Pith. sign in

REVIEW 3 major objections 5 minor 18 references

Next-Generation Sustainable Wireless Systems: Energy Efficiency Meets Environmental Impact

T0 review · 3 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read This paper claims that putting carbon emissions per bit directly into a 6G resource-allocation objective, solved with multi-objective reinforcement learning, reduces average emissions per bit by about 26% at matched delay and energy efficie

desk verdict Worth a serious referee because the carbon-per-bit control problem is real, but the MDP formulation and Pareto-optimality claims don't hold as written; expect major revisions. read the letter →

arxiv 2509.02395 v1 pith:4ZYJSMKN submitted 2025-09-02 cs.IT cs.NImath.IT

Christo Kurisummoottil Thomas , Omar Hashash , Kimia Ehsani , Walid Saad This is my paper
classification cs.ITcs.NImath.IT MSC 90C2968T0594A05
keywords sustainabilitymetriccarbonemissionsperbitmulti-objectivereinforcementlearning6GresourceallocationrenewableenergyPareto-optimalpolicyefficiencyqueuingdelay
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

This paper tries to show that wireless resource allocation in 6G can be controlled directly by an environmental objective—grams of CO2 emitted per delivered bit—rather than by energy efficiency alone, which does not guarantee low carbon output. It defines a closed-form sustainability metric from the energy mix of renewable, grid, and battery sources and folds it into a multi-objective optimization over computing, communication, and energy resources. To solve the non-convex problem, the paper proposes a multi-objective reinforcement learning scheme whose global policy is a weighted mixture of per-objective policies. If the argument holds, operators gain a principled knob for trading emission reduction against delay and energy use, with simulations showing roughly 26% lower average emissions per bit than baselines that ignore carbon. The reader should care because it moves sustainability from a side constraint into the objective being optimized.

What carries the argument

The load-bearing object is the emissions-per-bit sustainability metric, which compares actual carbon emitted per delivered bit against an all-non-renewable baseline, making it dimensionless and directly optimizable. The solver is a variational multi-objective reinforcement learning update in which each objective gets a local policy, and the global policy is the weighted average of those local policies, with preference weights selecting the operating point on the Pareto frontier. An LSTM predictive model of emissions evolution is used to turn the per-slot metric into a long-term objective.

What would settle it

Retrain the same multi-objective reinforcement learning controller with the state expanded to include battery level, queue backlog, and current channel coefficients for each server, keeping the same objectives and simulator. If the emissions-per-bit versus delay Pareto frontier does not shift, the emissions-only state suffices; if it shifts measurably, the claimed optimality rests on an incomplete state and the 26% reduction is not attributable to the stated mechanism.

Watch

Extended reading notes

Core claim

The central claim is that carbon emissions per bit can serve as a sustainable and actionable control objective for a downlink 6G system with heterogeneous CPU/GPU servers and mixed renewable/grid/battery energy. The paper defines the per-server emissions per bit and a network sustainability measure equal to the square of one minus the ratio of actual emissions per bit to the emissions per bit that would occur if all energy came from non-renewable sources. It then formulates a multi-objective problem maximizing long-term sustainability, minimizing energy per bit, and minimizing average delay, and solves it with multi-objective reinforcement learning. The resulting global policy is a convex co

Load-bearing premise

The approach assumes that knowing each server's current emissions per bit is enough to decide the best actions, even though delays and energy use also depend on battery levels, queue backlogs, and random channel and traffic conditions that are not part of that state.

Editorial extensions

If this is right

  • Operators can optimize a concrete, measurable target—CO2 per delivered bit—instead of treating carbon as a side effect of energy use.
  • At matched QoS, the reported simulation results indicate roughly 26% lower average emissions per bit under varying packet arrival rates than methods that ignore carbon in the control objective.
  • Changing the preference weights moves the operating point along an emissions-versus-delay Pareto frontier, from about 33% emission reduction at low delay to about 9% at high delay.
  • Because the metric is defined from server-level emissions per bit, it applies across heterogeneous CPU/GPU servers and mixed renewable/grid/battery energy, so the same controller can govern different hardware configurations.
  • Using a predictive model of emissions evolution favors reductions averaged over many time slots rather than instantaneous dips, which matters when renewable supply and battery levels change over time.

