{"id":"e95f2220-029a-4bd8-a4ab-d9c625224eb4","arxiv_id":"2502.06580","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"A dissipativity-based co-design of distributed inventory consensus controllers and communication topology keeps parallel supply chains synchronized while attenuating disturbances.","lead":"This paper designs communication topologies and consensus controllers together for supply chain networks, so multiple parallel supply chains keep inventories synchronized under demand and waste uncertainty. It uses dissipativity theory to turn the design into convex optimization problems and demonstrates improved consensus in simulations.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The L2-gain bound (33) is imposed on consensus error, which is blind to common-mode disturbances; it cannot by itself bound bullwhip or ripple, so the headline practical claim rests on an unproven structural proxy.","rationale":"The paper has real contributions: a convex LMI formulation for co-design, compositional use of dissipativity, and a publicly available simulator; the consensus simulation results are plausible. My concern is specifically with the strongest advertised claim. The robustness constraint (33) is a legitimate H-infinity-type performance objective for consensus, but it is not a bullwhip or ripple objective. The output z uses only between-chain differences (30); bullwhip is a within-chain, echelon-by-echelon amplification of demand variance, and ripple is the propagation of disruptions through the chain or network. A disturbance that hits all chains equally leaves z at zero while allowing arbitrary demand amplification, so even a perfect solution of (36) could have unbounded bullwhip. The reader flagged this as an asserted, unverified equivalence; I agree and sharpen it to a structural blind spot. The conditional verdict is appropriate: either add direct bullwhip and ripple metrics to the existing Monte Carlo framework, or soften the claims to robust consensus rather than bullwhip and ripple mitigation. Proof omission of Theorem 3 and Monte Carlo uncertainty quantification are secondary, because Theorem 3's first LMI already enforces the needed local dissipativity property and the consensus comparison is adequate for the narrower claim. No other issue is more load-bearing than this mismatch between the guaranteed quantity and the claimed practical phenomenon.","tokens_in":24151,"tokens_out":9307,"duration_ms":88119,"concrete_test":"In the existing Monte Carlo simulator, augment Section V with direct bullwhip and ripple metrics: for each method, compute the bullwhip ratio per echelon BW_{i,k} = Var_t(u_{i,k}(t))/Var_t(d_{i,n}(t)) (or the standard order-to-demand variance ratio), and for ripple compute the post-disruption inventory and order deviation energy, e.g., sum over t=240..720 of ||x_i(t)-x_i^*||^2, comparing DCC-C and DCC-U against LSSC, LSFC, and GCC. If DCC methods do not consistently reduce these metrics (not just PMAE), the claim that (33) mitigates bullwhip and ripple is unsupported. As a minimal analytical check, simulate the common-mode case: identical chains with identical demand and waste; if the consensus bound holds while order variance increases, the structural gap is confirmed.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The advertised benefit—mitigating bullwhip and ripple effects—is not entailed by the main result. Theorem 2 proves the closed-loop SCN (29)-(31) satisfies the L2-gain bound (33) from the aggregate disturbance r(t) to the consensus-error output z(t), where z is defined in (30) as a vector of pairwise differences of inventory errors. A disturbance component common to all supply chains cancels in z: if N identical chains receive the same demand and waste realizations, then y_i(t)=y_j(t) for all i,j, so z(t)=0 identically and the bound holds trivially even if orders and absolute inventories amplify demand arbitrarily. Bullwhip is precisely the within-chain amplification of demand variance into order variance, and ripple is the propagation of disruptions through the network; neither is a function of between-chain consensus error. Therefore minimizing gamma in (36) need not reduce either effect. Section III.E asserts that r-to-z attenuation will minimize both the bullwhip effect and the ripple effect, but no theorem or experiment measures bullwhip or ripple; Section V reports only consensus PMAE/APMAE/CAPMAE. This is a structural mismatch, not just missing evidence.