Cloud-based Management of Energy-Efficient Dense IEEE 802.11ax Networks

Alexander Krotov; Anton Kiryanov; Evgeny Khorov

REVIEW 4 major objections 4 minor 14 references

Cloud-based Management of Energy-Efficient Dense IEEE 802.11ax Networks

T0 review · 4 major / 4 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A cloud-based controller can jointly tune transmit power and schedules in dense Wi-Fi networks to maximize a fairness-aware energy-efficiency objective, and simulation shows up to 50% higher energy efficiency than throughput-only control.

desk verdict Sensible new fairness-aware energy-efficiency objective for dense Wi-Fi cloud control, but the global-optimality proof relies on an unstated monotonicity assumption that needs to be fixed. read the letter →

arxiv 1908.03803 v3 pith:XLTZ5GYX submitted 2019-08-10 cs.NI

Evgeny Khorov , Anton Kiryanov , Alexander Krotov This is my paper · ORCID

classification cs.NI

keywords IEEE802.11axenergyefficiencyWi-Fibranchandboundglobaloptimizationpowercontrolfairnesscloudmanagement

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper tries to establish that a cloud-based manager can solve the joint transmit-power and transmission-scheduling problem in dense IEEE 802.11ax Wi-Fi networks for a fairness-aware energy-efficiency objective, not just for throughput. It defines the network objective as the ratio of an alpha-fair mean data rate to the total power consumed by all access points, and shows that maximizing this ratio can be rewritten as a monotonic optimization problem solvable to global optimality with a branch-and-bound search. The paper reports that in simulated dense deployments this approach raises energy efficiency by up to 50% compared with the authors' earlier throughput-only power-control-and-scheduling algorithm, while keeping mean and geometric-mean throughput nearly unchanged. If the claim holds, operators of crowded Wi-Fi deployments can cut energy cost without sacrificing client fairness.

What carries the argument

The load-bearing mechanism is the decomposition of $\log\hat U$ into a difference of non-decreasing functions $V(r)-W(r)$. $V$ is the log of the $\alpha$-fair mean rate; $W(r)=\log\sum_i(p_i+p_c)$ is non-decreasing in rates because transmit powers increase with SINR, a monotonicity result the paper takes from its reference [12, Lemma 2]. Introducing the auxiliary variable $w\in[-W(b),-W(a)]$ turns the difference into the monotone constraint $W(r)+w\le0$, so the problem becomes maximizing the monotone objective $V(r)+w$ over monotone constraints, which branch-and-bound handles by splitting the box-shaped search space, bounding the best possible value in each box, and discarding boxes that cannot beat the incumbent. The carrier-sense constraint (9) and power limits are also monotone in $(r,w)$, so they fit the same pruning logic without extra heuristics.

What would settle it

On a small network of two or three links with known path losses, power caps, and carrier-sense thresholds, solve the same objective by an independent exhaustive grid over all feasible rate-power combinations and compare with the branch-and-bound result; any instance where the optimizer's energy efficiency falls below the exhaustively computed maximum, or where the solution changes with the initial box size, would show the claimed global optimum is not reached.

Watch

Extended reading notes

Core claim

On its own terms, the central discovery is that the global energy-efficiency problem for dense Wi-Fi can be cast as a difference of two monotone functions and then as a monotone optimization problem with a global optimum. The objective is $\hat U=U^{-1}(\frac{1}{N}\sum_i U(r_i))/\sum_i(p_i+p_c)$, where $U$ is the $\alpha$-fair utility (the geometric mean at $\alpha=1$); taking logarithms gives $\log\hat U=V(r)-W(r)$ with $W(r)=\log\sum_i(p_i+p_c)$. Because each link's transmit power is non-decreasing in its SINR, $W$ is non-decreasing in the rate vector, so an auxiliary variable $w$ converts the problem into maximizing $V(r)+w$ subject to $W(r)+w\le0$ plus the Wi-Fi carrier-sense and maximum-power constraints. The branch-and-bound search over boxes in $(r,w)$ then finds the global maximum. For time-varying deployments, the paper differentiates the objective and applies the same machinery to maximize the incremental gain. In packet-level simulation with 1 to 30 access points, the method uses up to 50% less energy per delivered bit than the previous throughput-optimal algorithm while delivering similar mean and geometric-mean throughput.

