A Fast Coordinate Descent Algorithm Improving both Coverage and Capacity in Cellular Networks

It is an essential task for mobile operators to extend the coverage and improve the capacity by adjusting the radio parameters in cellular networks, especially in self-organizing networks. In the existing algorithm, the decoration of target local performance is neglected, which severely deteriorates the algorithm efficiency. In this paper, we propose a hierarchy-induced coordinate descent algorithm, which focuses on the local decoration of the target function to solve the coverage and capacity optimization problems by obtaining the optimal antenna tilt and azimuth settings in cellular networks. Such algorithm alternatively executes the optimization step decision where the original complex constrained optimization problem is converted into a non-constrained simple one with the penalty term and the optimization variable selection that balances the influences of the hierarchy and the distance according to different system parameter settings. Besides, in the optimization step decision mentioned above, we adopt the non-linear interpolation to obtain the optimal optimization direction and its corresponding length.