Wlan Indoor Positioning Algorithm Based On Genetic Neutral Network

As the development of location-aware service for whole space in the future, the indoor positioning technology of wireless LAN, as an important component of this system, will gradually be paid more attention and favor. However, domestic and international research on the indoor location is yet not well-rounded, and lots of key technology and core issues are still in its infancy. From initial distance dependent algorithms to pattern matched indoor positioning algorithms, which including three artificial intelligence algorithms, that is, artificial neural networks, fuzzy inference systems and genetic algorithms, advantages of implementing wireless LAN to communicate in office buildings, shopping malls, apartments and other indoor environment are more and more obvious. With the diversity of network topology and high positioning accuracy, artificial intelligence algorithms, which take into account of complex network structure and dynamic environment characteristics, become a key algorithm for wireless LAN indoor location. Therefore, how to achieve effective integration of artificial intelligence positioning algorithms to make up for deficiencies is a significant issue for indoor location system.Based on defects of artificial neutral network, slow training rate, easy to fall into local minima and weak global search capability, by calculation of generalization capability, some related questions about design process of genetic algorithm optimized artificial neutral network are analyzed in this paper.Firstly, the indoor location math model of genetic algorithm optimized artificial neutral network is established. Because slow convergence of traditional back propagation artificial neural network, vulnerable to trap into local extreme point for larger search space, multi-peak and non-differentiable function, and lack of basis for determining initial weights, threshold, and the network structures, indoor positioning results are not stable enough. Therefore, in this section, based on the analysis of artificial neural network model and the indoor location algorithm, the design of genetic algorithm optimized artificial neutral network are proposed. Also, the model with hidden layer nodes optimized is established.Secondly, generalization capability of genetic algorithm optimized artificial neutral network are measured.Based on learning rules of neutral network and calculation of generalization capability, measuring method which adapted to dramatic indoor environment is proposed. According to the analysis of advantages and disadvantages of common coding types currently applyed to genetic algorithm, prove that by network ensuring criterion of generalization capability, selecting a most appropriate genetic coding type is feasible. Finally, generalization capability ensured artificial neutral network positioning algorithm which optimized by genetic algorithm is analysed. To begin with, implement genetic algorithm to make artificial neutral network equipped with self-evolved and adaptive capabilities, and construct a network with connectting weights, structure, and learing rules evolved. Then according to diversify discussion of genetic parameters and generalization capability ensured selection of coding types, also taking account of genetic fitness as measuring function, discuss selecting criteria for different reference point intervals. By theoretical analysis and simulation, priciples of network parameters, structure and network design are determined, and indoor positioning algorithm are also optimized.According to the analysis of generalization capability of genetic algorithm optimized artificial neutral network, the design method based on genetic coding type optimized artificial neutral network with guarantee of generalization capability is proposed. And this new method will be helpful to the efficient management and optimal network design of the future WLAN indoor location technology.