{"id":"67235d45-cbb4-49d2-bf7e-06d42df9a275","arxiv_id":"1908.02429","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"For status update systems over block-fading channels, the paper derives the average age of information under NACK-count-based power control and proposes a simulated annealing search for power policies that reduce age compared to constant power.","lead":"This paper studies how a status-updating transmitter should vary its power after repeated failed deliveries to keep the received information as fresh as possible. It derives a closed-form age-of-information formula and shows, by simulation, that a power control policy can cut the average age dramatically in low-power settings.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 1 is sound under the stated model; the load-bearing gap is that the 'age-optimal' title claim rests on a heuristic solver with no optimality guarantee.","rationale":"I stress-tested Theorem 1 first because it is the paper's main formal result. The renewal-reward proof in Appendix A is correct: the age process resets at each success, the inter-success intervals Y_k are i.i.d. under the stated independent block-fading model, and the identities E[Y]=Σξ_j and E[Y^2]=2Σjξ_j+Σξ_j follow by telescoping (Pr{Y=m}=ξ_{m-1}-ξ_m). The special-case check in Remark 2 is algebraically wrong: the correct constant-power AoI is (3-p)/(2(1-p)), not (3-2p)/(2(1-p)); this is a typo, not a flaw in Theorem 1. The reader's weakest assumption (independent fading) is explicit in Section II-A and used in Eq. (3); it limits scope to i.i.d. block fading but is not internally inconsistent. The concern I find load-bearing is the gap between the paper's 'Age-Optimal' title and the actual optimization evidence. Problem (11) is non-convex, and Algorithm 1 is a heuristic random search with no convergence proof, no feasibility repair for ε' outside [0,1], an undefined temperature counter, and no code. The numerical comparison to constant power supports a heuristic-improvement claim, not an optimality claim. The paper's own abstract limits itself to 'a feasible solution,' so the title overstates the contribution. This does not invalidate Theorem 1, but it should keep the verdict at CONDITIONAL, pending reframing or an explicit optimality benchmark.","tokens_in":6704,"tokens_out":15983,"duration_ms":174563,"concrete_test":"Run a truncated global-optimality benchmark: set M=10 (states 0..9), Rayleigh fading with unit mean, R=1, and Pbar=10^{-0.6}. Use a multistart global optimizer (e.g., differential evolution with population 10^4 plus local refinement) over P_0,...,P_9 with P_m≥0 and ∑P_m ξ_m/∑ξ_j ≤ Pbar, evaluating Δ from Eq. (9); compare its best Δ with Algorithm 1's output on the same instance. If the benchmark finds a lower Δ, the proposed solver is not age-optimal. If it matches, repeat at Pbar=10^{-1} and Pbar=10^{-0.2} to confirm the title claim.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The closed-form AoI in Theorem 1 is correctly derived from the renewal structure: after each success the Markov chain resets to state 0, and under the stated i.i.d. block-fading assumption the failure events are independent, so E[Y]=Σξ_j and E[Y^2]=2Σjξ_j+Σξ_j. The reader's independence concern is a scope limitation rather than an internal flaw. The genuinely load-bearing gap concerns the 'age-optimal' claim in the title and in Section III. Problem (11) is non-convex, and Algorithm 1 is a simulated-annealing/evolutionary random search with no proof of convergence to a global (or even local) optimum. It generates ε' by adding Cauchy noise, never clamps ε' to [0,1], accepts or rejects based on a temperature schedule with an undefined counter n, and checks the average-power constraint only at evaluated points. No KKT or optimality certificate is supplied. Consequently, the numerical 80% AoI reduction shows only that the heuristic beats constant-power on one simulation instance; it does not demonstrate that any solution of (11) is age-optimal. The paper's own abstract says only 'a feasible solution,' so the title overstates what is proven.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper considers a status update system in which the source transmits one fixed-rate packet per slot over a block fading channel and can adapt transmit power based on the number of consecutive NACKs received. The main contribution is a closed-form expression for the average AoI (Theorem 1) in terms of the outage probabilities ε_m. Based on this expression, the authors formulate a non-convex, average-power-constrained AoI minimization problem (11) and propose a simulated-annealing/evolutionary search algorithm (Algorithm 1), together with a simpler on-off power policy. Numerical results over Rayleigh fading show that the proposed policies reduce the average AoI relative to fixed-power transmission, particularly in the low-power regime.","tokens_in":6956,"tokens_out":5926,"duration_ms":58521,"significance":"If