Dynamic Resource Allocation and Balanced Cell Loading - a Stochastic Meanfield Control Approach

摘要

Mean-field theory is a significant recent step for the field of stochastic optimal control. By allowing the optimal control functions to take into account not only the state of the controlled node, but also the mean-field state of an entire ensemble of nodes, mean-field theory allows us to model interdependent networks of agents in an analytically tractable manner. In this paper, we show its application to a very standard problem of cellular network optimization, the cell loading problem. By modelling the cell-loading problem as a combination of the loading of the individual cell, as well as the loading of the entire network, we show that a distributed optimal control function exists that can be individually implemented at nodes, and that is capable of reaching network wide equilibrium.