Predefined-Time Distributed Optimal Control for Nonlinear Multiagent Systems via Reinforcement Learning — Review Speed Concern with Self-Consistent Numerical Data

Academic fraud report · Geng Detector

Summary

This report evaluates the article by Nannan Cai, Wei Wu, and Shaocheng Tong published in IEEE Internet of Things Journal (Vol. 13, No. 10; DOI: 10.1109/JIOT.2026.3665672). The verdict is "questionable" (yellow), driven primarily by an anomalously short review timeline rather than evidence of scientific fraud. The paper was received on 14 January 2026 and accepted on 12 February 2026 — a 29-day turnaround — which is inconsistent with the typical 3–6 month review cycle expected of a top-tier Q1 journal, given the manuscript's length, complex Lyapunov/HJB derivations, and heavy formula typesetting. By contrast, the numerical simulation data was independently re-derived from the reported initial conditions and compared against the authors' reported cost function values J_i. The recomputed theoretical optimal values V_i were 0.0566, 0.1312, and 0.5674, matching the authors' reported values of 0.0572, 0.1308, and 0.5694 to within negligible numerical discrepancy, confirming self-consistency. Image-level forensic analysis could not be performed due to lack of raw figures. Limitations include reliance on text-only extraction and inability to verify peer-review records.

Verdict

🟡 Questionable. The manuscript is internally self-consistent on the numerical points examined, but the publication timeline raises serious concerns about the integrity of the review process. No direct evidence of data fabrication, image manipulation, or plagiarism was found.

Key findings

Anomalous review timeline (high suspicion): Received 14 January 2026; accepted 12 February 2026 — a 29-day interval. For a 10-page manuscript in IEEE Internet of Things Journal (CAS Q1 / JCR Q1) containing extensive Lyapunov and HJB analysis, this turnaround is extraordinary and warrants scrutiny of the editorial and reviewer process.
Self-consistent numerical data (cleared): Recomputation of the initial tracking-error-derived cost function values from the reported initial conditions produced values that match the authors' reported simulation outcomes to within ~0.001–0.002, consistent with minor discretization/rounding effects.
Future-dated reference check (cleared): Multiple 2025 references are reasonable given a 2026 submission window.
Image forensics (untestable): Pixel-level inspection of Figs. 3–10 was not possible from the text-only PDF extraction; no duplicate-image or splicing conclusion can be drawn.

Evidence highlights

Reported cost-function values at T_c = 2 s: J_1 = 0.0572, J_2 = 0.1308, J_3 = 0.5694.
Recomputed from initial states x_0(0) = [0.005, −0.005, 0.002]^T, x_1(0) = [0.15, −0.18, 0.25]^T, x_2(0) = [0.65, −0.23, 0.15]^T, x_3(0) = [0.85, −1.05, 0.80]^T and the graph topology in Fig. 2:
δ_1 = x_1 − x_0 → V_1 = 0.5·δ_1^T δ_1 = 0.056577
δ_2 = x_2 − x_1 → V_2 = 0.5·δ_2^T δ_2 = 0.13125
δ_3 = x_3 − x_2 → V_3 = 0.5·δ_3^T δ_3 = 0.56745
Absolute differences between reported and recomputed values are on the order of 10⁻³, well within the range expected from numerical integration and rounding.
Submission-to-acceptance interval: 29 days (received 14 January 2026; accepted 12 February 2026).

Notes

Limitations: analysis was conducted on text-extracted content only; raw figure files were unavailable, so image duplication, splicing, or PS detection could not be performed. Verifying the review-process anomaly would require access to the publisher's editorial system and reviewer logs, which are not publicly available.
The mathematical self-consistency check supports the technical credibility of the reported simulations but does not validate all figures, intermediate numerical tables, or theoretical lemmas.
DOI: 10.1109/JIOT.2026.3665672 — preserved verbatim from the source.
This report does not constitute a formal finding of misconduct; final adjudication remains with the relevant institutional and editorial bodies.