چكيده به لاتين
Due to the ever-increasing growth of the number of users, the amount of traffic and the services required by users, the use of new generations of telecommunication networks is unavoidable in the near future. The next generation of communication networks is constantly evolving with the aim of maximizing the user experience, supporting the huge volume of mobile traffic and managing network resources. Due to the complexity and heterogeneity of mobile environments, the realization of this is faced with challenges such as resource management and simultaneous allocation of different resources, energy efficiency and multiple access. In addition, due to the limited resources, intelligent mechanisms are needed to share resources. Based on this, due to the limitation of radio resources (bandwidth and power), the issue of optimal allocation of resources from the point of view of economic efficiency has been specially noticed by mobile operators. In the fifth generation of communication networks (5G), the allocation of network resources is done using slicing technology according to service requirements. The use of network slicing provides different services in a common communication infrastructure by telecommunication network operators and increases network flexibility. In order to allocate resources, one of the common access techniques used in different generations of mobile networks is the separation of resources through time, frequency or code, known as orthogonal multiple access (OMA) techniques. In addition to this access technique, in 5G, the non-orthogonal multiple access (NOMA) technique has been proposed in order to support more than one user, through the non-orthogonal allocation of resources due to the need to increase spectral efficiency and support mass connections. Also, due to the progress of computing capabilities, improvement of machine learning algorithms and increase in the amount of available data, machine learning and its branches will play a very important role in resource allocation. Using these methods will increase efficiency in terms of calculation and performance. In this research, the combination of OMA and NOMA has been used in the fifth generation of communication networks in terms of network segmentation. For this purpose, the problem of resource allocation with the aim of simultaneously maximizing energy efficiency and throughput in 5G networks with hybrid access has been modeled as a multi-objective optimization problem and has been solved using CVX tools and machine learning-based approaches. The results obtained from the evaluations indicate the optimal allocation of resources and the achievement of the goal of the problem. Also, the evaluations in the machine learning section show that RNN-LSTM and Dense-Keras Tuner models have achieved losses of 3.7 and 3.1 in 200 epochs, respectively.
