Research On The Key Technologies Of Self-Optimization And Self-Healing In The LTE Network

As the advanced development of the third generation mobile communications system, LTE (Long Term Evolution) can provide the greater bandwidth, lower latency and the larger capacity. Moreover, with the widespread deployment of LTE networks, both the mass and the complex attributes of wireless network parameters are becoming much more serious, which leads to the failure of traditional optimization methods. Therefore, to reduce the costs as well as improve the efficiency of operation and maintenance simultaneously, major operators and research institutions have already initiated the study of SON (self-organizing network technology) in3GPP (3rd Generation Project Partnership). In fact, SON has some autonomous intelligence functions inclusive of self-configuration, self-optimization and self-healing. Herein, self-optimization function includes many use cases, such as capacity and coverage optimization, automatic neighbour relation function, energy saving, load balancing, random access channel selection and so on. Furthermore, handover self-optimization is one of the key technologies of SON. While COC (Cell Outage Compensation), a key self-healing technology of SON, is also becoming the focus of attention.Based on the analysis of both characteristics and optimization strategies of LTE networks, this paper studies the key handover self-optimizing and cell outage technologies of LTE networks in-depth in terms of the intellectuality and adaptivity of the optimization of communications networks with those methods related to pattern recognition, data mining and reinforcement learning theory. Moreover, we also proposed some handover self-optimizing, cell outage detection and compensation algorithms, and verified them by both theoretical analysis and simulations. The main innovative contributions are listed as follows.1. A new methodology based on Support Vector Machine (SVM) for the detection of handover-related radio link failures is proposed. Since the abnormal handover detection is regarded as a pattern recognition process, we first extract the abnormal handover features by setting the time points during the handover process, and then train and test data that are generated using these features. Finally, based on these data obtained before, the classification performance of the SVM-AID algorithm is tested and the effects of the parameters in SVM on the classification are analyzed. The experimental results show that the parameters should be chosen carefully because they have great effects on the classification.2. A dynamic self-optimization algorithm for handover (HO) parameters using the Q-Learning method is proposed. We introduce the enhanced Q-learning algorithm to the process of handover optimization to determine the optimization parameters, establish rewards and punishments Q-learning function based on call drop rate and ping-pong handover and construct Q-learning handover optimization system, making the whole process of self-parameter optimization have learning and adaptation capabilities. Simulation results show that this method can effectively optimize handover parameters improve the network performance by reducing the call drop rate from73%to23%.3. A cell outage detection method based on differential evolution algorithm and BP (Back Propagation) neural network cell is proposed. After selecting several characteristic parameters related to the outage cell as the input of the detection algorithm, we use the differential evolution algorithm for training the neural network to determine the thresholds and weights of BP neural network model so as to detect the outage cells. The proposed method combining the advantages of both differential evolution algorithm and neural network overcomes the disadvantages of BP network being easily trapped in local minimum and slow convergence. Experimental results show that the proposed detecting method can achieve higher detection accuracy (i.e., basically can reach more than90percent) than the traditional BP neural network.4. A self-organized approach for cell outage compensation through jointly adjusting both the tilt and transmit power of antennas based on fuzzy Q-learning is proposed. We first propose a cell selection mechanism to choose part of cells rather than all the cells to carry on compensation. Our strategy mainly considers the radiated power level and the load condition of the neighbor cells. Considering the poor performance of the traditional Q-learning in continuous state and action space, we combine the fuzzy theory with the traditional Q-learning to jointly adjust the transmit power and tilt of antennas to further compensate those outage cells. Moreover, we also design a reward function considering both the whole spectrum efficiency and compensation effect. Simulation results show that our approach can effectively alleviate the degradation of network performance in the outage cells.Finally, we summarized the whole paper by pointing out some problems and weaknesses of our research work, and gave the direction of our further research.