 | Evolutionary robotics problems such as the development of controllers for autonomous robots are being used as a test project for building a grid-based framework for Artificial Evolution application.
Image courtesy of Erol Sahin
|
Evolution. Survival of the fittest. It seems like a pretty efficient way to work out which combination of genes offers the best solution to the challenge of life under a certain set of conditions. In fact, evolution is such a clever system that we’re trying to simulate it. And its so complex that we need grid computing to power those simulations. Grid-based Artificial Evolution is the result: a relatively new approach to solving complex problems, inspired by the mechanisms of natural evolution. AE solves a given problem by first evaluating the “fitness” of each of a population of candidate solutions. This approach can be used to solve problems in fields ranging from engineering and robotics to social sciences and genetics. Like evolution, AE is incredibly complex: the fitness values of the candidates are used to produce a new generation of candidate solutions using genetic operators such as recombination and mutation. This “reproduction” process favors the “fitter” solutions, while reducing the probability that non-fit solutions will be included in the new generation. This process is iterated until solutions of a desired fitness value emerge. Since this typically requires a massive number of parallel calculations and iterations, it is perfectly suited to grid computing. GridAE is an application in progress at the Middle East Technical University, Turkey. The proposed work aims to create a grid-based framework for AE applications that will be able to distribute the execution of the evolution onto the grid. GridAE is funded by and uses grid infrastructure established as part of the SEE-GRID project. | 
Headlines | Site Map