Load-bearing premise

The argument assumes that the total power consumed by the network never decreases when any link's data rate rises, a monotonicity that holds only for rate vectors actually achievable under the power and carrier-sense limits; if the search visits infeasible rate vectors, the bound may mislead the branch-and-bound search.

Editorial extensions

If this is right

A cloud controller can use the algorithm to tune transmit powers and schedules together, so dense Wi-Fi energy consumption approaches the physical floor set by circuit power alone (the paper's simulation shows this).
The same solver covers the whole alpha-fair family, so it can maximize arithmetic-mean energy efficiency (the classical global energy-efficiency metric), geometric-mean energy efficiency, or any intermediate fairness level without changing the algorithm.
Because the objective and constraints are monotone, the branch-and-bound approach transfers the global energy-efficiency machinery developed for cellular networks to Wi-Fi's carrier-sense restrictions.
In dynamic scenarios, the differentiated objective gives a marginal-update rule: instead of re-optimizing from scratch, the controller maximizes the derivative of energy efficiency at each scheduling step.
The reported 50% energy-efficiency gain over the earlier algorithm comes with similar mean and geometric-mean throughput, implying the gain comes from power reduction rather than throughput loss.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The branch-and-bound solver could serve as an offline benchmark: operators could compute the global optimum for recorded traffic traces and measure how much efficiency their online heuristics leave on the table.
The same monotonic formulation may extend to objectives with per-link circuit powers or sleep modes, but then W(r) would no longer be a simple sum of constants and the monotonicity argument would need rework.
The fairness parameter alpha acts as an operator knob: lowering it favors total delivered data, raising it protects weak links, and the paper's solver covers the entire family, so a controller could adapt alpha to traffic demands without changing the optimization core.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Short take: this is a solid, conventional engineering paper with a genuinely new objective — alpha-fair mean throughput over total power — for cloud-coordinated dense Wi-Fi. The NS-3 results showing up to 50% energy-efficiency gains over the authors' previous throughput-only power control are plausible and consistent with the physics: cut transmit power where it doesn't hurt fairness.

What's new: formulation (3) is new for Wi-Fi; the branch-and-bound machinery is borrowed from Tuy's textbook and the authors' earlier work, but applying it to this objective while respecting carrier-sense constraints is a new application. The paper also deserves credit for implementing the whole thing in NS-3 and comparing against both legacy CSMA and their own previous algorithm — that's a reasonable evaluation baseline.

The soft spots, in proportion. The main one is the proof that branch-and-bound converges to the global optimum. The DC decomposition requires W(r) = log Σ(p_i+p_c) to be non-decreasing in r. The paper argues via [12, Lemma 2] that r'≽r implies γ'≽γ. That inference is not valid for the step-function rate-SINR model of Fig. 2: a non-decreasing step function maps different SINR values to the same rate, so r'≽r does not force γ'≽γ. To make the argument work you need to define W(r) using a minimal-power inverse (pick the smallest p achieving each r) and then prove monotonicity on the feasible set, including carrier-sense constraints. As written, the global-optimality claim is not established. I suspect it's fixable — the authors may have this in mind — but they need to write it down.

Secondary issues: the dynamic scheduling section (IV-C) derives a derivative and then says 'we use the approach described above,' which is hand-wavy and never evaluated. The simulation section lacks error bars and never states the circuit power p_c, even though the lower bound in Fig. 4 depends on that value. The carrier-sense threshold is also unspecified.

The citation pattern is fine. Building on your own prior work and on Tuy's standard text is normal for this incremental global-optimization genre.

Who this is for: anyone working on centralized Wi-Fi management, energy-efficiency metrics for dense WLANs, or cloud-based radio resource optimization. It deserves a serious referee. I'd send it to peer review with a conditional recommendation: the authors should patch the monotonicity proof, report all parameters, and either defend or trim the dynamic scheduling section.

Referee Report

4 major / 4 minor

Summary. The paper proposes a cloud-based centralized controller for dense IEEE 802.11ax networks. The controller maximizes a fairness-aware energy-efficiency objective: the α-fair mean throughput (via the α-fair utility function) divided by the total consumed power, which includes both transmit power and a per-device circuit power p_c. The optimization problem is rewritten as the difference of two monotone functions of the data-rate vector, and a branch-and-bound method is applied. A dynamic-scheduling extension is sketched by maximizing the time derivative of the objective. The proposed scheme is evaluated in NS-3 against legacy Wi-Fi, scheduling-only, and power-control-plus-scheduling baselines, with the headline result of up to 50% energy-efficiency improvement over the authors' prior power-control algorithm at similar throughput.

