# Maximizing Spatial -Fairness in Multi-Tier Multi-Rate Spatial Aloha ......

date post

19-Aug-2020Category

## Documents

view

0download

0

Embed Size (px)

### Transcript of Maximizing Spatial -Fairness in Multi-Tier Multi-Rate Spatial Aloha ......

1

Maximizing Spatial α-Fairness in Multi-Tier Multi-Rate Spatial Aloha Networks

Yujie Xu, Student Member, IEEE, Ben Liang, Fellow, IEEE, Gary Boudreau, Senior Member, IEEE, and S. Hossein Seyedmehdi

Abstract—We consider the maximization of α-fair utility in a generalized spatial Aloha network consisting of multiple tiers of transmitter-receiver (T-R) pairs each forming a Poisson bipolar process. The tiers are distinguished by transmission power and the T-R distance. Multi-rate communication between the T-R pairs is facilitated by multiple received signal-to-interference ratio (SIR) thresholds. We aim to optimize the transmission probability of each tier. This results in a complex non-convex optimization problem due to intra-tier and cross-tier interference. We propose a solution termed Minorize-Maximization with Tier Separation (MMTS), through designing an iterative sequence of lower bound problems that can be decomposed into tier- separable one-dimensional convex optimization problems and solved efficiently. Specific solutions are derived for the cases 0 ≤ α < 1, α = 1, and α > 1. We show the convergence of MMTS to the objective value of a Karush-Kuhn-Tucker (KKT) point of the original problem and further identify several conditions under which it finds the global optimum. Numerical results demonstrate the near optimality of MMTS and substantial performance advantage over existing alternatives.

Index Terms—Spatial Aloha networks, utility maximization, transmission probability, minorize-maximization, tier separation.

I. INTRODUCTION

Direct transmission between devices in proximity is a well promoted paradigm to allow ad hoc network access and to increase wireless spectrum utilization [1]. In particular, device- to-device communication has become an important aspect of next-generation wireless standardization [2]. One of the main challenges in direct communication among devices is how to allocate the common wireless spectrum in an efficient and fair manner over a large-scale network, such as in the Internet-of- Things environment.

Utility maximization in random access networks has re- ceived much attention in the literature. The authors of [3] first studied the problem of maximizing network α-fair utility for α > 1 using the protocol model. They showed that the problem can be recast as a convex optimization problem and proposed a distributed scheduling based on Lagrangian dual decomposition. Then [4] further proposed a method based on coordinate descent to solve the utility maximization problem by observing that the problem is convex in the transmission

Y. Xu and B. Liang are with the Department of Electrical and Computer Engineering, University of Toronto, Toronto, ON, M5S 3G4, Canada (e- mail: {yujiexu, liang}@ece.utoronto.ca). G. Boudreau and S. H. Seyedme- hdi are with Ericsson Canada, Ottawa, ON, K2K 2V6, Canada (e-mail: {gary.boudreau, hossein.seyedmehdi}@ericsson.com).

This work was supported in part by research grants from Ericsson and the Natural Sciences and Engineering Research Council of Canada.

probability of each T-R pair separately for all α values. In addition, [5] studied the utility maximization problem in the single signal-to-interference-plus-noise ratio (SINR)-threshold model instead of the protocol model, and show that the single SINR threshold model yields higher throughput. However, these works all required that the exact position of each node is known, which may be difficult to obtain. This inspires researchers to study random access networks with a random topology defined only by its statistics.

In the presence of a large number of direct communication pairs, acquiring the exact topology information is a prohibitive task, especially in networks with high mobility. Thus, research on wireless network models with random topology has drawn much attention. In the celebrated spatial Aloha model [6], only statistical information of the topology is available. Each transmitter in the network randomly transmits following the slotted Aloha medium access control (MAC) protocol. Both the Poisson point process (PPP) and Poisson bipolar process (PBP) are commonly used to model the location of transmitters and receivers in spatial Aloha networks. In the PBP model, the transmitters form a PPP, and each transmitter is paired with a dedicated receiver at some distance away.

