Research On A Novel Swarm Intelligence Algorithm Inspired By Beans Dispersal

Many complex self-adaptive phenomena in the nature often give us inspiration. Some scholars were inspired from these natural bio-based phenomena and many nature-inspired optimization algorithms have been proposed by them. When solving some complex problems which the traditional optimization algorithm can not solve easily, the nature-inspired optimization algorithms have its own unique advantages. Inspired by the transmission mode of seeds, a novel evolutionary algorithm named Bean Optimization Algorithm (BOA) is proposed, which can be used to solve complex optimization problems by simulating the adaptive phenomenon of plants in the nature. BOA is the combination of nature evolutionary tactic and limited random search. BOA has stable robust behavior on explored tests and stands out as a promising alternative to existing optimization methods for engineering designs or applications. This thesis carries out a comprehensive research on BOA.The main work of this thesis can be introduced as follows:1. Learn from the seed dispersal mode and population distribution of evolution in nature, this thesis proposes a novel swarm intelligence algorithm-BOA. This optimization algorithm is provided with relatively new design ideas and clear meaning of bionic. Through research and study on the relevant research results of biostatistics, three distribution models of population evolution for BOA are built. These three models are respectively distribution model based on piecewise function, normal distribution model and negative binomial distribution model. Three kinds of BOA algorithms are presented based on the three distribution models respectively. In order to verify the validity of the three algorithms, function optimization experiments are carried out, which include eleven typical benchmark functions. The results of BOA in the experiments are made a comparative analysis with that of particle swarm optimization. From the results analysis, we can see that BOA is significantly better than PSO algorithm. We also research on the characters of BOA. A contrast experiment is carried out to verify the research conclusions about the relations between the algorithm parameters and its performance.2. A preliminary adaptive optimization strategy selection mechanism of BOA is proposed based on inductive reasoning. This thesis firstly lists the main parameters of the existing BOA algorithms and develops a kind of adjusting order of sub options in optimal strategy. Then the relevant parameter adjustment rules are sort out. This thesis also builds a performance evaluation method to evaluate the ability of optimization and convergence speed of the selected optimization strategy. Finally, two 500-dimensional multi-peak benchmark functions are used in the experiments to verify the validity of the adaptive optimization strategy of BOA. The experimental results show that the adaptive selection mechanism for optimal strategy can adjust the parameters to achieve a better performance than that of the BOA algorithm with fixed parameters.3. As BOA is a novel proposed optimization algorithm, theoretical analysis for the algorithm is still very preliminary at present. This thesis carries out the research on the state transfer process and the convergence behavior of BOA. Firstly several basic definitions of BOA are re-defined in strict mathematical description. Then the Markov chain model of BOA is established and the analysis of the Markov chain is made. Then I prove that BOA satisfies the two convergence conditions of the random search algorithms. Finally according to Global Search Convergence Theorem, BOA will converge to the global optimal solution with probability 1. The research conclusion is of great significance for BOA’s understanding, improvement and application.4. Algorithm application is the best way to test the algorithm. So in this thesis, BOA algorithm is applied to solve three typical optimization problems. The problem of parameter estimation for frequency-modulated (FM) sound waves is to generate a sound similar to target sound. It is a six dimensional optimization problem with the properties of highly complex multimodal and strong epistasis. It is also the No.1 problem of“Testing Evolutionary Algorithms on Real World Optimization Problems”in IEEE-CEC2011. BOA is used to solve the FM problem and the result is compared with that of DE-RHC. By comparing the results, we can see that BOA is better than DE-RHC. In China, it is important to establish a complete post-disaster reconstruction and rehabilitation system to respond to sudden natural disasters. In order to get post-disaster restoration items scheduling, we invited experts to give the fuzzy preference relation coefficients on the restoration items. Then an optimization model was proposed based on the fuzzy preference relation. BOA is proposed to solve the model and obtain ranking values of the items. The experimental result gives priority to the restoration of basic livelihood of the people and public service facilities. TSP is a typical discrete optimization problem. This thesis presents a discrete BOA for solving discrete optimization problems based on the ideas of population migration and information cross-sharing. This method overcomes the defect of basic BOA algorithms which do not suit for solving discrete optimization problems. Finanlly, five typical TSP problems are used to verify the validity of discrete BOA. The results of the experiments compared with that of the Cross-PSO and MAX-MIN AS show that discrete BOA gets better results than that of the other two algorithms and it suits for solving the discrete optimization problems. This research also deepens the BOA algorithm theory and expands the application areas of it.