Centralized self-optimization of eICIC with varying traffic in LTE-A

Network Quality-of-Service (QoS) in terms of capacity can be brought about by enhanced Inter-Cell Interference Coordination (eICIC). eICIC techniques are gaining interest these days due to the ability to mitigate the interference at the cell-edge of the low power nodes (typically pico cells) by blocking sub-frame transmissions of the high power nodes (macro cells). In our previous work, we had proposed a novel technique for centralized self-optimization using an iterative surrogate management framework which involves the iterative use of: 1-a surrogate to model the functional relationship between the Network Parameters (NPs) and Key Performance Indicators (KPI) and 2- a subsequent use of a pattern search algorithm for eICIC optimization. The study assumed that the traffic was not varying in intensity over the optimization period. In this work we propose to introduce: 1- a time-varying traffic eICIC optimization scenario and 2- a new optimization algorithm based on the “Expectation of Improvement” criteria within the surrogate management framework. Results using a system level network simulator reveals the practicality of the proposed optimization approach for varying traffic and also demonstrates the superiority of the Expected Improvement algorithm in terms of incorporating the uncertainty of the noisy KPI observations during the optimization process.