Research On Key Technology For Performance Optimization Of Cognitive-Mechanism-based Ad Hoc Network

With the continuous development of computer network technology, wireless network communication technology is constantly changing the style of people’s life and communication. Aiming at improving on the traditional wireless communication technology, researchers have developed a new type of wireless communication network architecture, which is represented by Ad Hoc network. Ad Hoc network is a centralized self-organizing wireless multi-hop network. Different from traditional wireless communication networks, Ad Hoc networks do not rely on infrastructure, all nodes in the network is responding as users, whose node functions depends on the needs of location and path. This architecture is of a high degree of flexibility and autonomy, brings the network management and maintenance activities complex and interesting.Currently both at home and abroad, Ad Hoc network, especially the research study in autonomy characteristics, is the hot spots of academia of wireless communication networks. Possessing a variety of network performance standards and variable parameters, together with the typical characteristics of complex networks, it’s difficult for selecting an ideal style of network operation mode for Ad Hoc network responding to environmental changes with great autonomy. Traditional Ad Hoc networks, which perpetually using Event-Condition-Act response model for network management and optimization, is somehow of a passive characteristic. The current Cognitive Science and Cognitive Mechanism related to computer network communications applications, provide possibilities to inspire a bran-new approach for the network management activities.By studying the Cognitive Mechanism based Ad Hoc network, we can effectively improve the network performance. The so-called Cognitive Mechanism based Ad Hoc network is a type of Ad Hoc network. It would dynamically adapt the complex end to end network conditions for performance optimization through learning and reasoning by the collection of parameters. In this scenario, the users take out the policy on overlap network level objective rather than a single network device. The key issue of this thesis is to introduce the Cognitive Mechanism into Ad Hoc Networks. The main research step of this study is as follows:(1) Based on existing research results on Cognitive Mechanism and cognitive processes, this thesis presents a Cognitive Mechanism functional framework for Ad Hoc network. According to the characteristics of distribution and flexibility of Ad Hoc network, and also the power consumption constraints, we propose a distributionization program for the Cognitive Mechanism functional framework and a distributed policy management system for Ad Hoc network. Finally, a collision avoidance strategy for experimental cases is brought to verify the effectiveness of related work.(2) According to the Game Theory Model and Multiple Objective Optimization theory, this thesis proposes research on formalization of Ad Hoc network, and describes the relationship between the Ad Hoc network model and game theory model. Concerning the mechanism of cognitive processes in the network model on this basis, we propose the concept of decision-making behavior characteristics, define three types of decision-making behavior characteristics, which are selfishness, ignorance and control, and by that we propose the theory of characteristic value for decision-making behavior in the cognitive processes as the basis for evaluation and decision-making.(3) We add some constraints on formal recognition mechanism based Ad Hoc networks, and constrain the formal network model into the Potential Game Model and Quasi-Concave Class Game Model for optimization based model. Combined with game theory, we prove these two models are convergence which can be optimized by three kinds of convergence methods——Uryasev, Gabay and Yates. Further more, we bring out the way of calculation for the characteristic behavior vector of the convergence process, or the best response decision-making vector sequence. Finally, with the basis of research on the convergence process, we put forward the research and evaluation on the characteristics value of decision-making behavior on the specific model.(4) In connection with general issues about life cycle optimization for Multicast Tree in Ad Hoc Network, by the Cognitive Mechanism and game model we proposed, and also energy efficiency, directed receiving antenna, Signal to Interference plus Noise Ratio requirements we introduced for the research work on optimization for Multicast Routing life cycle, we propose a distributed solution based on Cognitive Mechanism, and also bring out a type of comparable quantitative evaluation for the solution with Mixed Integer Linear Programs Model. Finally, we evaluate decision-making behavior characteristics in the solution with the theory of value of decision-making behavior characteristic, and reveal the significance of all threes kinds of decision-making behavior characteristics in the decision-making solution. In summary, this thesis targets the performance optimization for Ad Hoc network, introduces the concept of Cognitive Mechanism to the network model, investigates deeply on the formalization issue of Ad Hoc Networks model and the issue of solving network game model with constraints, and describes the solution process on behavior vector for the network model to turn from any initial state to the best state, and proves the network model is convergence and can be optimized, finally gives the relevant experiments and cases.