Significance. If the mathematical derivation is made rigorous and the simulations are properly reported, this is a useful contribution: it brings fair energy-efficiency optimization into the domain of monotonic optimization and branch-and-bound, and it provides a concrete centralized design for emerging dense Wi-Fi deployments. The paper's strengths are that the objective is well motivated, the proposed algorithm is actually implemented in NS-3, and the comparison includes both legacy behavior and the authors' previous algorithms. The main risk is that the global-optimality claim rests on an incompletely justified monotonicity step, and the empirical headline is not supported by the reported simulation detail.

major comments (4)

[Section IV-A, Eqs. (4)-(8)] The assertion that W(r) = log(Σ_i(p_i+p_c)) is a non-decreasing function of the rate vector r is not sufficiently justified. The text cites [12, Lemma 2] for the monotonicity of transmit powers in SINR vectors, but it does not verify the preconditions. First, under the step-function rate-SINR model in Fig. 2, a rate vector r does not determine a unique SINR vector or a unique power vector; one must fix a well-defined minimal-power inverse, for example γ_i = inf{γ: f(γ) ≥ r_i}, before W(r) is even well-defined, and this inverse is never stated. Second, [12] applies to feasible SINR targets and does not by itself cover the constraints 0 ≤ p_i ≤ p̂_i and the carrier-sense constraint (9); the paper does not show that every r in the branch-and-bound search space is feasible under these constraints, nor that monotonicity survives when a link is silenced by (9) or hits its power limit. If W(r) is undefined or non-monotone on a box, the upper bounds used in the branch-and-bound search are invalid and the claimed global optimum of (3) is not established. The authors should either prove the required monotonicity on the feasible set with explicit assumptions or restrict the search to a set on which it is guaranteed.
[Section V, Table I and Figs. 3-5] The circuit power p_c is never given, although it appears directly in the objective (3), in the lower-limit line in Fig. 4, and in all energy-efficiency values. Without this parameter, the reported energy-efficiency curves and the 'up to 50%' improvement cannot be reproduced or quantitatively interpreted. The paper should state the value of p_c used in the simulation and, ideally, show the sensitivity of the results to this parameter.
[Section V, Figs. 3-5] The simulation results show no error bars, confidence intervals, or number of random topologies, even though clients are randomly placed. It is therefore impossible to assess whether the observed differences between the proposed algorithm and the 'power control and scheduling' baseline are statistically significant or within the noise of the random placement. The authors should report the number of independent runs and standard deviations or confidence intervals, and identify at which AP counts the 50% energy-efficiency gain is actually observed.
[Section IV-C, Eq. (13)] The dynamic-scheduling extension is not rigorously connected to the branch-and-bound framework. The derivative objective (13) is a weighted sum of instantaneous rates minus a power-consumption term, which is not shown to fit the same monotonic difference-of-functions structure that justifies the algorithm in Section IV-A. The text simply says 'we use the approach that is described above' after Eq. (13), but no proof or precise restatement is given. If the dynamic part is intended only as a heuristic, that should be stated explicitly; if global optimality is claimed, a supporting argument is needed. This section is also not evaluated in the numerical results.

minor comments (4)

[Section IV-A, between Eqs. (6) and (7)] The interval [W(a), W(b)] is called the 'target set' of the function W, but the vectors a and b are never defined; the authors should specify the lower and upper bounds of the search box used in the branch-and-bound method.
[Section V, pathloss formula] The distance variable in the pathloss formula is also denoted r, which conflicts with the rate vector r used throughout the paper; renaming the distance variable, e.g. d, would improve readability.
[Section V, Fig. 4] The vertical axis label 'Wt' should be 'W' for watts; this is a minor typo but should be corrected.
[Section IV-A, reference [12]] The cited [12, Lemma 2] is from a 1967 paper and is used as a black box; since the monotonicity claim is load-bearing, the authors should either state the relevant lemma explicitly and check its hypotheses, or provide a short self-contained proof for the power-SINR relation used here.

