Research On Artificial Immune-based Multi-Agent Systems And Its Applications

With the development of computation and communication, especially artificial intelligence, the theories and applications of Multi-Agent Systems have been one of emergent interdisciplinary research fields generated by computer science and life science. Artificial Immune Systems is a novel nature-inspired research field after Artificial Neural Network and Evolutionary Computation, and it has been successfully used in many applications. At the same time, as a typical multi-agent system, biological immune system contains abundant information processing mechanism, and this brings many inspirations for modeling and designing a novel multi-agent system.The purpose of this dissertation is to further explore the efficient combination of the complex information processing mechanism in biological immune system and current Multi-Agent Systems research, and to extend the applications of multi-agent systems based on artificial immune systems. In this paper, biological immune theories and models are firstly introduced, and some relative information processing mechanisms in biological immune system are analyzed. Then a novel multi-agent system that can be used to efficiently solve practical problems is designed by the bottom-up design processes. Finally the corresponding algorithms for solving practical problems are given in this paper, and these algorithms adequately demonstrate that the self-adaptive ability and dynamic regulation function of biological immune system can be successfully used in a multi-agent system.The main research works of this dissertation can be summarized as follows:(1)A novel immune recognition algorithm based on the clonal selection principle is proposed. This general search algorithm is based on the adaptive immune recognition mechanism, and combines affinity maturation, negative selection, immune memory, gene library evolution and meta-dynamics and other key factors in the course of clonal selection. It can automatically extract and accumulate related knowledge of the search space, and regulate the population memory and gene-library memory, and efficiently solve the question with limited resource. The experiments on STSP problems are done to prove the performance of this algorithm.(2) A novel multi-agent system model that can be used for efficiently solving practical problems is established. This model adopts the immune recognition algorithm based on the clonal selection principle as its primary control algorithm of each agent, and all agents in this model depend on the self-adaptive and strong search ability of the primary control algorithm of each agent, co-evolve, and demonstrate high intelligent problem-solving ability. This model is a novel co-evolution model. It uses multi-point random search based on the population, and all agents execute in parallel. Two complex optimization problems, K-TSP and CVRP, are used in experiments to test the performance of this model. And the experimental results prove that this model has a good performance.(3) The artificial immune network and its application to designing the routing protocols in wireless sensor networks are studied. By applying the self-organizing, self-adaptive, self-learning and other good characteristics to improve the routing solution of network nodes in the directed diffusion protocol, an efficient dynamic routing solution is established. Therefore, the wireless sensor network becomes a multi-agent system based on artificial immune systems. The results of simulation experiments prove the validity of the routing solution based on artificial immune network.