Theorem 1 stands, it provides a clean and useful objective for NACK-based power adaptation, which was not available in closed form for this model. The paper validates the formula against simulation and identifies a simple on-off policy that nearly matches the heuristic search, which is a useful engineering insight. However, the optimality claim in the title is not supported: Algorithm 1 is a heuristic with no convergence or optimality certificate, and the algorithm as stated contains undefined variables and does not enforce that candidate probabilities lie in [0,1]. The closed-form derivation and numerical comparison are nevertheless valuable, and the issues are correctable by reframing the claims and tightening the algorithm.","major_comments":[{"comment":"The paper's title and Section III claim 'age-optimal power control', but Algorithm 1 is a stochastic search with no proof of convergence to a global or even local optimum of (11). It generates candidates via Cauchy perturbations in (12) without enforcing ε'∈[0,1], and it checks the average-power constraint only at evaluated points without a feasibility certificate for the continuous problem. The abstract's phrase 'a feasible solution' is the accurate description. Please revise the title and main claims to reflect that the contribution is a heuristic with empirical improvement, or provide an optimality/convergence analysis.","section":"Section III, Eq. (11), Algorithm 1"},{"comment":"The fixed-power average AoI is computed incorrectly. Substituting ξ_j = p^j into (9) yields Δ = 3/2 + p/(1-p) = (3-p)/(2(1-p)), not (3-2p)/(2(1-p)). This baseline is used in Section IV, so the displayed formula and any theoretical fixed-power curve derived from it must be corrected.","section":"Remark 2"},{"comment":"The temperature schedule in line 5, Tn = T0/n, uses an undefined counter n; as written, n is not initialized or incremented, so the temperature does not decrease. Additionally, lines 7–8 are underspecified: what is 'A' and how is the average power P̂ computed from it? Because candidate ε' is not clamped to [0,1], the Markov chain in (1) may be invalid for generated candidates. Please specify the algorithm precisely and add bounds on ε'.","section":"Algorithm 1"}],"minor_comments":[{"comment":"References [8] and [9] are identical in title and venue; one should be removed or corrected.","section":"References"},{"comment":"Reference [12] has incomplete publication data (\"vol. pp, no. 99\"); please provide the full citation.","section":"References"},{"comment":"The system model states that 'no retransmission is needed' yet the NACK-based power adaptation implies the source tracks failed packets; clarify that each slot transmits a new packet and NACKs only determine the power level.","section":"Section II-A"},{"comment":"The on-off power policy's switch point τ is not indicated in Fig. 3(b); adding a marker or annotation would improve readability.","section":"Fig. 3(b)"}],"recommendation":"major_revision","confidential_remarks":"The paper is in scope for the journal and Theorem 1 appears correct. The primary concern is the overclaiming 'optimal' in the title; I would be satisfied with a revision that reframes the contribution as a heuristic and fixes the algorithm details, the Remark 2 error, and the reference issues."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nThe one thing to know: the paper gives a clean closed-form expression for average AoI when transmit power is chosen based on the number of consecutive NACKs. Theorem 1 is correct under the stated i.i.d. block-fading model, and the result is worth having. The 80% AoI reduction in the low-power regime is a real numerical finding, but it comes from a heuristic optimizer, not from an age-optimal proof. The title oversells what is established.\n\nWhat is new is the state-dependent policy P_m and the formula Δ = 3/2 + (Σ j ξ_j)/(Σ ξ_j) with ξ_j = Π ε_m. The renewal-style derivation in the appendix checks out: after each success the chain resets, E[Y] = Σ ξ_j and E[Y^2] = 2Σ j ξ_j + Σ ξ_j, so (9) follows. The simulation matches the analytic curve, which is decent evidence the modeling is faithful. The on-off policy is a useful practical add-on.\n\nThe soft spots are mostly in the optimization half. Problem (11) is non-convex, and Algorithm 1 is simulated annealing with Cauchy perturbations and a temperature schedule; there is no convergence proof or optimality certificate, so calling it 'age-optimal' is not justified. The abstract is more careful, saying 'a feasible solution,' which suggests the title got ahead of the content. Remark 2 has a small algebraic slip: fixed-power AoI should be (3-p)/(2(1-p)), not (3-2p)/(2(1-p)). The pseudocode also leaves some details loose, e.g., the counter n in the acceptance probability is not defined in the loop, and ε' is not clamped to [0,1] before evaluating the objective. These are minor but should be cleaned up. No code is provided, so the simulation result cannot be reproduced exactly.