Circularity Check

0 steps flagged · score 1.0 of 10

No material circularity: the optimization derivation is grounded in external references, and the paper's self-citations serve only as baselines and implementation pointers.

full rationale

The paper's central derivation is the decomposition of the log objective (4) into V(r)-W(r) with W(r) shown nondecreasing via [12, Lemma 2], and then the transformation into a monotonic optimization problem (7)-(8) solved by branch-and-bound from [9, Theorem 11.1] and [9, 11.1.2]. Both [12] (Stefanyuk-Tsetlin) and [9] (Tuy) are external references, not self-citations, so the global-optimality machinery does not reduce to the authors' own prior results. The proposed objective (3) is a genuinely new combination of a fairness utility with total network power. The NS-3 rate-SINR step function in Fig. 2 is an approximate planning model used inside the optimizer, while performance is evaluated with the NS-3 simulator; it is not a fitted parameter renamed as a prediction. The paper's self-citation to [11] appears as a comparison baseline and as a pointer for the detailed branch-and-bound implementation, but the underlying convergence argument is supplied by [9], so this self-citation is not load-bearing. The monotonicity precondition for W(r) regarding feasibility of rate vectors and uniqueness of the power inverse is a correctness concern, not a circularity, because the paper does not define W(r) in terms of the target result or fit the target quantity. Overall, no step in the derivation chain reduces by construction to its own inputs.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central claim rests on the monotonic decomposition of the objective, which depends on a cited power-control monotonicity lemma, and on several simulation inputs (circuit power, rate model, carrier sense threshold) that are not fully specified.

free parameters (3)

Circuit power p_c = not specified
Used in the objective (3) and in the energy consumption plots, but its numerical value is never given; the EE results and the minimum-power line in Fig. 4 depend on it.
Rate-SINR step function = values shown only in Fig. 2, not tabulated
The optimizer maps SINR to rate using a step function fitted to NS-3 Minstrel results; exact thresholds and rates are absent from the text.
Carrier sense threshold c_hat = not specified in Table I
Constraint (9) uses a threshold for acceptable interference from other transmitters; the simulation parameter table lists only Receiver Sensitivity, dBm -96, which may not equal c_hat.

assumptions (4)

domain assumption Power monotonicity with SINR ([12, Lemma 2]): if gamma' >= gamma componentwise, then the corresponding transmit power vectors satisfy p' >= p.
Used in Section IV-A to prove W(r) is non-decreasing in rates. The paper cites [12] but does not restate the proof or the feasibility conditions required in this interference model.
domain assumption Rate is a non-decreasing function of SINR, r_i = f(gamma_i(p)), and the optimizer uses a step-function approximation of it.
Section IV-A and V; the monotonicity is needed for the DC decomposition and the step function is the planning model.
domain assumption CSMA/CA can be abstracted as the deterministic constraint max_j b_ij p_j <= c_hat for each active transmitter.
Constraint (9); real Wi-Fi carrier sensing is stochastic with backoff and hidden nodes; the paper relies on this abstraction for the optimizer's feasible set.
standard math Tuy's DC programming theorem [9, Theorem 11.1] transforms a difference of monotonic functions into a monotone maximization over a normal set.
Used in Section IV-A to rewrite (4) as (7)-(8).

how reviews work

0 comments

Cite this review

Pith. "Pith review of Cloud-based Management of Energy-Efficient Dense IEEE 802.11ax Networks." pith.science (2026). https://pith.science/paper/XLTZ5GYX

@misc{pith2026190803803,
  author       = {Pith},
  title        = {Pith review of: Cloud-based Management of Energy-Efficient Dense IEEE 802.11ax Networks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XLTZ5GYX}},
  note         = {Machine review of arXiv:1908.03803}
}

read the original abstract

During the last decade, the number of devices connected to the Internet by Wi-Fi has grown significantly. A high density of both the client devices and the hot spots posed new challenges related to providing the desired quality of service in the current and emerging scenarios. To cope with the negative effects caused by network densification, modern Wi-Fi is becoming more and more centralized. To improve network efficiency, today many new Wi-Fi deployments are under control of management systems that optimize network parameters in a centralized manner. In the paper, for such a cloud management system, we develop an algorithm which aims at maximizing energy efficiency and also keeps fairness among clients. For that, we design an objective function and solve an optimization problem using the branch and bound approach. To evaluate the efficiency of the developed solution, we implement it in the NS-3 simulator and compare with existing solutions and legacy behavior.

Figures

Figures reproduced from arXiv: 1908.03803 by the authors.

