Research On Modeling And Robustness Analysis Of Immune Computation

Human immune System is very important for human health, because it is able to detect, recognize, memorize and eliminate foreign viruses and inner faults, which are sometimes unknown and even quite complex. Inspired from nature, artificial immune system is very important for computer world, because it is used to detect, recognize, learn, memorize and eliminate special objects, which are possibly unknown and even quite complex, such as computer viruses, faults and so on. However, due to incomplete theories of immunology, two bottlenecks prevent the artificial immune system from developing. First, traditional detection approaches against viruses and faults are based on matching the features of the viruses and faults, and the features of unknown viruses and unknown faults are possibly unknown, thus 100% detection is impossible in theory. Second, the faulty mechanism for detecting viruses and faults causes lower possibility for recognizing the viruses and faults, and affects extent and efficiency for repairing the damaged computer system. To overcome the bottlenecks of research on the artificial immune system and improve research on the basis of the anti-worm application and software fault diagnosis, the normal model of the system is built with the space-time properties of the components and the normal model is stored in the self memorizer. The approach for detecting selfs and non-selfs is designed with the normal model, 100% detection is realized in theory on the conditions that the data of space-time properties are correct and the operator for detecting selfs and non-selfs is normal, and the rate for recognizing the non-self and the rate for repairing the system are both increased.To build the models for immune computation, the natural computing model of immune computation is proposed and the mapping from the natural immune system to the artificial immune system is established. The normal model of the immune system is proposed. On the normal model, the tri-tier visual immune model of the natural immune system is proposed and the three tiers include innate immune tier, adaptive immune tier and immune cell tier. The normal model of the artificial immune system is proposed, and the space-time properties are used to identify the normal state of the system uniquely. On the basis of the normal model, the tri-tier immune computation model of uncertainty and limited computing is proposed for the artificial immune system. The three tiers include innate immune computing tier, adaptive immune computing tier and parallel immune computing tier. Moreover, the self/non-self detection approach that detects non-selfs by detecting selfs is proposed on the normal model, and in theory the detection rate is 100% on the conditions that the data of space-time properties are correct and the operator for detecting selfs and non-selfs is normal. The model for learning unknown non-selfs is proposed, and the features of the unknown no-selfs are learnt with those of the known non-selfs. The approach for eliminating non-selfs and the automatic approach for repairing the damaged computer system on the normal model are proposed, and this approach is useful for increasing efficiency to repair the damaged system.In designing the immune operators, the operator for building the normal model of the artificial immune system is proposed, and the data of the normal model are added into the set of space-time properties in the self memorizer. The operator for detecting selfs and non-selfs on the normal model is proposed, and the operators for recognizing known non-selfs and unknown non-selfs are proposed. The algorithm for eliminating non-selfs is designed, and the algorithm for repairing the system on the normal model is proposed.To analyze the features of the immune computation, the uncertainty features of the natural immune system and the artificial immune system are proposed, and the uncertainty feature of the artificial immune system inspires researchers to emphasize more on investigating the problems of detecting self/non-self and repairing the damaged system by itself. The computing limit of the immune computation is proposed, and the parallel computer is used to increase the load limit of the immune computation and decrease the probability that the artificial immune system destroys itself when its load exceeds the limit. The theorem of determining robustness for the artificial immune system is proposed, and on the basis of the theorem, the model and theorem of reducing robustness for the ideal distributed artificial immune system are proposed.In the real applications of the immune computation, an anti-worm immunization schema for the static web system is designed and tested in a lot of experiments. The four-element structure and natural computing architecture of immune control are proposed. The artificial immune system for detecting and diagnosing software faults are designed and tested for some mobile robots.