\n\nThe i.i.d. block-fading assumption is a scope limitation, not an internal flaw. If fading is correlated, the Markov transition structure and the factorization in (2)-(3) break; the authors state this assumption clearly in Section II-B, and the theorem is valid within that model.\n\nThis paper is for people working on AoI for IoT-style status updates over fading channels. The closed-form result is a legitimate contribution and will get cited; the heuristic optimization part needs a caveat or a stronger algorithm before it should be treated as optimal.\n\nRecommendation: yes, send it to peer review. The AoI derivation and simulation validation deserve referee time. Ask the authors to fix Remark 2, clarify the algorithm, and reframe the title so 'age-optimal' means 'low-AoI via a proposed heuristic' unless they can prove otherwise.","headline":"Solid closed-form AoI for NACK-state-dependent power control, but the 'age-optimal' title outruns a heuristic optimizer.","tokens_in":7480,"tokens_out":3297,"would_cite":true,"duration_ms":32955,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["94A15","60J20"],"pacs":[],"model":"deepseek-v4-flash","headline":"Average age of information under NACK-based power control has a closed form, and optimized power cuts age by over 80 percent at low power.","keywords":["age of information","power control","block fading","status update systems","NACK feedback","average power constraint","simulated annealing","on-off power control"],"falsifier":"Run the same NACK-based power policy over a block-fading channel whose gains follow an autoregressive process with correlation coefficient $\\rho$ between consecutive slots instead of independent draws. For $\\rho$ large enough, simulate the average age and compare it with Theorem 1's $\\Delta$: a systematic gap that grows with $\\rho$ would confirm that the independence assumption is load-bearing, whereas at $\\rho = 0$ the formula should match within Monte Carlo error.","tokens_in":6501,"feed_emoji":"📡","tokens_out":6371,"duration_ms":62301,"temperature":0.7,"pith_summary":"This paper asks how a source that sends one fresh status packet per time slot over a fading wireless channel should choose transmit power when it only learns, through NACK feedback, how many packets in a row have failed. It establishes a closed-form expression for the average age of information under such NACK-based power control, and it shows that minimizing age under an average power constraint becomes an optimization over per-state failure probabilities. The central result is that average AoI equals $\\Delta = \\frac{3}{2} + \\frac{\\sum_{j=0}^\\infty j \\xi_j}{\\sum_{j=0}^\\infty \\xi_j}$, where $\\xi_j$ multiplies the failure probabilities of the first $j$ states. The paper reports that the optimized power policy reduces average age by more than 80 percent at low average power compared with constant power, and that a simple on-off policy performs nearly as well.","feed_headline":"Power control cuts status-update age by over 80% at low power","feed_subtitle":"With a closed-form age formula, transmitters can scale power after each failure and keep monitors fresher on fading channels.","key_machinery":"The carrying object is a countable-state Markov chain whose state is the number of consecutive NACKs, with failure probability $\\epsilon_m$ in state $m$ and transition back to state 0 on success. The closed-form AoI follows from $\\xi_j$, the probability of surviving $j$ consecutive failures, which telescopes the renewal intervals $\\tilde{Y}_k$ into the standard formula $\\Delta = 1 + E[\\tilde{Y}_k^2]/(2E[\\tilde{Y}_k])$. Theorem 1 evaluates that ratio as $3/2 + \\frac{\\sum j \\xi_j}{\\sum \\xi_j}$, giving a directly computable objective for the power-allocation search.","core_discovery":"For a status update system in which a new fixed-rate packet is generated every slot, no retransmission is used, and the transmit power $P_m$ is chosen only according to the number $m$ of consecutive NACKs already received, the average age of information over independent block fading is $\\Delta = \\frac{3}{2} + \\frac{\\sum_{j=0}^\\infty j \\xi_j}{\\sum_{j=0}^\\infty \\xi_j}$ with $\\xi_0 = 1$, $\\xi_j = \\prod_{m=0}^{j-1} \\epsilon_m$, and $\\epsilon_m = \\Pr\\{z < (2^R - 1)/P_m\\}$. This makes age a function only of the state-dependent decoding failure probabilities, so power-control design becomes a nonconvex optimization over those probabilities subject to the average power constraint $\\sum_m P_m \\pi_m \\le \\bar{P}$. In Rayleigh fading with $R=1$, the optimized policy jumps transmit power upward after roughly seven consecutive failures, shortening the longest failure runs and cutting average age by more than 80 percent at $\\bar{P} = 10^{-0.6}$ W; simulations confirm the closed form.","pith_inferences":["The numerical threshold at state 7 suggests a conjecture the paper does not prove: for Rayleigh fading and fixed