**Figure 1.** Cloud-based Wi-Fi Management System thresholds, and even time-division multiplexing, including those enabled by IEEE 802.11ax. Besides, large Wi-Fi deployments may contain hundreds or even thousands of APs. In such a scenario, it becomes important to take into account the amount of energy consumed by each AP, and try to design energy-efficient solutions. In this paper, we consider the problem of cooperation of Wi-Fi… view at source ↗

**Figure 2.** Data rate as a function of SINR Let P and Ri be the total energy spent from the beginning of the experiment and the total amount of data transmitted over the link i from the beginning of the experiment. In this case, we can rewrite the objective function as follows: log Uˆ = log U −1 [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 3.** Throughput Table I SIMULATION PARAMETERS Parameter Value AP location height, m 3 Client location height, m 1 Maximum transmit power, mW 40 Noise, dBm/Hz -174 Channel width, MHz 80 Amplifier noise, dB 7 Receiver Sensitivity, dBm -96 Rate control algorithm Minstrel HT 0 5 10 15 20 25 30 Number of access points 0 10 20 30 40 Energy consumption, Wt Legacy Scheduling Power control and scheduling Energy efficient power co… view at source ↗

Figures from the paper (2 more)

**Figure 5.** Figure 5: Energy efficiency APs. For a single AP, the results of all solutions except for the one proposed in this paper coincide. This happens because for a single AP it does not matter how access to the medium is organized. In any case, this AP receives the whole channel. Howe…

**Figure 4.** Figure 4: Energy consumption 2) Scheduling. We schedule transmissions of APs in order to maximize geometric mean throughput. For that, we also solve the optimization problem described in [11]. 3) Power control and scheduling. In addition to scheduling, we tune transmit power as …

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

14 extracted references · 14 canonical work pages

[12]

Power regulation in a group of radio stations,

V . L. Stefanyuk and M. L. Tsetlin, “Power regulation in a group of radio stations,” Problemy Peredachi Informatsii, vol. 3, no. 4, pp. 49–57, 1967

work page 1967
[1]

A Tutorial on IEEE 802.11ax High Efﬁciency WLANs,

E. Khorov, A. Kiryanov, A. Lyakhov, and G. Bianchi, “A Tutorial on IEEE 802.11ax High Efﬁciency WLANs,” IEEE Communications Surveys Tutorials, vol. 21, no. 1, pp. 197–216, Firstquarter 2019

work page 2019
[2]

Cloud Control to Optimize Real-Time Video Transmission in Dense IEEE 802.11aa/ax Networks,

E. Khorov, A. Ivanov, A. Lyakhov, I. F. Akyildiz, “Cloud Control to Optimize Real-Time Video Transmission in Dense IEEE 802.11aa/ax Networks,” in Mobile Ad-hoc and Sensors Systems . IEEE, 2018

work page 2018
[3]

A survey of energy-efﬁcient techniques for 5g networks and challenges ahead,

S. Buzzi, I. Chih-Lin, T. E. Klein, H. V . Poor, C. Yang, and A. Zappone, “A survey of energy-efﬁcient techniques for 5g networks and challenges ahead,” IEEE Journal on Selected Areas in Communications , vol. 34, no. 4, pp. 697–709, 2016

work page 2016
[4]

Energy-constrained error control for wireless channels,

M. Zorzi, R. R. Rao et al., “Energy-constrained error control for wireless channels,” IEEE Personal communications , vol. 4, no. 6, pp. 27–33, 1997

work page 1997
[5]

Energy-efﬁcient wireless communications: tutorial, survey, and open issues,

G. Y . Li, Z. Xu, C. Xiong, C. Yang, S. Zhang, Y . Chen, and S. Xu, “Energy-efﬁcient wireless communications: tutorial, survey, and open issues,” IEEE Wireless Communications , vol. 18, no. 6, 2011

work page 2011
[6]

Interference- aware energy-efﬁcient power optimization

G. Miao, N. Himayat, Y . G. Li, A. T. Koc ¸, and S. Talwar, “Interference- aware energy-efﬁcient power optimization.” in ICC, 2009, pp. 1–5

work page 2009
[7]

Energy-efﬁcient scheduling and power allocation in downlink ofdma networks with base station coordination,

L. Venturino, A. Zappone, C. Risi, and S. Buzzi, “Energy-efﬁcient scheduling and power allocation in downlink ofdma networks with base station coordination,” IEEE transactions on wireless communications , vol. 14, no. 1, pp. 1–14, 2015

work page 2015

Show all 14 references

[8]