An interesting design problem in spatial Aloha networks is to optimize the transmission probability. This is a challenging problem, often with a non-convex objective function due to signal interference. In [6]–[11], this problem is studied where all transmitters are assumed to use the same transmission prob- ability if the exact location of nodes is unknown. This model is suitable only when the network is uniform. In many practical scenarios, the transmitters may have different powers and the T-R distance may be different for different T-R pairs, so that the transmitters should use different transmission probabilities. Some previous studies have addressed this problem in static Aloha networks [3]–[5], [12]–[14], but none of them allows randomness in the network topology.

Furthermore, most existing works assume single-rate com- munication either based on a single received SIR threshold [5]–[11], such that the data rate is log(1 + Th) if the received SIR is above some threshold Th, and is zero otherwise; or based on the “protocol model” [3], [4], [12]–[14] wherein the data rate is some fixed term if the nearby transmitters do not transmit. In terms of physical implementation, both cases correspond to the usage of only a single modulation- coding scheme at the transmitter. Such a model simplifies mathematical analysis but has limited application in more sophisticate multi-rate systems.

In this work, we extend the spatial Aloha model of [7]–[9]

2

to the multi-tier, multi-rate regime. Each tier of the network is defined by the power of the transmitters and the T-R distance. We consider both intra-tier and inter-tier interference. We also accommodate multi-rate communication through multiple received SIR thresholds. We aim to optimize the transmission probability of each tier, to maximize a general α-fair utility function.

Our main contributions are as follows: • We first derive a closed-form expression of the average

throughput of T-R pairs in each tier, which takes into account the random location of multi-tier interferers and multiple received SIR thresholds. This is then used in the formulation of an optimization problem to maximize the network-wide spatial α-fair utility, which is generally non-convex.

• We propose a computationally efficient iterative algo- rithm, termed Minorize-Maximization with Tier Separa- tion (MMTS), to address this optimization problem. By exploring the partial-convexity and partial-concavity of the objective function when 0 ≤ α ≤ 1 and α > 1, respectively, we develop special lower bounds to the α-fair objective, which we dynamically update in each iteration through solving an optimization sub-problem. Furthermore, the lower bounds are designed so that these sub-problems can be decomposed into one-dimensional convex optimization problems that are separable accord- ing to T-R tiers, which drastically reduces the computa- tional complexity.

• We show that MMTS converges to the objective value of a KKT point of the optimization problem. We further provide various sufficient conditions under which the KKT point is a global optimizer. Numerical evaluation results demonstrate that MMTS is near optimal over a wide range of parameter settings, and it substantially outperforms existing alternatives.

The rest of the paper is organized as follows. In Section II, we summarize the related work. In Section III, we present the system model and formulate our optimization problem. In Sections IV, we derive the average throughput of com- munication pairs in different tiers. In Section V and VI, we present MMTS and discuss its convergence and optimality, respectively. In Section VII, we present numerical evaluation results. Conclusions are drawn in Section VIII.

II. RELATED WORKS

There has been a large amount of research into ad hoc, device-to-device, or direct-transmission networks that employ the Aloha MAC protocol [3]–[21]. Among them, [3]–[5], [12]–[14] consider a fixed transmitter/receiver topology, [6], [17]–[21] use the PPP model, [7], [10], [11], [15], [16] use the PBP model, while [8], [9] use the PBP model with partial topology information. The networks in the latter three groups are commonly termed spatial Aloha networks. In this section, we briefly review works in optimizing transmission probabilities in spatial aloha networks. We further describe the Minorize-Maximization framework and its application in communication systems.

TABLE I: Table of Notations

Notation Description N Number of tiers λn Intensity of PBP of transmitters in tier n Rn T-R distance in tier n Pn Transmission power of transmitter in tier n pn Transmission probability of transmitter in tier n γ Pathloss exponent L Number of SIR thresholds to modulate the received signal Tl lth SIR threshold α Fairness index in the utility function

A. Spatial Aloha Networks

Several studies consider the optimization of transmission probability in spatial Aloha [6]–[11]. However, in these works all transmitters are assumed to use the same transmission probability when the exact location of nodes is unknown. In our work, we design different transmission probabilities for different tiers of the network based on T-R distance and transmission power. Our numerical results show that this can lead to substantial performance improvement.

Furthermore, all of

*View more*