Self-Organized Resource Allocation for LTE Pico Cells: A Reinforcement Learning Approach

摘要

This articles proposes a smart resource sharing algorithm to manage interference in LTE networks. Relying on reinforcement learning theory, the proposed Cyclic Multi Armed Bandit (CMAB) algorithm steers each cell in the choice of the most suitable frequency band portions, in autonomous manner. Thanks to the traffic aware feature, the algorithm adapts as well to the real needs of the cell. Adding to that, a refinement of the decision function is proposed to speed up the convergence time. The proposed approach is tested in LTE compliant simulator and the results shows its efficiency compared to conventional static reuse schemes.