Multi-objective Immune Algorithm And Its Application On Cloud Workflow Scheduling

Optimization problems are widely encountered in scientific research and engineering applications.Thereinto,multi-objective optimization problem is regarded as one kind of difficult problem due to its objective confliction,and has received more and more attentions.The early works had tried to use the methods like linear weighted,objective programming and Max-Min to solve multi-objective optimization problems.They need lots of prior knowledge and are easily affected by Pareto front shape,and they also show high computation complexity.All these methods only obtain one optimal solution in a single run.Evolutionary algorithms become popular in solving multi-objective optimization problems,as they can generate a set of Pareto-optimal solutions in a single run and don't have the defects mentioned above.On the other hand,inspired from the biologic immune system,multi-objective immune algorithm is a new kind of algorithms that extended from evolutionary algorithms.Compared to evolutionary algorithms,multi-objective immune algorithms show better convergence ability and high search efficiency when solving the multi-objective problems.The main works in this paper are on the basis of the clonal-selection-based multi-objective immune algorithm.In this paper,after detailed analyzing the merits and drawbacks of multiobjective immune algorithms,an improved multi-objective immune algorithm is proposed.Furthermore,multi-objective immune algorithm is applied to solve cloud workflow scheduling problems and an immune-based multi-objective cloud workflow algorithm is designed.The works of this paper are summarized as follow.Firstly,in order to remedy the disadvantage of lacking population diversity in multiobjective immune algorithms when dealing with complex problems,this paper proposes an adaptive immune-inspired multi-objective algorithm with multiple DE strategies,called AIMA.The main contribution of AIMA is that multiple DE strategies are combined with the basic framework of multi-objective immune algorithm and an adaptive DE strategy selection operator is designed.The strong search abilities of DEs help to improve the population diversity and enhance the robustness of AIMA when tackling different kinds of multi-objective optimization problems.In addition,a hybrid selection operator,which simply combines the clonal selection and the non-dominant sorting and crowding distance based selection method in NSGA-II,is also proposed in this paper to further enhance the population diversity.Secondly,this paper applies multi-objective immune algorithm to solve cloud workflow scheduling problems and proposes an immune-based multi-objective cloud workflow scheduling algorithm,called IMCWS.IMCWS is designed according to the previous works in EMS-C.It replaces the framework of evolutionary algorithm in EMS-C with the framework of multi-objective immune algorithm,which helps to speed up convergence.Meanwhile,this paper improves the crossover operator in EMS-C and proposes a novel crossover operator with stronger search ability.The new crossover operator can improve the convergence speed and population diversity simultaneously.