Reading between the lines

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

  • Editorial extension: the metric's relative form—actual carbon per bit divided by an all-fossil baseline—is not tied to 6G specifics, so it could be reused for edge data centers, IoT gateways, or any system that can attribute energy to delivered bits.
  • Editorial extension: the 26% figure comes from a simulator with relatively stable emission factors and channel statistics; in the field, emission factors vary hourly, so a deployment would need online estimates, and the realized reduction could differ.
  • Editorial extension: the most direct test of the modeling assumptions is to add battery levels, queue backlogs, and channel states to the reinforcement learning state; whether the gains survive a richer state is an open question.
  • Editorial extension: because the preference weights are exposed parameters, the same framework could be adapted to per-user carbon budgets or regulatory emissions caps by adding constraints on cumulative emissions per bit.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes a carbon-emissions-per-bit sustainability metric S(t) (Eq. (12)) for a multi-server downlink system with renewable and grid energy sources, and formulates a multi-objective resource allocation problem P1 (Eq. (13)) that jointly maximizes long-term sustainability, minimizes energy per bit, and minimizes average delay. The solution method is an MORL framework: the state is defined as the vector of per-server emissions per bit, an LSTM is used as an environment model, and a variational inference procedure over three Q-functions produces a global policy π(a_t|s_t,V). The paper claims this policy achieves a Pareto-optimal tradeoff and reports about 26% lower average emissions per bit than baselines that do not include carbon emissions in their control objective.

Significance. If the formal claims are supported, the paper would provide a concrete, operationalizable way to include carbon footprint in 6G resource allocation and would demonstrate a tradeoff between sustainability, energy, and delay. The emissions-per-bit metric itself, normalized by an all-non-renewable baseline, is a reasonable and useful construct, and the paper explicitly targets joint computing, communication, and energy control rather than energy efficiency alone. However, the theoretical guarantees are not currently backed by a well-posed MDP, the central policy-composition proof is omitted, and the headline 26% improvement is a direct consequence of training on S(t). The significance of the work therefore depends on whether the MDP can be reformulated and whether the matched-QoS and statistical claims can be substantiated.