rate, the optimal NACK-based policy may be exactly on-off; proving that would explain why the on-off heuristic performs so well.","The formula is specific to independent block fading; under correlated fading or finite-blocklength coding, the factorization $\\epsilon_m = \\Pr\\{z < (2^R-1)/P_m\\}$ changes, so the closed form should be treated as a baseline rather than a universal law.","The reported 80 percent reduction is demonstrated for a particular Rayleigh scenario with $R=1$ and $\\bar{P}=10^{-0.6}$ W; transferring that number to other channel distributions or rates is an extrapolation, not a claim of the paper."],"forward_implications":["The closed-form age expression gives a tractable objective, so the power-control policy can be designed offline once channel statistics are known, with each search iteration requiring only the evaluation of the formula.","In the low-power regime, the optimized policy concentrates power on later failure states, indicating that breaking long failure streaks matters more than protecting the first transmission attempt.","At high average power, constant-power transmission nearly matches the optimized policy, and both approaches approach the lower bound $\\Delta \\ge 1.5$.","An on-off policy that transmits only after at least $\\tau$ consecutive failures is nearly age-optimal at low power, offering a simple implementation without per-state power tuning."],"supporting_citations":[{"why":"Supplies the reindexing of successful updates and the trapezoid age computation that Theorem 1 builds on.","marker":"[11]"},{"why":"Defines age of information as time since the last successfully received status update, the quantity this paper minimizes.","marker":"[1]"},{"why":"Provides the simulated annealing power-allocation procedure that Algorithm 1 modifies.","marker":"[12]"},{"why":"Gives the fast simulated annealing temperature schedule used in Algorithm 1.","marker":"[13]"},{"why":"Justifies the Cauchy-distributed mutation that speeds convergence in the evolutionary search.","marker":"[14]"}],"fun_headline_variants":["Age-optimal power control cuts status age by 80%","Closed-form AoI formula enables 80% age cut on fading channels","Power scaling after NACKs slashes status age by 80%","Adaptive power after failures trims AoI by 80%","Optimal power jumps cut status-update age 80%"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The derivation assumes the channel fades independently from slot to slot, so the probability that the next attempt fails depends only on the power just used and not on the history of failures; if fading is correlated across slots, the formula does not hold.","fun_headline_variants_meta":{"raw":{"variants":["Age-optimal power control cuts status age by 80%","Closed-form AoI formula enables 80% age cut on fading channels","Power scaling after NACKs slashes status age by 80%","Adaptive power after failures trims AoI by 80%","Optimal power jumps cut status-update age 80%"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000651,"raw_usage":{"total_tokens":2995,"prompt_tokens":961,"completion_tokens":2034,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":577,"completion_tokens_details":{"reasoning_tokens":1953}},"tokens_in":577,"tokens_out":2034,"duration_ms":16740,"temperature":1.0,"reasoning_tokens":1953,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:44:10.251520+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the same NACK-based power policy over a block-fading channel whose gains follow an autoregressive process with correlation coefficient $\\rho$ between consecutive slots instead of independent draws. For $\\rho$ large enough, simulate the average age and compare it with Theorem 1's $\\Delta$: a systematic gap that grows with $\\rho$ would confirm that the independence assumption is load-bearing, whereas at $\\rho = 0$ the formula should match within Monte Carlo error.","supporting_citations":[{"cited_title":"Energy-age tradeoff in stat us update communication systems with retransmission,","cited_arxiv_id":null,"evidence_quote":"Supplies the reindexing of successful updates and the trapezoid age computation that Theorem 1 builds on."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Defines age of information as time since the last successfully received status update, the quantity this paper minimizes."},{"cited_title":"On optimi zing power allocation for reliable communication over fading ch annels with uninformed transmitter,","cited_arxiv_id":null,"evidence_quote":"Provides the simulated annealing power-allocation procedure that Algorithm 1 modifies."},{"cited_title":"Fast simulated annealing,","cited_arxiv_id":null,"evidence_quote":"Gives the fast simulated annealing temperature schedule used in Algorithm 1."},{"cited_title":"Evolutionary programming ma de faster,","cited_arxiv_id":null,"evidence_quote":"Justifies the Cauchy-distributed mutation that speeds convergence in the evolutionary search."}],"review_version":1}