{"id":"5ef1d9cc-97dc-4983-a1eb-1af17b32e7dc","arxiv_id":"1906.09762","paper_version":1,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":5.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":0,"one_line_summary":"Derives a closed-form multi-level water-filling computation offloading policy for delay minimization in MEC systems by approximating an MDP to a virtual continuous time system and using dynamic rate estimation.","lead":"This paper models delay-optimal computation offloading in mobile edge computing systems with coupled queues at mobile terminals and servers as an MDP, then approximates it to derive a closed-form multi-level water-filling policy. A smart generalist might read it to see how queue-aware offloading decisions can be made in real time for lower latency in wireless edge networks.","discovery_kind":"new_method","skeptic_critique":{"model":"grok-4.3","headline":"VCTS approximation + dynamic rate estimation step lacks error bounds or validation against original MDP","rationale":"The reader's weakest_assumption directly identifies the same technical hinge (VCTS + rate estimation). Because the full text was not previously accessible, the current pass confirms that this step remains the least-secured link in the argument; simulations alone cannot substitute for an error analysis of the approximation itself. This moves the verdict from UNVERDICTED to CONDITIONAL pending the proposed check.","tokens_in":1766,"tokens_out":346,"duration_ms":10411,"concrete_test":"For the single-MT single-server case, compute the exact average-cost MDP value function via value iteration on a truncated state space (queues up to 20) and compare it to the VCTS-derived priority function evaluated at the same states; if the policy extracted from the approximate priorities yields >15% higher average cost than the exact MDP policy on the same instances, the closed-form claim does not hold.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central claim rests on moving from the discrete-time MDP (with coupled LQSI/RQSI queues) to a VCTS with reflections, then applying dynamic instantaneous rate estimation to obtain closed-form approximate priority functions that enable the multi-level water-filling policy. No section provides a rigorous error bound, convergence rate, or regime where the approximation error remains small; the derivation instead invokes the VCTS and rate estimation directly to claim closed-form optimality. If the instantaneous-rate step deviates materially from the true discrete dynamics (especially under cascade queue coupling), the resulting priority functions and water-filling solution lose their claimed delay-optimality guarantee.","agreement_with_reader":"agree"},"referee_report":{"model":"grok-4.3","summary":"The paper models delay-optimal computation offloading in MEC systems with cascade-coupled queues at the mobile terminal and MEC server as an infinite-horizon average-cost MDP. It approximates the discrete MDP by a virtual continuous-time system (VCTS) with reflections, applies dynamic instantaneous rate estimation to obtain closed-form approximate priority functions, and derives a multi-level water-filling offloading policy that incorporates both LQSI and RQSI. The approach is extended to multi-MT/multi-MEC scenarios, with simulation results claimed to outperform conventional schemes.","tokens_in":1875,"tokens_out":547,"duration_ms":11639,"significance":"If the VCTS approximation and rate-estimation step are shown to be accurate, the closed-form multi-level water-filling policy would offer a tractable, analytically insightful solution for delay minimization under queue coupling, which is a practically relevant extension beyond single-queue MEC offloading analyses.","major_comments":[{"comment":"§4 (VCTS approximation step): the manuscript invokes the VCTS with reflections to replace the original discrete-time MDP but supplies no error bound, convergence rate, or regime of validity for the approximation error relative to the coupled LQSI/RQSI dynamics; without this, the subsequent priority functions cannot be guaranteed to inherit delay optimality.","section":"§4"},{"comment":"§5 (dynamic instantaneous rate estimation): the derivation of the closed-form approximate priority functions relies on this step, yet no quantitative comparison (e.g., mean-squared deviation or policy-value gap) is provided against the true discrete MDP transition probabilities, especially under cascade queue coupling; this is load-bearing for the central claim of closed-form optimality.","section":"§5"},{"comment":"Simulation section (results and baselines): outperformance is asserted without reported error bars, sensitivity to rate-estimation parameters, or explicit description of the conventional schemes and MDP solver used for comparison, preventing assessment of whether the gains are attributable to the proposed policy or to the unvalidated approximation.","section":"Simulation results"}],"minor_comments":[{"comment":"Notation for the priority functions and water-filling thresholds should be introduced with explicit dependence on the estimated rates to avoid ambiguity when moving between the VCTS and the original MDP.","section":"§5"},{"comment":"The extension to multiple MTs/MECs in the final section would benefit from a brief statement of how the per-user priority functions are coordinated across servers.","section":"Multi-user extension"}],"recommendation":"major_revision","confidential_remarks":null},"author_rebuttal":{"model":"grok-4.3","summary":"We thank the referee for the constructive feedback on our manuscript. The comments highlight important aspects regarding the theoretical justification of our approximations and the presentation of simulation results. We address each major comment below and indicate