| The traditional genetic algorithms (GA's) have a set of relatively perfect algorithm systems and have been applied in many optimization problems successfully, but there still exist some drawbacks such as lack of local search ability, premature convergence and random walking etc in the algorithms themselves, which lead to the weak capability of convergence. These drawbacks hamper the wide application of GA's. The study on how to improve the search capability of GA's so that they can be used in practice more effectively is an important task for many researchers all the while. Recently the study on biology shows that the immune action can prevent the premature and raise the optimizing speed effectively-thus the immune principle can give important edification on how to enhance the performance of GA's.This dissertation proposes a new and effective optimization--Immune GeneticAlgorithm (IGA) on the analysis of the drawbacks of traditional GA's and developing the useful and discarding the useless of existing immune theories. This dissertation aims to make the designed algorithm resolve the contradiction of local search capability and global search capability effectively and keep the population diversity during the evolving progress so as to remedy the demerits of traditional GA's. To make further study on the optimization capability of IGA and other correlative capabilities mathematically, stochastic theory is used to analyze IGA. As a result, the global convergence of IGA and other correlative characteristics are got. Meanwhile, the convergent effect and the mechanism on prevention from premature of IGA is studied so that the effectiveness and the superiority of the algorithm proposed in the dissertation are proved theoretically. Next, by optimization experiments on several kinds of standard test functions and the comparisons to another optimization algorithm the effectiveness and the superiority of IGA is proved in another way. In application, the problems on how to uniquely determine the kinematics inverse solution and how to constitute and simplify the optimization model using IGA are mainly considered for redundant manipulator trajectory planning, in the meanwhile, the problem of realtime optimizing the control paramatres using IGA in CSTR tracking control is also investigated. The experiments on the two examples in engineering show that IGA can be used in practice successfully and the results are satisfactory.The main contributions of this dissertation are summarized as follows:(1) By developing the useful and discarding the useless of existing immune theories and simulation of the behavior of the biological immune system, some corresponding immune optimization operators are designed, and the practical andeffective optimization algorithm (IGA) is proposed. Compared to the previous immune optimization algorithms, the algorithm in this dissertation can simulate the real immune actions such as chaotic proliferation and metadynamic function more generally and truly.(2) Stochastic theory and other correlative theories are used to analyze IGA, and the immune extend population sequence formed by IGA is proved to be an aperiodic irreducible ergodic Markov chain. Next, the global convergence of IGA is proved. Meanwhile, extend discussions are carried out on the general convergence of IGA in the pure mathematical sense, IGA is proved to be weakly convergent in the sence of probability, weakly convergent a.e., asymptotically convergent. In addition, the mechanism on the prevention from premature of IGA is studied.(3) By the optimization experiments on several kinds of standard test functions, the calculation of off-line performances and the comparisons to another optimization algorithm, the effectiveness and the superiority of IGA are got.(4) Multi-object strategy that consists of the joints' best compliance critic and the location critic is proposed to uniquely determine the inverse kinamatics solution in planar redundant manipulator trajectory planning. Next, b...
|
PDF Full Text Request