Immune-Genetic Algorithm for Traveling Salesman Problem

Abstract

Traveling Salesman Problem, Theory and Applications 82 applications in engineering optimization.However, GA also has some significant drawbacks, for instance, the pre-mature convergence of computations, the poor use of system information during computational evolutions, expensive computation from evolutional procedures, and the poor capability of local search.The immune system, which is made up of special organs, tissues, cells and proteins, is the body's defence against infectious organisms and other invaders (Liu, 2009).The immune system detects and attacks antigens that invade the body through different types of lymphocytes.Artificial immune systems are adaptive systems inspired by the functions, principals and models of the vertebrate immune system.When artificial immune systems are attacked, the immune mechanisms are started to guarantee the basic functions of the whole intelligent information system.Researches on artificial immune systems aim to set up engineering models, algorithms and advanced intelligent information system through intensive study on the information processing mechanisms of biological immune systems.In the 1970s, Jerne first propounded the hypothesis of the immune network system and founded the basic theories of the artificial immune systems, Jerne's idiotypic network model.Perelson studied on a number of theoretical immune network models proposed to describe the maintenance of immune memory, which accelerated the development of artificial immune systems in computer science.In 1986, Farmer built a dynamic model of the immune system and brought in the concept of learning.Farmer's work contributed to turning artificial immune systems to practical application.One important aspect of the research on artificial immune systems is to develop effective learning and optimization algorithms.Immune algorithms are one of heuristic search algorithms inspired by immune principals.In 1990, Bersini put immune algorithms into practice for the first time.By the end of the 20th century, Forrest et al. started to apply immune algorithms to computer security field.At the same time, Hunt et al. began to use immune algorithms in machine learning.Immune algorithms (IAs), mainly simulate the idea of antigen processing, including antibody production, auto-body tolerance, clonal expansion, immune memory and so on.The key is the system's protection, shielding and learning control of the attacked part by invaders.There are two ways considered to design an immune algorithm: one is to abstract the structure and function of the biological immune system to computational systems, simulating immunology using computational and mathematical models; the other is to consider whether the output of the artificial immune systems is similar with that of the biological immune system when the two systems have similar invaders.The latter doesn't focus on the direct simulation of the process, but the data analysis of the immune algorithm.As the immune system is closely related to the evolutionary mechanism, the evolutionary computation is often used to solve the optimization problem in immune algorithms.The research on artificial immune systems lays a foundation for further study on engineering optimization problems.On the one hand, it aims to build a computer model of biological immune system, which contributes to the study of the immune system operation.On the other hand, it supplies an effective way to solve many practical problems.Immune Algorithms inspired by biological immune mechanism, can make full use of the best individuals and the information of the system, and keeps the diversity of the population.In the optimization process, Immune Algorithms take the useful ideas of existing optimization algorithms, combine random search with deterministic changes, reduce the impact of the random factors to the algorithm itself and can better eliminate the premature convergence and oscillation.www.intechopen.com