Application of Game Theory for Load Balancing in Long Term Evolution Networks

Abstract

Game theory offers a set of effective tools to be applied in autonomous and distributed self-organizing networks. A typical use case is load balancing which aims at increasing the overall network capacity in case of unequal traffic distribution. The 3GPP Long Term Evolution (LTE) standard provides the means that enables the handover of users from highly loaded cells to the lower loaded neighbors. The method is based on exchanging load information between neighboring cells. However, the computation of the amount of load that a cell should either offload or accept is running autonomously in each cell and ismost likely not generally specified, but rather vendor specific. In this case, a network-wide algorithm for load balancing may not be possible to use if eNodeBs from distinct vendors are deployed. In this game-theoretic analysis for load balancing, we consider each cell as a rational player that decides, in the worst-case, on the amount of load that maximizes its payoff in an uncooperative way. The simulation results for LTE network show that the resulting Nash equilibrium is able to achieve most of the gain expected from a strictly cooperative load balancing scheme. Though each cell acts independently, the Nash equilibrium almost provides the sameperformanceof a network-wide algorithm for load balancingwhichwould ease the players to decide on strategies that aremore collaborative. The capacity gain of theNash equilibrium is verified for the homogeneous network layout, different scenarios and parameter configurations. Moreover, to take real network effects, such as different cell sizes and number of neighbor cells into consideration, the Nash equilibrium is also tested in the heterogeneous network layout. Index Terms – Game theory, Load balancing, Self optimizing network, Distributed systems.