Spectrum And Power Resource Allocation Based On Machine Learning In Two-tier Femtocell Networks

In recent years, with the development of internet technology, people prefer to enjoy a variety of internet services at home, thus improving indoor wireless coverage has become very important. Femtocell, also known as the home base station, is a kind of a small base station apparatus which is used to solve the indoor coverage problem, it is deployed by the user in an indoor environment, connected to the operator’s core network through IP technology, thereby to provide users with high-rate indoors data services. The cost of femtocell is very low, the installation and maintenance is very easy, and femtocell has many other advantages such as low power consumption, good coverage and so on, so it’s the best solution to improve the quality of indoor communication. Machine learning, through the study of computer simulation of human learning ability to realize intelligent and automatic control of equipment, using machine learning methods to solve the spectrum and power resource allocation problem in femtocell two-layer network, make the system more intelligent, is an important direction in femtocell application fields.Firstly, a two-layer Femtocell network architecture which consists of macrocell and several femtocells is studied in this thesis, meanwhile the principle of femtocell is introduced and the interruption problem is analyzed. Then, two typical methods of machine learning are used for spectrum and power resource allocation: ant colony optimization and Q-learning. Ant colony optimization can dynamic sense the spectrum occupancy status of system, and it is used for spectrum resources allocation, as a result, the interference is effectively avoided; Q-learning is used for automatic femtocell transmit power control, and it can reduce interference while ensuring the communication capacity of femtocell and macrocell. Finally, these two machine learning methods are simulated based on theoretical derivation, and the application result shows that systems can intelligently allocate spectrum and power resources based on the current status of environment, and greatly reduces the co-layer interference and cross-layer interference to improve the performance of system.