Energy efﬁciency in wireless networks via fractional programming theory,

A. Zappone, E. Jorswieck et al., “Energy efﬁciency in wireless networks via fractional programming theory,” F oundations and TrendsR© in Com- munications and Information Theory , vol. 11, no. 3-4, pp. 185–396, 2015

2015
[9]

Tuy, Convex Analysis and Global Optimization

H. Tuy, Convex Analysis and Global Optimization . Springer, 2016, vol. 110

2016
[10]

Globally optimal energy-efﬁcient power control and receiver design in wireless networks,

A. Zappone, E. Bj ¨ornson, L. Sanguinetti, and E. Jorswieck, “Globally optimal energy-efﬁcient power control and receiver design in wireless networks,” IEEE Transactions on Signal Processing , vol. 65, no. 11, pp. 2844–2859, 2017

2017
[11]

Joint power control and time division to improve spectral efﬁciency in dense Wi-Fi networks,

E. Khorov, A. Kiryanov, and A. Krotov, “Joint power control and time division to improve spectral efﬁciency in dense Wi-Fi networks,” in Black Sea Conference on Communications and Networking (BlackSea- Com), 2018 IEEE International . IEEE, 2018, pp. 199–203

2018
[13]

S. Merlin. TGax simulation scenarios. [On- line]. Available: https://mentor.ieee.org/802.11/dcn/14/ 11-14-0980-16-00ax-simulation-scenarios.docx
[14]

[Online]

The NS-3 network simulator . [Online]. Available: http://www.nsnam.org/

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [12]

Power regulation in a group of radio stations,

V . L. Stefanyuk and M. L. Tsetlin, “Power regulation in a group of radio stations,” Problemy Peredachi Informatsii, vol. 3, no. 4, pp. 49–57, 1967

work page 1967

[2] [1]

A Tutorial on IEEE 802.11ax High Efﬁciency WLANs,

E. Khorov, A. Kiryanov, A. Lyakhov, and G. Bianchi, “A Tutorial on IEEE 802.11ax High Efﬁciency WLANs,” IEEE Communications Surveys Tutorials, vol. 21, no. 1, pp. 197–216, Firstquarter 2019

work page 2019

[3] [2]

Cloud Control to Optimize Real-Time Video Transmission in Dense IEEE 802.11aa/ax Networks,

E. Khorov, A. Ivanov, A. Lyakhov, I. F. Akyildiz, “Cloud Control to Optimize Real-Time Video Transmission in Dense IEEE 802.11aa/ax Networks,” in Mobile Ad-hoc and Sensors Systems . IEEE, 2018

work page 2018

[4] [3]

A survey of energy-efﬁcient techniques for 5g networks and challenges ahead,

S. Buzzi, I. Chih-Lin, T. E. Klein, H. V . Poor, C. Yang, and A. Zappone, “A survey of energy-efﬁcient techniques for 5g networks and challenges ahead,” IEEE Journal on Selected Areas in Communications , vol. 34, no. 4, pp. 697–709, 2016

work page 2016

[5] [4]

Energy-constrained error control for wireless channels,

M. Zorzi, R. R. Rao et al., “Energy-constrained error control for wireless channels,” IEEE Personal communications , vol. 4, no. 6, pp. 27–33, 1997

work page 1997

[6] [5]

Energy-efﬁcient wireless communications: tutorial, survey, and open issues,

G. Y . Li, Z. Xu, C. Xiong, C. Yang, S. Zhang, Y . Chen, and S. Xu, “Energy-efﬁcient wireless communications: tutorial, survey, and open issues,” IEEE Wireless Communications , vol. 18, no. 6, 2011

work page 2011

[7] [6]

Interference- aware energy-efﬁcient power optimization

G. Miao, N. Himayat, Y . G. Li, A. T. Koc ¸, and S. Talwar, “Interference- aware energy-efﬁcient power optimization.” in ICC, 2009, pp. 1–5

work page 2009

[8] [7]

Energy-efﬁcient scheduling and power allocation in downlink ofdma networks with base station coordination,

L. Venturino, A. Zappone, C. Risi, and S. Buzzi, “Energy-efﬁcient scheduling and power allocation in downlink ofdma networks with base station coordination,” IEEE transactions on wireless communications , vol. 14, no. 1, pp. 1–14, 2015

work page 2015