Strategies In Swarm Intelligence Algorithm For Complex Optimization Problems With Application

Swarm intelligence algorithm is a heuristic search algorithm based on the behavior of biological groups.Its main idea is that using individuals in the group to interact and collaborate with information to achieve the goal of optimization.This algorithm has a simple structure and is easy to implement,so it can be used to solve various complex optimization problems.Particle swarm optimization(PSO)algorithm is a representative algorithm among swarm intelligence algorithms,which realizes its search on the solution space by imitating the social behavior of birds.Due to its characteristics of simple algorithm principle,easy implementation and few control parameters to be adjusted,PSO has been widely concerned by scholars and researchers in various fields at home and abroad.As an optimization tool,this algorithm has been employed to solve a variety of optimization problems in practical applications,such as path planning,feature selection,image processing,wireless communication,circuit design and so on.Quantum-behaved particle swarm optimization(QPSO)algorithm is a variant of PSO,inspired by the trajectory analysis of PSO and the basic theory of quantum mechanics.Compared to PSO,QPSO has no velocity vectors for particles,has simple framework and needs fewer parameters to adjust.Besides,its ability of global searching is better.Nevertheless,the QPSO algorithm still succumbs to the issue of premature convergence,just like PSO and other PSO variants.In particular,when solving complex optimization problems,it converges too fast and the swarm diversity declines rapidly,which makes it difficult for particles to get rid of the local optima area.Considering this issue,this dissertation designs a series of strategies for the QPSO algorithm,in order to enhance its performance when dealing with complex optimization problems,including the single-objective optimization problems and the multi-objective optimization problems.Meanwhile,the improved versions of QPSO proposed in this dissertation are applied to solve the protein-ligand docking problems,and obtain a good docking performance.The main research work is summarized as follows:Firstly,an entropy-based QPSO algorithm is proposed.The concept of entropy is introduced into the QPSO algorithm,and then particles in the swarm are accordingly classified into two categories,i.e.,the crowded particles and the uncrowded ones.Different particles are updated according to different rules.The purpose of this strategy is to improve the swarm diversity,and avoid falling into the local best area,especially at the later stage of the search process.The experimental results on benchmark functions show that this proposed algorithm is effective in solving complex optimization problems and that it has an advantage over its competitors.Secondly,an improved QPSO based on dynamic grouping searching strategy is proposed.The particle swarm is randomly divided into two groups at the beginning of the search process,and the local attractors for particles in each group are generated according to different rules,making the particles in two groups focus on exploration and exploitation,respectively,and thus achieving a good balance between the global search ability and the local search ability.Meanwhile,an opposition-based computation is employed to improve the swarm diversity and to help particles escape from the local optima.The experimental results on benchmark functions show that this proposed algorithm is efficient in solving complex optimization problems and that it has better searching performance than its competitors.Thirdly,a decomposition-based multi-objective QPSO algorithm is proposed.In this dissertation,the QPSO algorithm is integrated into the framework of the multi-objective evolutionary algorithm based on decomposition(MOEA/D),which uses the PBI(Penalty-based Boundary Intersection)approach to decompose the multi-objective optimization problem into a set of single-objective subproblems and optimizes them simultaneously.A diversity measuring mechanism is employed in this algorithm to avoid the premature convergence,for the purpose of making particles escape from the local best area with a high probability.Meanwhile,a number of non-dominated solutions are introduced to generate the global best for guiding particles in the swarm.The experimental results on benchmark functions show that the proposed algorithm could achieve a good search performance on most of the complex multiobjective optimization problems.Lastly,considering the characteristics of the problems,only the entropy-based QPSO and the dynamic grouping searching based QPSO algorithms are employed to solve the proteinligand docking problems.These two algorithms are combined with the Solis & Wets local search method,and then the corresponding hybrid algorithms are applied in the docking environment of the latest version of Auto Dock software in order to enhance the docking accuracy and the docking efficiency of the molecular docking software.The experimental results show that both of the two algorithms can be used for solving the protein-ligand docking problems and that their performance are better than their competitors on most of the test cases,especially on highly flexible ligand docking problems.In summary,this dissertation focuses on QPSO algorithm,a variant of PSO,and designs a series of strategies to improve the search performance of QPSO when solving complex optimization problems.What’s more,two single-objective QPSO algorithms are further applied to the protein-ligand docking problems in this dissertation,providing some new solutions.Therefore,it can be concluded that the research work of this dissertation has certain academic value and application value.