| Artificial endocrine system is a new subject of artificial intelligence models and methods, which is based on information processing mechanisms contained in human endocrine system. Just like nervous system, immune system, endocrine system also have highly complex and unique information processing mechanisms, so analyzing and extracting these mechanisms and establishing corresponding model have the important theoretic meaning and application value. This thesis will do some research in the modes and its applications of artificial endocrine system from different aspects.1. Research in biological endocrine system. Endocrine system is a complex regulatory system in human body, it plays an important role in the maintenance of human homeostasis. At present, theoretical studies of artificial endocrine are few, works of summarize endocrine system's characteristics of regulatory mechanisms and compare with related computational intelligence models are even fewer. From the computer information processing point of view, this thesis will tell about the basic concepts, functions and regulatory mechanisms of the endocrine system, summarize characteristics of regulatory mechanisms and compare with related computational intelligence models, all these have important theoretic meaning in research of the models and application of artificial endocrine system.2. Researches in nero-endocrine compute model and its application. Nervous system and endocrine system have interrelation and mutual effect in many ways. On the one hand, nervous system plays a role in control endocrine system, nervous system accept nerve signals from vivo environment, then stimulate endocrine system; On the other hand, the hormones which endocrine system secrete have feedback effects on the nervous system. This called neuroendocrine system in physiology. Inspired by the regulation mechanism of biological neuro-endocrine system, a neuron-endocrine computation model is proposed. In the model, the endocrine system regulates the learning and memorization behaviors of the neural system by feedback control, which could adjust the agent's behavior in time and improve its ability to learn and adapt to the unknown environment. To verify the validity of the model, we performed a simulation experiment, in which the proposed model and Q-learning with discrete states were utilized respectively to help a robot learn the ability of obstacle avoidance. Experimental results show that the proposed model is more effective than the Q-learning with discrete states. The model and algorithm are the combination of neural network and endocrine system, which reflect the regulatory mechanisms of neroendocrine system, and provides some reference for agents'behavior control.3. Researches in artificial endocrine system based on hormone's reaction-diffusion mechanism. Although the endocrine system is regulated by nervous system and other systems, but its physiological functions mainly achieve through self-organization by its internal components. Biological endocrine system is composed by large number of mutual stimulation and inhibition endocrine cells, inspired by the characteristics of its biological mechanism, a artificial network model and algorithm based on hormone's reaction-diffusion is proposed, in the model, we take the endocrine system as a complex network, endocrine cells as network's nodes, these cells secrete hormones, through hormones'reaction-diffusion mechanism, the endocrine system maintain homeostasis. To verify the validity of the model, we used it in control of multi-robots system, experimental results show that the proposed model and algorithm could induce the adaptive ability in multi-robots systems. The model and algorithm provide reference for establishment of artificial endocrine system model.In conclusion, the endocrine system contains special information processing mechanisms, analyze these mechanisms from different points of view, and create models and algorithm which have endocrine's characters have important theoretic and application value. The thesis has done some trial research on artificial endocrine model and its applications. How to create a perfect artificial endocrine model is the key point of future research. |
