimation Density Algorithms.

  • Evaluation of the new developed techniques: Comparative evaluation of the designed  techniques. Their flexibility, scalability and performance, in comparison with the most used methods in the present.
  • Study of the appropriateness of coevolutive approaches: To study the way coevolutive approaches could be used in the frame of multiobjective. To design and  develop a coevolutive method competitive with the rest of methods.
  • Study of the appropriateness of hybrid approaches Analyse the way of combining some techniques and to evaluate possibles candidates to combine, how they could be merged together and their efficiency .
  • Solve problems in Economy and finance with efficiency/accuracy above the existing  solutions, paying special attention to the optimal assignment of radio-frequency spectrum problem, and other assignment problems using auctions.
  • Solve problems in the context of classification problems from a multiobjective  point of view.
  • Solve problems in Brain Computer Interface.