Optimal Cross-Tier Power Allocation for D2D Multi-Cell Networks

Abstract

Efficient transmission power control is indispensable for cellular networks. It not only provides a high energy efficiency, but also maintains reliable connections. With the emergence of 5G mobile technology, the presence of device-to-device (D2D) communications within the cellular network has stimulated research on radio resource sharing. In this paper, we consider an underlay D2D network operating in a Rayleigh fading channel and propose a power allocation method that assigns transmit power levels to D2D UEs (DUEs) and cellular UEs (CUEs) such that the joint connection probability of DUEs and CUEs is maximized. The approach is formulated as a optimization problem, and we prove that the problem is log concave. Hence, the optimum powers for active UEs can be found easily using modern computational methods. Both the theoretical and simulated results show that the joint connectivity probability is improved by one to two orders of magnitude by applying the optimization procedure compared to conventional LTE open loop power allocation. This dramatic improvement comes at the cost of an increase in UE average transmit power. Thus, the proposed technique is well suited to 5G public safety and disaster relief communication modes where enhanced connectivity is the top priority.

Joint Coverage Enhancement by Power Allocation in Poisson Clustered Out-of-Band D2D Networks

“…Therefore, the lower bound on p nc ij given by ( 15) is logconcave inP κ . The objective in ( 19) is also log-concave, because the product of log-concave coverage probabilities is still log-concave [52]. This concludes our proof.…”

“…The coherent and single-cluster approximated coverage probability formulations have similar mathematical structures to the one in [52], within which the proof of Theorem 1 is given. The proof of non-coherent lower-bound coverage probability formulation is given as follows:…”

“…For a non-coherent coverage probability between the transmitting node i and receiving node j given by (17), applying variable transformationsP i = ln P i and taking the natural logrithm on both sides of the equation, we have ln p nc ij = ln p sc ij + ln L I β Ge (dij ) exp(Pi) . ln p sc ij is concave inP κ [52] and , where the right side of the inequality is also concave inP i ∈ P κ , since its second order derivative with respect toP i equals −A( 2 α ) 2 exp(− 2 αP i ) < 0, where A = πλ o m (βd α ij P d ) 2/α sinc(2/α) . Therefore, the lower bound on p nc ij given by ( 15) is logconcave inP κ .…”