A Stochastic Gradient Descent Algorithm for Antenna Tilt Optimization in Cellular Networks

The electronic tilts of antennas are considered as critical parameters for the network coverage maximization. The existing approaches to optimize antenna tilts are mainly gradient-free methods. As an approximation of the original non-differentiable coverage ratio, a novel continuous coverage ratio is introduced to quantitatively represent the quality of service of network coverage in connection with both the Reference Signal Receiving Power (RSRP) and the Signal to Interference plus Noise Ratio (SINR), which offers a differentiable objective function to support the maximization based on its gradient. We thus propose a stochastic gradient descent-based antenna tilt optimization algorithm. Moreover, experiments show the proposed algorithm performs better in network coverage maximization than the traditional gradient-free algorithms.