Методи керування інтелектуальних телекомунікаційних мереж

Abstract

UA: Дисертаційне дослідження присвячене розробці методів керування інтелектуальних телекомунікаційних мереж з метою підвищення ефективності їх функціонування. В дисертаційній роботі було розроблено метод вибору каналів когнітивного радіо при множинному доступі первинних та вторинних користувачів з використанням технології «ENERGY HARVESTING» під керуванням нейронної мережі, що дозволяє зменшити ймовірність помилок при виявленні вільних каналів та при передачі інформації вторинними користувачами. В роботі удосконалено метод керування середовищем когнітивної радіосистеми з використанням нейронної мережі, який дозволяє забезпечити адаптацію до змін навколишнього середовища і збільшення швидкості передачі даних. Також було розроблено метод множинного виявлення мобільних користувачів на основі оптимізації рою частинок в когнітивній радіомережі, що дозволяє підвищити ефективність виявлення мобільних користувачів інтелектуальних телекомунікаційних мереж. Розроблено алгоритм управління середовищем з використанням нейронної мережі. Алгоритм має таку особливість, що нейромережеві системи управління більш гнучко налаштовуються на реальні умови, утворюючи моделі повністю адекватні когнітивним системам. В ході досліджень було розроблено архітектуру множинного виявлення, яка відрізняється від відомих тим, що кожна локація розділена на підзони. В результаті, мобільні користувачі можуть в різних підзонах отримувати різні результати вимірювань в одному і тому ж каналі.

EN: The dissertation research is dedicated to developing of methods of management of intelligent telecommunication networks in order to increase the efficiency of their functioning. In the dissertation the researches of the level of access control to the environment, methods of management and control of the systems of cognitive radio showed that the known methods have significant disadvantages of fault tolerance, accumulation of the made decisions, mechanisms of realization of dynamic change of the environment, not fully using the spectrum, which leads to deterioration to reduce the number of customer service. Improved cognitive radio environment management method using neural network. The proposed method has the feature that the neural network is located in each base station (BS) and interacts with other WRANs in accordance with IEEE 802.22. A network environment may consist of other WRANs with which it may interact. These interactions may include data sharing and resource allocation coordination. This ensures adaptation to environmental changes and an increase in performance of 1.67 times. This established fact can be explained by the fact that such architecture more than meets the parameters set by the standard IEE802.22, in comparison with other existing methods. Neuronetic control systems are also more adaptable to real-world conditions, creating models that are completely adequate to cognitive systems. In addition, the management of the environment using the neural network is implemented on the basis of a special case of radial basis networks - probabilistic neural network. This reduces the Bayesian error probability and increases network performance. This established fact can be explained by the fact that the hybrid form of control is used. For the first time, a method for multiple detection of mobile users based on particle swarm optimization in a cognitive radio network is proposed. The proposed method differs from the known ones in that it is more correctly tuned to the real conditions inherent in the architecture of cognitive systems. This method uses particle swarm optimization under the control of a convolutional neural network. This provides rigorous access to spectrum analysis based on the energy component of mobile users. Through the use of microparticle architecture and convolutional neural layers, the detection efficiency and global particle positioning are more precisely determined. Also, this implementation provides a more realistic search space and speed of detection of users with found channels. For example, multiple detection efficiency for 50 mobile users has several convolutional layers that are generalized to each other. Such an architecture can be a confirmation that the neural network chosen in a practical way is more satisfying for the task. For example, at 20 locations, detection efficiency using the method increased by 10%, at 25 locations - by 20%, at 30 locations - by 20%, at 35 - by 20%, at 40 did not change. For the first time, a method for selecting cognitive radio channels with multiple access for primary and secondary users using "ENERGY HARVESTING" technology is proposed. To solve the problem of competition for channels among secondary users, a hybrid transmission model for one secondary user was developed. Each secondary user can transmit data in an unoccupied channel or "collect" energy from the channel in which the data is transmitted taking into account the competing set and the state of the uninterruptible power supply. The method can also provide higher bandwidth than random policy. That is, the proposed method of combined spectrum analysis and the mechanism for determining the channel selection criterion exceeds the known in terms of error probability, average bandwidth, average latency and efficiency of "collecting" the energy of the secondary user. For example, with 4 busy channels, the average waiting time decreased by one microsecond. At 10 busy channels it decreased by 20mx when compared to the combination methods and hybrid ones. An algorithm for managing the environment using a neural network has been developed. The algorithm has the feature that neural network control systems are more flexible to real-world conditions, forming models that are completely adequate to cognitive systems. In the course of the research, a multiple detection architecture was developed that differs from the known ones in that each location is divided into subzones. As a result, mobile users may receive different measurement results in different sub-bands in the same channel. A method for determining the criterion for channel selection was also developed. Developed policies can provide higher bandwidth than random policies.