Research On Evolutionary Algorithm Optimization Based On Extension Transformation

The computer is tending to more and more intelligence, and nomatter what kind of applications, intelligence has been the mostimportant factor. In the 1980s, the artificial intelligence theorybased on the evolution of structure—computational intelligencewas fast becoming the new mainstream. Computational intelligenceconsists of a wide range of research areas, which have profoundlinks and promote each other, and evolutionary computation is justan important field of these areas.The thinking of evolutionary robotics mainly comes fromevolutionary computation. In evolutionary robotics, the primarywork of designers is to decide evolutionary framework andassessment strategies. The system behaviors depend to a highdegree on the fitness function used in assessing. In designingmethods of fitness function there are many ways, for examples:linear transformation, power transformation, indextransformation, variance adjustment, etc. Meanwhile, theparameters in evolutionary framework, such as the selectionprobability, crossover probability and mutation probability arecrucial to the evolutionary process and results. The target of thispaper is to achieve the integration of the populations' diversityand convergence, evolution speed and performance, therebyoptimize the design of evolutionary framework and assessmentstrategies, with the design of fitness function and evolutionaryparameters. This is of great significance to the development ofevolutionary robotics and evolutionary computation.Extension study is a new subject, which deals specificallywith the issue of conflicts of thinking model, it is bound to infiltrate into artificial intelligence and related disciplines,and produce a new type of intelligent robots extenics-based robot.This paper introduces the existing genetic algorithm andExtension GA. On the basis of the extension transformation ofExtension GA, this paper studies the performance optimization ofrobot obstacle avoiding. Through the extension analysis of theobstacle-avoiding robot, I have brought forward a matter elementmodel of the obstacle-avoiding robot and design the correspondingextension fitness function. Through experiments and simulation inthe Evorobot platform, results showed that, the performance of theextension fitness function is better than the original fitness.And in certain circumstances, its performance improves moreobviously. By introducing the extension transformation theory,considering the extension conversion of the barriers in thespecific circumstances, this paper introduce the new methods ofexpression for the intelligent robot sensor information and thenew methods for modeling of the surrounding environment. It hasraised the awareness of the obstacle-avoiding robot to theenvironment. And thereby it has increased the adaptability of theobstacle-avoiding robot.The innovation of this paper: based on the extension geneticalgorithm, it has constructed its own obstacle avoidance robotelement model, and analyzed the extensibility. It has expanded thematter elemental transformation mode of the surroundinginformation to the obstacle-avoiding robot. And it has designedthe corresponding extension fitness function. It has reached thepurpose of evolutionary algorithm optimization. According to thenew model and algorithm, I have simulated it in the Evorobot SystemSimulation. Through analysis and evaluation of the simulationresults, I have drawn the corresponding conclusions.