major comments (3)
  1. [§III-A, Eqs. (9)-(13)] The MORL state is not sufficient and the transition is not deterministic. Section III-A defines s_t = (Cb_1(t), ..., Cb_M(t)) and states that "state transitions are deterministic and follow (11)". This is load-bearing and incorrect. From (11), c_k(t) is a linear combination of energy decisions, and Cb_k(t) = c_k(t)/(T Σ_{u∈M_k} R_u(t)); R_u(t) depends on the stochastic channel h_u(t) and the transmit powers, so the next Cb is not a deterministic function of (s_t, a_t). Moreover, constraint (13d) and battery dynamics (2) require b_k(t−1) in the state, and the delay objective (10) depends on the queue backlog Ω_k(t) evolving stochastically through (9). Two states with identical Cb_k(t) can have different feasible action sets and different delay consequences. Consequently, the Q-functions in (14), the variational updates (16)–(18), and the Bellman-style argument underlying Lemma 1 are not j
  2. [§III-A, Eq. (16) and Lemma 1] The proof of the central policy-composition result is omitted, and the "local optimum" claim is unsupported. Lemma 1 states that the global policy is π = Σ_r ζ_r q_r, and the text concludes that alternating optimization "leads to a local optimum". The proof is explicitly omitted ("We omit the details due to space constraints"), and the variational derivation after (16) is also skipped ("we skip the details here"). The result is not immediate because q_r in (19) itself depends on the current π via p(Z_r,t | s_t, a_t), so the alternating scheme is a fixed-point iteration; no convergence or local-optimality proof is provided. Since this is the mechanism behind the claimed Pareto-optimal global policy, the gap is material.
  3. [Abstract, §IV, Eq. (12) and Fig. 2] The headline 26% reduction in emissions per bit is partly by construction. The reward Q1 in (14) maximizes S(t), and S(t) in (12) is a monotone transform of the ratio Σ_k Cb_k(t)/Cb_ψ(t); hence the agent is trained specifically to minimize emissions per bit. Comparing this agent with baselines whose reward does not contain that term will naturally produce a reduction in that same metric. The non-circular content of the claim is that energy and delay are matched and that the reduction is statistically meaningful. However, Figs. 2 and 3 show no error bars or multiple-seed variance, and the paper does not report whether constraints (13d)–(13g) are actually satisfied. Please report the achieved energy and delay values, confidence intervals, and constraint-violation rates.
minor comments (5)
  1. [Eq. (14)] Q2 contains a double negative "--lim" before the first term; this is presumably a typo and should be a single minus sign.
  2. [Eq. (6) and preceding rate expression] The interference index is written as "i≠k" in one place and "i≠u" in another; the notation for channel coefficients also switches between h_u and h_k. Please make the indices consistent.
  3. [Fig. 4] The horizontal-axis label appears corrupted ("10-4" appears in the axis title), and the delay units are unclear. Please fix the axis labels and state whether delay is in milliseconds or seconds.
  4. [Algorithm 1 and §III-A] The LSTM architecture, prediction horizon T, number of epochs, and the update rule for the Lagrange multipliers λ_i are not specified. These details are needed for reproducibility and for judging whether the predictive model actually tracks (11) faithfully.
  5. [Related work, §I-A] Reference [8] already defines network carbon intensity as the ratio of total carbon emissions to total data traffic; the distinction between Cb_k and NCI should be stated more explicitly, since the novel content appears to be the joint optimization enabled by the metric rather than the metric itself.

Circularity Check

1 steps flagged · score 6.0 of 10

Headline 26% emissions-per-bit reduction is the optimized objective itself; MORL state excludes variables needed for the MDP.

  1. self definitional [Section II-C Eq. (12) and Section IV (results)]
    "S(t) = (1 - Σ_{k=1}^M Cb_k(t)/Cb_ψ(t))^2 ... our proposed solution can achieve a 26% lower average carbon emissions for varying packet arrival rates in comparison to the baselines, while achieving the same energy efficiency and average delay values."

    The first objective in (13a) is to maximize the long-term average of S(t). By (12), S(t) is a squared decreasing function of the emissions-per-bit ratio (Σ Cb_k / Cb_ψ), so a policy trained to maximize S(t) is, by construction, a policy trained to minimize emissions per bit. Reporting a 26% reduction in emissions per bit compared to baselines that do not include this objective is therefore a measure of on-objective performance, not an independent prediction. The claim would only be non-circular if energy efficiency and delay were demonstrated to be unchanged; the paper asserts this but provides no supporting data.

full rationale

The main quantitative claim—26% lower emissions per bit—reduces by construction to the reward function being optimized: S(t) is a monotone transform of the emissions-per-bit ratio, so maximizing it is equivalent to minimizing emissions per bit. This is a partial circularity, not a validated out-of-sample prediction. The paper's Pareto-optimality claim is further weakened by a non-circular correctness gap: §III-A defines the state as only emissions per bit and asserts deterministic transitions via (11), but the constraints and delay objective depend on battery levels, queue backlogs, and stochastic channels, so the Bellman optimality argument does not hold as stated. There are no self-citations or imported uniqueness theorems; the variational derivation is borrowed from [18], an external source, and Lemma 1's proof is omitted but standard. Because the central 26% figure is the value of the optimized objective, and the equal-energy/delay assertion is unsupported, the circularity score is 6.

