A neural network based resource allocation scheme for 4G LTE heterogeneous network

Long Term Evolution network or LTE network is a 4G network extension that provides higher data rate by combining multiple technologies and channels through a demand based selection of the appropriate one. This implements an evolved node base station (eNB). The Integrated eNB's in 4G network are responsible for seamless switching between different technologies and channels. Such a switching decision is often affected by several parameters like available bandwidth, desired load, inter channel interference, available signal to noise ratio, fading, downlink power and so on. Regular switching between channels result in waste of a portion of bandwidth in control signal whereas delayed switching results in low packet delivery ratio and in turn low data rate. Hence, an optimal solution is desired to maintain high data rate under scalable 4G network. In this work, we propose an optimal resource allocation and power allocation using a Feed Forward Neural Network based machine learning framework for channel condition classification. The system is trained with probe packets under controlled conditions. The eNB nodes then manage the channel resources based on resource constraints during communication of performance parameters. Results shows that the proposed system improves effective data rate and packet delivery ratio in comparison with conventional linear techniques.