Immune Optimization Algorithms In Complex Static Or Noisy Environments And Their Applications

The biological immune system is a parallel, distributed and adaptive information processing system with highly intelligent performance that plays an important role in solving engineering problems. This results that it has become a research focus to exploit new optimization algorithms based on the immune system. Under this background, this dissertation proposes a suite of immune optimization algorithms to deal with optimization problems in complex static or noisy environments respectively. Through comparisons with numerical experiments and practical applications, the results illustrate their validity and feasibility. The acquired achievements of this dissertation are summed as follows:A. An improved immune optimization algorithm, suitable for static single-objective optimization with high dimensions, is proposed. For the algorithm, relying on the idea of parallel coevolution, any evolving population is partitioned into many sub-populations. Each of which is required to evolve independently in terms of a simple immune evolving mechanism; further, all these sub-populations are intercommunicated periodically so as to achieve the purpose of global optimization. This algorithm is applied to optimize the weights of one neural network, while successfully dealing with the problem of optimal control for micro hard disk servo dual-stage control systems. The comparative experiments show that this algorithm is of great potential for practical applications.B. A novel adaptive sampling immune optimization algorithm with less parameters and availability is proposed to tackle with single-objective optimization problems in noisy environments. In design of the algorithm, one novel adaptive sampling scheme is designed to enhance its performance efficiency, while one simple immune evolving mechanism is developped based on immune metaphors. Through comparison with other algorithms and the practical application to the problem of system identification for the micro hard disk servo system, the experimental results show that the algorithm can achieve satisfactory behaviors beween performance effect and performance efficiency.C. A multiobjective immune optimization algorithm in noisy environments, suitable for multiobjective optimization problems with noise, is proposed. In the algorithm, introduce T cells to adjust the evolving directions of antibodies from the current evovling population, and attach a life-time to each of the antibodies. Besides, a novel comparative strategy between antibodies is designed based on the concept of probability dominance. The comparative experiments show that this algorithm can rapidly obtain satisfactory effect in different noisy environments.D. A multiobjective evolutionary algorithm for complex static multiobjective optimization problems with high dimensions is proposed. The key of design of the algorithm is to construct three operations: dynamical encoding strengthening the capability of global and local search, individual evaluation improving population diversity, as well as population isolating and adaptive mutation rules of gene segments. Compared to several representative multiobjective evolutionary algorithms, the proposed algorithm is of great potential for extremely difficult multiobjective optimization problems with high dimensions.