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

The central claim rests on three sets of unstated parameters (emission factors, compute-energy constants, MORL hyperparameters), an assumed Markov state that omits queue and battery dynamics driving two of the three objectives, and unverified carry-over of the [18] variational guarantees. No new physical entities are postulated; the sustainability metric S(t) is a derived quantity, not an entity.

free parameters (3)
  • Emission factors w_u, w_g, w_d = not reported in Table I
    Eq. (11) requires these factors to compute emissions and the sustainability metric S(t). Section II-C declares them unknown and time-varying, yet the reward Q1 in (14) and every reported emissions-per-bit value depend on their values. No values or inference procedure are given.
  • Computing energy model constants (alpha_k, beta_k, A_s, B_s, C_s) = not reported in Table I
    Eq. (4) defines CPU/GPU energy consumption using these architecture-dependent and application-dependent constants. Table I omits them, so the computing component of (7) and hence the 26% result are not reproducible.
  • MORL hyperparameters (zeta_r, rho_r, lambda_i), look-ahead horizon T, LSTM configuration = not reported
    Section III-A uses preference weights zeta_r, temperature rho_r in (20), bisection-computed multipliers lambda_i, and an LSTM for the approximate model P-hat. None are specified, so the Fig. 4 tradeoff points and the convergence claims cannot be re-run.
assumptions (4)
  • ad hoc to paper The emissions-per-bit vector is a sufficient state for the multi-objective MDP, and state transitions are deterministic via Eq. (11)
    Section III-A defines s_t as only (Cb_1(t), ..., Cb_M(t)) and claims transitions are deterministic following (11). Queue backlogs, battery levels, and channel and arrival randomness drive the delay and energy objectives but are absent from the state, so the MDP is potentially non-Markov.
  • domain assumption The variational derivation of [18] carries over to this constrained wireless setting, including the KL-based local optimality argument
    Section III-A: 'We have followed similar derivation as in [18] but extended to a wireless network model and hence, we skip the details here.' The paper then invokes convexity of KL to assert a local optimum for non-convex policy search, which does not follow from KL convexity alone.
  • standard math Little's law yields the long-run average queuing delay in (10)
    Eqs. (9)-(10) invoke Little's law [15]; standard under stationarity assumptions, but the paper does not verify those assumptions for the simulated arrival and service processes.
  • domain assumption Capacity-achieving Gaussian downlink with known channels as in (5)-(6)
    Section II-B models the downlink as capacity-achieving Gaussian signaling with known effective channel coefficients h_u(t); this assumes perfect CSI and Gaussian codebooks, which is standard but an idealization.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Next-Generation Sustainable Wireless Systems: Energy Efficiency Meets Environmental Impact." pith.science (2026). https://pith.science/paper/4ZYJSMKN

@misc{pith2026250902395,
  author       = {Pith},
  title        = {Pith review of: Next-Generation Sustainable Wireless Systems: Energy Efficiency Meets Environmental Impact},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4ZYJSMKN}},
  note         = {Machine review of arXiv:2509.02395}
}
read the original abstract

Aligning with the global mandates pushing towards advanced technologies with reduced resource consumption and environmental impacts, the sustainability of wireless networks becomes a significant concern in 6G systems. To address this concern, a native integration of sustainability into the operations of next-generation networks through novel designs and metrics is necessary. Nevertheless, existing wireless sustainability efforts remain limited to energy-efficient network designs which fail to capture the environmental impact of such systems. In this paper, a novel sustainability metric is proposed that captures emissions per bit, providing a rigorous measure of the environmental footprint associated with energy consumption in 6G networks. This metric also captures how energy, computing, and communication resource parameters influence the reduction of emissions per bit. Then, the problem of allocating the energy, computing and communication resources is posed as a multi-objective (MO) optimization problem. To solve the resulting non-convex problem, our framework leverages MO reinforcement learning (MORL) to maximize the novel sustainability metric alongside minimizing energy consumption and average delays in successfully delivering the data, all while adhering to constraints on energy resource capacity. The proposed MORL methodology computes a global policy that achieves a Pareto-optimal tradeoff among multiple objectives, thereby balancing environmental sustainability with network performance. Simulation results show that the proposed approach reduces the average emissions per bit by around 26% compared to state-of-the-art methods that do not explicitly integrate carbon emissions into their control objectives.

