Optimal resource allocation for sensing-based spectrum sharing D2D networks

Abstract

To improve the spectral efficiency of a cellular network, sensing based spectrum sharing device-to-device (sensing-based D2D) technique is introduced in this paper. To be specific, efficient algorithms are introduced to find the optimal sensing time and power allocation for single-subband sensing-based D2D systems under cellular link’s rate loss constraint. For multi-subband systems, D2D users may have limited sensing capability where only a part of the subbands can be sensed. Accordingly, a partially observable Markov decision process (POMDP) model is introduced and an efficient algorithm is introduced to address the resource allocation problem.

Simulation results suggest that sensing-based D2D outperforms overlay D2D when the transmitter of D2D link is far away from the receiver of cellular link in the single-subband single-subcarrier case. In the single-subband multi-subcarrier case, sensing-based D2D can achieve 15.2% increase in spectral-efficiency over overlay D2D. In the multi-subband case, when D2D users have limited sensing capability, our proposed Separate Principle method can further improve D2D link’s spectral-efficiency by $5\%\sim 15\%$ compared to the Greedy algorithm.

Introduction

The global mobile data traffic is predicted to increase nearly 11-fold between 2013 and 2018, and there will be over 10 billion connected mobile devices by 2018 [2]. Accordingly, current deployed 4G cellular technologies, which have extremely efficient physical and MAC layer performance, are still lagging behind mobile users’ future data demand. To address this challenge, device to device (D2D) communication, an evolving product of cellular networks, has been regarded as a promising component in the next generation cellular technologies, and has attracted attention from both industry and academia. D2D communication in cellular network is defined as direct communication between two mobile devices without traversing the Base Station (BS) or the core network [3]. Compared to conventional cellular communications where mobile data has to be sent to and from base stations, D2D communications will require less communication resources (time, spectrum) thus significantly improve the spectral efficiency of the overall system. Furthermore, when mobile devices are close to each other, the transmit power of D2D devices can be low, in this way the intra- and inter-cell interference can be reduced improving the energy/spectral efficiency of the whole network. Therefore, D2D communications bring a promising communication paradigm for beyond 4G/5G networks and the preliminary version of D2D networks (called proximity service) has been adopted in 3GPP Release 12 LTE-Advanced systems [4].

Based on different spectrum sharing strategies, D2D communication can be classified into two types [5]: in-band D2D, where D2D links utilize cellular spectrum and out-of-band D2D, where D2D links exploit unlicensed spectrum (e.g. 2.4 GHz ISM band) [3]. The key problem in out-of-band D2D is that D2D users have to compete with various users in the unlicensed band leading to low quality of service (QoS) guarantee. This paper focuses on in-band D2D systems which can be further divided into overlay D2D and underlay D2D according to D2D links’ different spectrum access strategies. If the cellular link and D2D link use orthogonal cellular spectrum, this paradigm is called overlay D2D [6]; otherwise, if the D2D link access to the same spectrum concurrently with cellular links, this paradigm is called underlay D2D [7]. By borrowing the idea of sensing based spectrum sharing in cognitive radio networks, a new spectrum access paradigm, sensing based spectrum sharing D2D communications, is introduced in [8], [1]. In this paper, we analyze optimal resource allocation strategies for sensing based spectrum sharing D2D networks.

A sensing based spectrum sharing D2D system can be viewed as a hybrid spectrum access paradigm. One important motivation of sensing based D2D is that it is impossible for a base station to assign spectrums for each D2D link as it does in overlay or underlay cases because of the huge amount of connected mobile devices and the corresponding control overhead. Instead, sensing-based D2D enables each D2D user the ability of spectrum sensing, and based on different sensing outcomes, D2D users will cognitively access the cellular spectrum using different level of transmit power. Specifically, if a D2D user senses the cellular spectrum idle, it will transmit at a higher power level to improve its throughput; otherwise, if the cellular spectrum is sensed busy, it will transmit at a lower power level to mitigate its interference to cellular links. As a result, sensing-based D2D provides D2D users with more transmitting opportunities compared to overlay D2D. Compared to underlay D2D, sensing-based D2D increases the throughput by allowing mobile users to transmit at a higher power. It is important to note that from network design perspective sensing-based D2D cellular networks assumes that the network infrastructure releases certain authority (spectrum sensing, resource allocation, and data transmission) to mobile devices which coincides very well with recent evolution paths of mobile cellular networks.