the revisions we will make.","responses":[{"response":"The VCTS approximation is employed to enable tractable closed-form derivations for the priority functions, consistent with fluid-limit techniques used in related MDP analyses of queueing systems. The resulting policy is presented as an approximate solution rather than an exactly optimal one. We will revise §4 to explicitly clarify the approximate nature of the approach, discuss the expected validity regime (e.g., large queue lengths or high system load), and add a remark noting that a rigorous error bound remains an open direction for future work.","revision_made":"partial","referee_comment":"[§4] §4 (VCTS approximation step): the manuscript invokes the VCTS with reflections to replace the original discrete-time MDP but supplies no error bound, convergence rate, or regime of validity for the approximation error relative to the coupled LQSI/RQSI dynamics; without this, the subsequent priority functions cannot be guaranteed to inherit delay optimality."},{"response":"We acknowledge the absence of direct quantitative validation for the rate-estimation step against the exact MDP. In the revision, we will add a new subsection or appendix that provides numerical comparisons in small-scale instances (where the true MDP can be solved exactly), including metrics such as the deviation in priority function values and the resulting average delay gap between the approximate policy and the optimal MDP policy. This will strengthen the evidence for the accuracy of the estimation under cascade coupling.","revision_made":"yes","referee_comment":"[§5] §5 (dynamic instantaneous rate estimation): the derivation of the closed-form approximate priority functions relies on this step, yet no quantitative comparison (e.g., mean-squared deviation or policy-value gap) is provided against the true discrete MDP transition probabilities, especially under cascade queue coupling; this is load-bearing for the central claim of closed-form optimality."},{"response":"We agree that the simulation results section requires additional details for reproducibility and robustness assessment. In the revised manuscript, we will: include error bars or standard deviations for all plotted metrics across multiple random seeds; provide explicit descriptions of the conventional schemes (e.g., threshold-based and myopic policies) and the numerical MDP solver employed; and add sensitivity plots with respect to key rate-estimation parameters. These changes will allow readers to better evaluate the source of the observed performance gains.","revision_made":"yes","referee_comment":"[Simulation results] Simulation section (results and baselines): outperformance is asserted without reported error bars, sensitivity to rate-estimation parameters, or explicit description of the conventional schemes and MDP solver used for comparison, preventing assessment of whether the gains are attributable to the proposed policy or to the unvalidated approximation."}],"tokens_in":1438,"tokens_out":666,"duration_ms":22523,"standing_objections":["A formal derivation of error bounds or convergence rates for the VCTS approximation under cascade-coupled LQSI/RQSI dynamics would require substantial new theoretical analysis that is beyond the scope of the current work."]},"desk_editor":{"model":"grok-4.3","letter":"The paper models delay-optimal offloading in computation-constrained MEC as an average-cost MDP with cascade-coupled queues, approximates it via a virtual continuous-time system with reflections, derives approximate priority functions through dynamic instantaneous rate estimation, and arrives at a closed-form multi-level water-filling rule that incorporates both LQSI and RQSI. It also sketches an extension to multiple MTs and servers. That specific combination of steps to reach an analytical policy is the concrete new element; most prior MEC work either solves the MDP numerically or ignores the remote queue state. If the approximation is tight, the result would be practically useful for implementation without online optimization. The simulations are said to beat conventional schemes, which is at least consistent with the claim. The central weakness is exactly the one flagged in the stress test: the derivation treats the VCTS and rate-estimation steps as directly yielding optimality without any error bound, convergence rate, or numerical check against the discrete MDP. When queue coupling is strong, even modest deviation in the instantaneous-rate estimate could invalidate the priority functions and therefore the water-filling thresholds. No sensitivity analysis or regime of validity appears in the abstract, and the lack of error bars or baseline details in the reported simulations leaves the performance claims hard to interpret. This is the kind of paper that belongs in a specialized wireless resource-allocation venue. Readers working on analytical approximations for edge-computing queues could extract the derivation steps even if they later have to supply their own validation. It is worth sending to peer review so that referees can require the missing error analysis and simulation details; the technical machinery is non-trivial and the problem is relevant, but the current evidence for the optimality claim is thin.","headline":"The closed-form multi-level water-filling policy for coupled-queue MEC offloading rests on a VCTS approximation plus dynamic rate estimation whose error relative to the original MDP is never bounded or validated.","tokens_in":2402,"tokens_out":419,"would_cite":false,"duration_ms":19066,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":{"model":"grok-4.3","evidence":[],"headline":"Standard MDP-to-VCTS control derivation with rate estimation; no RS cost, ratio symmetry or forcing