Figures

Figures reproduced from arXiv: 2509.02395 by the authors.

Figure 1
Figure 1. System model of the proposed sustainable wireless network. functions potentially requiring artificial intelligence. This vari￾ation in computational complexity motivates distributing the tasks across different servers or computing clusters to ensure efficient utilization of computing resources and energy [10]. As such, the computation and transmission loads as described by the operations in (1) are split among M CSs… view at source ↗
Figure 2
Figure 2. Emissions per bit versus mean packet arrival rate 0 10 20 30 40 50 60 110 115 120 125 130 135 140 145 150 Proposed MORL Approach Baseline: Optimizing Energy Efficiency [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Emissions per bit vs time for a fixed packet arrival rate. as shown in (11). In fact, our proposed solution can achieve a 26% lower average carbon emissions for varying packet arrival rates in comparison to the baselines, while achieving the same energy efficiency and average delay values [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

18 extracted references · 17 canonical work pages

  1. [8]

    Carbon Data Intensity for Network Energy Performa nce Moni- toring,

    ITU, “Carbon Data Intensity for Network Energy Performa nce Moni- toring,” Recommendation ITU-T L.1333 , 2022

    work page 2022

  2. [1]

    Envisioning the UN Sustainable Development Goals (SDGs) through the lens of energy sustainability (SDG 7) in the post-COVID-19 world,

    R. M. Elavarasan et al., “Envisioning the UN Sustainable Development Goals (SDGs) through the lens of energy sustainability (SDG 7) in the post-COVID-19 world,” Applied Energy , vol. 292, pp. 116665, 2021

    work page 2021

  3. [2]

    Towards sus tainable edge computing through renewable energy resources and onli ne, dis- tributed and predictive scheduling,

    G. Perin, M. Berno, T. Erseghe, and M. Rossi, “Towards sus tainable edge computing through renewable energy resources and onli ne, dis- tributed and predictive scheduling,” IEEE Transactions on Network and Service Management , vol. 19, no. 1, pp. 306–321, 2021

    work page 2021

  4. [3]

    Online resource management of heterogeneous cellular networks po wered by grid-connected smart micro grids ,

    L. Liu, Z. Zhang, N. Wang, H. Zhang, and Y . Zhang, “Online resource management of heterogeneous cellular networks po wered by grid-connected smart micro grids ,” IEEE Transactions on Wireless Communications, vol. 21, no. 10, pp. 8416–8430, 2022

    work page 2022

  5. [4]

    Multi-user Goal-oriented Communications with Energy-efficient Edge R esource Management,

    F. Binucci, P . Banlli, P . D. Lorenzo, and S. Barbarossa, “ Multi-user Goal-oriented Communications with Energy-efficient Edge R esource Management,” IEEE Transactions on Green Communications and Networking, May 2023

    work page 2023

  6. [5]

    Joint Optimization of Internet-of- Things and Smart Grid for Energy Generation, Battery (Dis) charging, and Inf ormation Delivery,

    L. Qi, B. Wu, et al., “Joint Optimization of Internet-of- Things and Smart Grid for Energy Generation, Battery (Dis) charging, and Inf ormation Delivery,” IEEE Internet of Things Journal , vol. 19, no. 1, 2024

    work page 2024

  7. [6]

    Waste Factor and Waste Figure: A Unified Theory for Modeling and Analyzing Wasted Power in Radio Access Networks for Improved Sustainability

    T. S. Rappaport, M. Ying, N. Piovesan, A. D. Domenic, and D . Shakya, “Waste factor and waste figure: A unified theory for modeling a nd analyzing wasted power in radio access networks for improve d sus- tainability,” arXiv preprint arXiv:2405.07710 , 2024

    work page Pith review arXiv 2024

  8. [7]

    Evolution of Sustainability Indicators f or Next- Generation Radio Network Technologies ,