There are two key issues in sensing-based D2D: the protection of cellular links and the sensing-throughput trade-off for D2D links. Traditionally, interference power constraint is considered in underlay systems and miss detection constraint is considered in overlay ones [9]. However, both of these constraints can not directly measure the QoS at which cellular links’ performance are affected by D2D users. A rate loss constraint (RLC) is introduced in [10] as a new metric in cognitive radio networks which can be directly related to primary users’ (PU) QoS. The new criterion is shown to provide improved capacity performance over conventional criterions. In this paper, we adopt rate loss constraint in our sensing-based D2D networks. This means that the rate loss of cellular links caused by D2D links’ interference should not exceed a certain tolerable threshold. By adopting the rate loss constraint in the design of sensing-based D2D systems, cellular links’ performance can be well protected.

Another issue in sensing-based D2D system is the sensing-throughput trade-off which is to balance between the protection of cellular link’s quality and the maximization of D2D links’ throughput. Sensing-throughput trade-off has been studied in cognitive radio networks in [9], [11], [12], [13], [14]. The sensing module should select the best spectrum resource to sense and optimize the sensing time in order to maximize secondary users’ (SU) throughput in cognitive radio networks. Different algorithms are proposed to optimize sensing time, sensing threshold, and power allocation under different constraints and difference scenarios. In [12], an algorithm to find the optimal sensing time and power allocation in spectrum overlay case using interference power constraint is introduced. In [13], the problem of optimizing sensing time and power allocation in sensing based spectrum sharing cognitive radio networks under interference power constraint is studied. In [15], optimal power and subcarrier allocation strategy to maximize SU’s throughput subject to SU’s quality of service constraint as well as PU’s rate loss constraint is introduced for multi-subcarrier systems.

Most of the existing work in cognitive radio networks or D2D networks are based on the assumption that D2D links can sense all the subbands during a time-slot. However, due to the limitation of hardware, mobile devices may have limited sensing capability, i.e., mobile users can only sense part of the available subbands and the sensing result may not always be perfect. In [16], a partially observable Markov decision process (POMDP) framework is introduced to solve the subband selection, sensing and spectrum access optimization problem in cognitive radio networks. In [17], the myopic subband selection policy that maximizes the immediate one-step reward is shown to be optimal in the POMDP framework when the channel state of subbands transitions are positively correlated over time or the number of subbands is limited to two or three. Power allocation in cognitive radio network using the POMDP framework is considered in [18], however, the sensing result is assumed to be perfect. In [19], the problem of whether to sense and how long SU should sense in an energy-constrained single-subband cognitive radio network is investigated using the POMDP framework.

In this paper, we study the optimal sensing time and power allocation strategy for sensing-based D2D networks with cellular links’ RLC. To be specific, we first characterize the optimal system operation and compare the performance (D2D link’s spectral-efficiency and optimal sensing time) of sensing-based D2D with that of overlay D2D systems in single-subband case. In multi-subband systems, when D2D users have limited sensing capability, i.e., D2D users can only sense one subband at a time-slot, we introduce a POMDP model to conduct subband selection, power allocation, and sensing time optimization. Our paper is trying to bridge the gap between [18], [19] and [20], [21] for sensing-based D2D networks under the rate loss constraint of cellular links. The remainder of this paper is organized as follows. In Section 2, we introduce the system models. We present the optimal power allocation and sensing time algorithm in Section 3. Section 4 shows our detailed simulation results. Summary and conclusions are given in Section 5.