Energy Efficient QoS Aware Machine Learning Model for Scheduling Users in NOMA Heterogeneous Networks

The massive growth in mobile devices and machine type communication devices which demands higher performance leads to higher data traffic and spectrum scarcity problem. Multiplexing of time frequency components with code division enhances better capacity in data transmission that are best suited for Non-orthogonal multiple access (NOMA). Deployment of NOMA helps to explode data traffic in a heterogeneous network and that plays an important model for next generation wireless networks. Joint optimization problems over the channel assignment, user decoding, and allocation of power are formulated to maximize system throughput. Not only throughput, achievable sum rate is also maximized at power allocation. Mixed-integer non-linear problem are resolved for continuous variables as an optimization sub-problem (P1) and integer variables as a matching sub-problem (P2). A power control scheme is focused for resource allocation policy to improvise average performance of sum rates with aid of reinforcement learning. With this keyhole, approaches are addressed with two issues for allocation of resource in NOMA namely dynamic user allocation and resource blocks and network traffic balancing. Results validate the proposed scheme can significantly enhance user sum data rates and thus utilities are compared with known Q-learning based strategy.