Improved Immune Clone Selection Algorithms And Its Applications

There exist many optimization problems in various fields of science and engineering. Therefore optimization methods have high value both theoretically and practically. Traditional deterministic optimization methods have many limitations. They are not adequate to solve the increasingly complicated problems in today’s society. In recent years, multidisciplinary research provides new ideas to solve such problems. Artificial immune optimization algorithm based on the biological immune mechanism has shown excellent performance in various fields. And it has become a powerful tool for solving complex optimization problems.Immune clone selection algorithm is a new intelligent optimization algorithm inspired by clone selection theory in the biological immune system. The algorithm combines the prior knowledge of the problem and the adaptive abilities of biological immune system. Therefore it is more robust in information processing, as well as faster convergence to the global optimal solution for searching. This dissertation summarizes the theory and characteristics of the simple immune clone selection algorithm, analyzing its shortcomings, improving the algorithm by summing up various immunological and evolutionary ideas, and applying the improved algorithm to several typical optimization problems. Simulation examples verify the effectiveness and practical value of the improved algorithms. The main contents and conclusions of the dissertation are as follows:(1) The biological principles and its bionic mechanism of artificial immune system are thoroughly analyzed. The specific research content and scope of the artificial immune system are elaborated in details. The mechanism of the clone selection theory is studied deeply.?? The drawbacks of simple immune clone selection algorithm in some areas are summarized. The fundamental research directions and its improvements are established.(2) Aiming at the slow population convergence and easy to fall into local optimum problems of simple immune clone selection algorithm, an immune dominance clone selection algorithm (IDCSA) is proposed which is inspired by the characteristics of the dominant gene in biological immune system. The proposed strategy strengthens the exchange of information among the antibodies in population and effectively leads the antibodies searching on the future direction, through improving the entire population in the iteration via making some antibodies with high fitness value. The strategy not only maintains diversity of the population, but also ensures the quality of solutions. Adopting the random mutation strategy with exponential distribution, the algorithm fully searches solution space to jump out of local optimum. Vehicle routing problems in logistics are tested to show the effectiveness of the improved algorithm. Different Benchmark problems in logistics are simulated, summarized, and compared to verify the validity of improved algorithm.(3) Aiming at the early convergence problem of the simple immune clone selection algorithm in a single population, a master-slave immune clone selection algorithm (MSICSA) based on a multi groups strategy to strengthen parallel search capability of the immune clone selection algorithm. The algorithm has a master-slave structure with a master population on top layer and multiple populations on the bottom layer. There are immigration and emigration operations between the mater and the slave populations to enhance the exchange of information among different populations, so as to improve the quality of the master population. Chaotic sequences are adopted as random number of the mutation to enhance the randomness of the search. The master-slave immune clone selection algorithm is used for the task assignment problem with detailed the solving procedures in the specific application example. Comparisons of the improved algorithm with other algorithms in the example show the effectiveness of the algrithm.(4) An adaptive global immune clonal selection algorithm(AGICSA) is proposed to overcome the shortcomings of fast convergent rate in early phase of optimization and slow convergent rate in final phase, and the lack of randomness and local search capabilities of simple immune clone selection algorithm. A global adaptive mutation strategy of Gaussian distribution is introduced to AGICSA. A normal cloud model related to fitness of antibodies is introduced for mutant rate. With dynamic adjusting the intensity of mutation, the algorithm introduces adaptive mutation operation of Gaussian distribution, which performs with high probability globally, uniformly and dynamically in σ neighbor of antibodies meeting the mutant rate, so as to strengthen randomness and stable orientation of the searching. In the specific application example, a Fuzzy Energy Management Controller(FEMC) for Electric Assist Control Strategy of Parallel Hybrid Electric Vehicles(PHEV) is used. AGICSA is used to optimize the membership function of Fuzzy Logic Controller(FLC), which further improves fuel economy and comprehensive energy consumption index. The improved algorithm and other algorithms are embedded in PHEV model respectively for different test road cycles. Simulations show its effectiveness.The improved immune clone selection algorithm is comprehensively studies through the above research work and simulation results. Improvements are made for handling large-scale complex optimization problem, which are implemented and simulated. At last, the issues for study further on the development of immune clone selection algorithm are discussed.