    White Paper, “Evolution of Sustainability Indicators f or Next- Generation Radio Network Technologies ,” NextG Alliance , 2024

    work page 2024

Show all 18 references
  1. [9]

    Optimal des ign of hybrid renewable energy systems in buildings with low to h igh renewable energy ratio,

    M. Sharafi, T. Y . ElMekkawy, and E. L. Bibeau, “Optimal des ign of hybrid renewable energy systems in buildings with low to h igh renewable energy ratio,” Renewable Energy, , no. 83, 2015

    2015

  2. [10]

    Splitwise: Efficient generative llm inferen ce using phase splitting,

    P . Patel, E. Choukse, C. Zhang, A. Shah, I. Goiri, S. Male ki, and R. Bianchini, “Splitwise: Efficient generative llm inferen ce using phase splitting,” in Proceedings of ACM/IEEE 51st Annual International Symposium on Computer Architecture (ISCA) , Jun. 2024

    2024

  3. [11]

    All ocation of computing tasks in distributed MEC servers co-powered by re newable sources and the power grid,

    D. Cecchinato, M. Berno, F. Esposito, and M. Rossi, “All ocation of computing tasks in distributed MEC servers co-powered by re newable sources and the power grid,” in Proceedings of IEEE International Conference on Acoustics, Speech and Signal Processing , May 2020

    2020

  4. [12]

    On the Effects of Battery Imperf ections in an Energy Harvesting Device,

    A. Biason and M. Zorzi, “On the Effects of Battery Imperf ections in an Energy Harvesting Device,” in Proceedings of the International Conference on Computing, Networking and Communications (I CNC), May 2016

    2016

  5. [13]

    Energy-efficie nt Resource Management for Federated Edge Learning with CPU-GPU Hetero ge- neous Computing,

    Q. Zeng, Y . Du, K. Huang, and K. K. Leung, “Energy-efficie nt Resource Management for Federated Edge Learning with CPU-GPU Hetero ge- neous Computing,” IEEE Transactions on Wireless Communications , vol. 20, no. 12, pp. 7947–7962, 2021

    2021

  6. [14]

    Power capping of cpu-gpu hetero geneous systems using power and performance models,

    K. Tsuzuku and T. Endo, “Power capping of cpu-gpu hetero geneous systems using power and performance models,” in Proceedings of International Conference on Smart Cities and Green ICT Syst ems, Jun. 2015

    2015

  7. [15]

    A Proof for the Queuing Formula: L= λ w ,

    J. D. Little, “A Proof for the Queuing Formula: L= λ w ,” Operations research, vol. 9, no. 3, pp. 383–387, 1961

    1961

  8. [16]

    Electric efficie ncy indicators and carbon dioxide emission factors for power generation by fossil and renewable energy sources on hourly basis,

    E. Marrasso, C. Roselli, and M. Sasso, “Electric efficie ncy indicators and carbon dioxide emission factors for power generation by fossil and renewable energy sources on hourly basis,” Energy Conversion and Management, , no. 196, pp. 1369–1384, 2019

    2019

  9. [17]

    CO2 emissions from renewable and non-renewable electrici ty gener- ation sources in the G7 countries: static and dynamic panel a ssessment ,

    L. C. V oumik, M. A. Islam, S. Ray, N. Y . Mohamed, and A. R. R idzuan, “ CO2 emissions from renewable and non-renewable electrici ty gener- ation sources in the G7 countries: static and dynamic panel a ssessment ,” Energies, vol. 16, no. 3, 2023

    2023

  10. [18]

    On Multi-objective Policy Opti mization as a Tool for Reinforcement Learning: Case Studies in Offline RL and Finetuning,

    A. Abdolmaleki, S. Huang, G. V ezzani, B. Shahriari, J. T . Springenberg, S. Mishra, and D. Tirumala, “On Multi-objective Policy Opti mization as a Tool for Reinforcement Learning: Case Studies in Offline RL and Finetuning,” arXiv preprint arXiv:2106.08199 , 2021

    2021 arXiv

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.