توسعه يك روش حل مسائل برنامه ريزي دو سطحي با استفاده از الگوريتم رقابت استعماري

چكيده انگليسي :

Developing a Method for Solving Bilevel Programming Problems Using Imperialist Competitive Algorithm ICA Mansoureh Moadi m moadi@in iut ac ir 2011 Department of Industrial Engineering Isfahan University of Technology Degree M Sc Language FarsiDr R Hejazi Assoc Prof Supervisor E mail rehejazi@cc iut ac irDr H Mirmohammadi Assist Prof Advisor E mail h mirmohammadi@cc iut ac irAbstractMultilevel programming techniques have been developed to solve decentralized planning problems withmultiple decision makers in hierarchal organizations the Bi Level Programming BLP problem is a specialcase of multilevel programming problems with a two level structure due to the hierarchical structure BLPPbelongs to a class of non convex and nondifferentiable global optimization problems a bi levelprogramming problem is formulated for a problem in which two Decision Makers make decisionssuccessively there are many practical cases in real world for which the bi level programming technique issuitable for modelling but rarely it has been applied to modeling these case due to the lack of effectivesolution methods there are several exact methods for solving BLPPs but they are not efficient in large scaleproblems hence the metaheuristic algorithms have been considered as an appropriate approach in largescales Imperialist Competitive Algorithm ICA as a novel metaheuristic algorithm inspired by socio political imperialistic competition has an excellent capability in several optimization problems like otherevolutionary algorithms it starts with an initial population of individuals called countries each country iseither a colony or it is an imperialist an imperialist with its colonies forms an empire There is animperialist competition among these empires during weak empires collapse and powerful ones takepossession of their colonies therefore the final outcome of this competition is a unique empire which hastaken the possession of all countries the empire is considered as the final solution of the ICA this thesispresents an imperialist competitive algorithm combined by enumerative techniques witch solves the linearbi level problem in which both objective functions are linear and the feasible region determined byconstraints is polyhedron Since the optimal solution of a linear programming can be found through thefinite set of extreme points the enumerative techniques make the ICA more effective in solving bilevelprogramming problems process the goal is optimization of the first level objective function on accessibleregion in this process finding the best solution in accessible region is modeled with the cost function ofimperialist competitive algorithm for each produced country proposed algorithm tries to produce feasibleand accessible countries taking into account In imperialist competitive algorithm in every empires theimperialist countries have the less cost function value in proposed algorithm only imperialist country areaccessible countries and colonies are feasible countries in every empire if a colony has a better costfunction value than it s imperialist it can be placed as imperialist when it is accessible for investigatingperformance of proposed algorithm the algorithm was compared with proposed genetic algorithm GA method In Proposed genetic algorithm genetic algorithm which was introduced by calvete etal 37 isdeveloped results showed that proposed algorithm is successful in solving small problems and in solvingmedium problems genetic algorithm solved more problems than proposed ICA algorithm but proposed ICAalgorithm take less time than genetic algorithm results of a variety of problems show that the algorithmprovides an effective and useful solution approach to complex problems Keywords Bi level programming Imperialist competitive algorithm simplex method Geneticalgorithm

معادي، منصوره

توسعه يك روش حل مسائل برنامه ريزي دو سطحي با استفاده از الگوريتم رقابت استعماري

روش سيمپلكس