Distributed learning approach for joint channel and power allocation in underlay D2D networks

Abstract

This paper investigates the problem of distributed joint channel and power allocation using game theoretic learning solutions in an underlay Device-to-Device (D2D) network where device pairs communicate directly with each other by reusing the spectrum that is being used by cellular users. This reuse, in addition to increasing the capacity, results in new cross-tier and co-tier interference cases. In order to garner the advantages of the resource reuse in the form of higher bit rates and increased energy efficiency, efficient and scalable resource allocation schemes are required. In the next generation ultra-dense networks, the focus is on distributed solutions with low computational complexity and signaling requirements. We formulate the joint channel and power allocation problem as a multi-agent learning problem with discrete strategy sets and suggest a fully distributed learning algorithm to determine the channel index and power level to be used by each device pair. The distributed joint channel and power allocation problem is formulated as an interference mitigation game, where the utility of each player is a function of its experienced expected weighted interference. We then propose a completely distributed and uncoupled stochastic learning algorithm which converges to pure strategy NE in a time-varying radio environment.