A deep learning based energy efficient downlink power control mechanism for cellular networks

Management of radio resources in wireless communication has always been a challenging task. The complexity of this resource management is high, because wireless channels always contribute to different interference levels which ultimately results in degradation of the quality of service (QoS). To combat these challenges, many power control algorithms are developed. In this paper, we propose a deep learning (DL) based mechanism for power controlling. A convolutional time-series prediction model is developed which predicts future signal-to-noise-to-interference-ratio (SINR) and allocates power to maintain minimum required SINR level subject to overall amount of power consumed remaining minimum. The results generated by this method are benchmarked with greedy iterative SINR target setting power control technique and the result shows significant improvement in total power consumption and energy efficiency (EE).