structure","alignment":"orthogonal","rationale":"The paper's core machinery (infinite-horizon average-cost MDP approximated by reflected VCTS, HJB PDE solved via dynamic instantaneous rate estimation to obtain closed-form priority functions and multi-level water-filling) is conventional stochastic control. It contains no J-cost functional, reciprocal symmetry, golden-ratio identities, 8-tick periodicity, or parameter-free constant derivations. The cascade-queue balance is handled by ad-hoc rate-difference estimation rather than any RS-style cost minimization. Hence the construction is orthogonal to the RS forcing chain.","tokens_in":63019,"confidence":"high","tokens_out":149,"duration_ms":5398,"cache_read_input_tokens":38528,"cache_creation_input_tokens":0},"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"grok-4.3","headline":"A closed-form multi-level water-filling solution optimizes delay in MEC offloading by using both local and remote queue states.","keywords":["computation offloading","mobile edge computing","delay optimization","Markov decision process","water-filling","queue state information","virtual continuous time system"],"falsifier":"Compare average delay of the proposed policy against the optimal policy obtained by solving the original MDP (or a high-fidelity simulator) in a scenario where remote queue length varies rapidly while local queue length is held fixed.","tokens_in":2651,"feed_emoji":"📡","tokens_out":649,"duration_ms":13701,"temperature":0.7,"pith_summary":"The paper models computation offloading in resource-constrained mobile edge computing as an infinite-horizon average-cost MDP where the task queue at the mobile terminal and the constrained queue at the MEC server are coupled in cascade. It approximates the discrete MDP by a virtual continuous time system with reflections and derives closed-form approximate priority functions through dynamic instantaneous rate estimation. From these functions the authors construct a multi-level water-filling policy that explicitly incorporates both local queue state information and remote queue state information. The same construction is extended to the multi-user multi-server case. A reader would care because the resulting policy is closed-form and therefore implementable without solving the original MDP at runtime.","feed_headline":"Water-filling offloading uses both local and remote queues for delay optimality","feed_subtitle":"Closed-form multi-level solution accounts for coupled queues at mobile terminals and MEC servers.","key_machinery":"Closed-form multi-level water-filling computation offloading solution derived from approximate priority functions of the virtual continuous time system approximation.","core_discovery":"A closed-form multi-level water-filling computation offloading solution, obtained from approximate priority functions derived via dynamic instantaneous rate estimation in the virtual continuous time system, characterizes the influence of both local and remote queue state information and thereby achieves delay optimality in computation-constrained MEC systems.","pith_inferences":["The same rate-estimation-plus-water-filling pattern could be tested on other cascade-queue resource-allocation problems that are currently solved only by value iteration.","Because the policy is closed-form, it could be embedded in low-power firmware without requiring an online MDP solver.","The approximation quality may degrade when channel coherence time becomes comparable to the MDP decision interval; that regime could be checked by varying Doppler spread in simulation."],"forward_implications":["The policy extends directly from single-MT single-server to multiple-MT multiple-server scenarios.","Offloading decisions now depend on both local and remote queue lengths rather than local information alone.","Several explicit insights on the relative weighting of LQSI and RQSI follow from the water-filling levels.","Simulation results indicate lower average delay than conventional schemes that ignore the remote queue."],"fun_headline_variants":["Closed-form water-filling uses local and remote queue state info","Multi-level water-filling for delay-optimal MEC with coupled queues","Approximate priority functions enable closed-form offloading policy","Delay-optimal offloading solution incorporates LQSI and RQSI"],"cache_read_input_tokens":2112,"weakest_assumption_plain":"The dynamic instantaneous rate estimation step yields sufficiently accurate closed-form approximate priority functions for the virtual continuous time system approximation of the original MDP.","fun_headline_variants_meta":{"raw":{"variants":["Closed-form water-filling uses local and remote queue state info","Multi-level water-filling for delay-optimal MEC with coupled queues","Approximate priority functions enable closed-form offloading policy","Delay-optimal offloading solution incorporates LQSI and RQSI"]},"model":"grok-4.3","cost_usd":0.004689,"raw_usage":{"total_tokens":2310,"prompt_tokens":654,"num_sources_used":0,"completion_tokens":61,"cost_in_usd_ticks":46887000,"prompt_tokens_details":{"text_tokens":654,"audio_tokens":0,"image_tokens":0,"cached_tokens":256},"completion_tokens_details":{"audio_tokens":0,"reasoning_tokens":1595,"accepted_prediction_tokens":0,"rejected_prediction_tokens":0}},"tokens_in":654,"tokens_out":61,"duration_ms":11976,"temperature":1.0,"reasoning_tokens":1595,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-05-25T17:41:13.237556+00:00","model_set":{"reader":"grok-4.3"},"falsifier":"Compare average delay of the proposed policy against the optimal policy obtained by solving the original MDP (or a high-fidelity simulator) in a scenario where remote queue length varies rapidly while local queue length is held fixed.","supporting_citations":[],"review_version":1}