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a dissipativity-based co-design of distributed inventory consensus controllers and communication topologies for a network of parallel supply chains. Each chain is modeled as a series of transportation/inventory links with delays, waste, and demand disturbances; local state-feedback controllers are designed to enforce equilibrium-independent dissipativity, and a global LMI problem then synthesizes the consensus coupling gains together with a sparse communication topology. The main theoretical claim is Theorem 2: the closed-loop SCN satisfies an L2-gain bound from the aggregate disturbance to a consensus-error output when the global controller is synthesized by the LMI co-design problem. The paper also claims that this robustness property mitigates bullwhip and ripple effects, and presents simulations of a three-chain network comparing the proposed DCC-C and DCC-U strategies with local steady-state control, local feedback control, and global consensus control.","tokens_in":24378,"tokens_out":9003,"duration_ms":168190,"significance":"If the main theorem and its application to bullwhip/ripple mitigation were fully established, the paper would offer a computationally attractive, scalable way to co-design control and topology using only per-chain dissipativity data, with a public simulator and an unusually complete set of LMI derivations. The paper also gives credit to the prior framework in [46],[47] and includes the proof of Proposition 2 in the appendix. However, the claimed practical benefit is not demonstrated: the L2-gain notion is applied to consensus error, not to bullwhip or ripple, and no direct bullwhip or ripple metric is simulated. There is also an input-output dimension mismatch between the local dissipativity certificates used in the simulations and the interconnection structure required by the global co-design theorem. These are load-bearing gaps that prevent the paper, in its present form, from supporting its headline claims.","major_comments":[{"comment":"The claim that attenuating the L2-gain from r(t) to z(t) mitigates the bullwhip and ripple effects is not established. The output z(t) in Eq. (30) is a vector of pairwise differences of inventory errors, so any disturbance component common to all chains cancels in z: if N identical chains receive the same demand and waste realizations, z(t) is identically zero even if within-chain orders amplify demand variance arbitrarily. Bullwhip is precisely within-chain demand-to-order amplification, and ripple is disruption propagation through the network; neither is a function of between-chain consensus error. Section V reports only the consensus metrics PMAE, APMAE, and CAPMAE and never measures order variance, bullwhip, or ripple. The paper needs either a theorem that connects the r-to-z gain to a formal bullwhip/ripple metric or new simulations that measure those effects directly.","section":"Section III.E, Eq. (33)"},{"comment":"There is a structural mismatch between the local dissipativity certificates used in the numerical DCC methods and the input-output pair required by the global co-design theorem. The networked system in Eq. (31) uses the virtual output y_i(t)=C_i x_i(t) of dimension n, with M_{ηy}=BK and K_{ij}∈R^{n×n}; Theorem 2 therefore requires X_i-EID from η_i(t) to C_i x_i(t). Remark 7 and Theorem 3, however, replace C_i by I and certify X_i-EID from η_i(t) to the full state x_i(t) of dimension n_i. The paper asserts in Remark 7 that this extension \"does not impact the networked system view\" or the global design, but no output-reduction lemma is provided, and because chains have different n_i when delays differ, full-state outputs cannot be fed into the consensus controller (28) and the interconnection matrices in Eq. (31). Since the DCC simulations use Theorem 3 for local control design, the simulated system does not instantiate the hypotheses of Theorem 2 as proved.","section":"Remark 7, Theorem 3 vs. Section III.E, Theorem 2"},{"comment":"The proof of Theorem 3 is explicitly omitted, with the text stating \"The proof is omitted here.\" This theorem supplies the local controllers used in all proposed DCC simulations, so the omission is not merely cosmetic. A journal version should either provide the complete proof (including the derivation of the necessary condition (44) from Remark 9) or state precisely which parts are assumed and which are routine algebra.","section":"Section IV.C, Theorem 3"},{"comment":"The proof of Theorem 2 is a direct specialization of Proposition 2 to the SCN interconnection (31), and Proposition 2 is itself the prior result from [46],[47]. Although the appendix proves Proposition 2, the novelty of Theorem 2 over [46],[47] should be delineated more carefully; as written, the global co-design result reads as an application of the existing framework rather than a new synthesis result. The authors should state explicitly which ingredients are new: the supply-chain model, the steady-state control law, the local tuning conditions, or the topology objective formulation.","section":"Section IV.B and Theorem 2"}],"minor_comments":[{"comment":"The Monte Carlo comparison reports point values of final CAPMAE without confidence intervals, standard errors, or any statistical significance test; with 1,000 realizations, such information would substantially strengthen the claim that DCC-U outperforms the baselines.","section":"Section V.C"},{"comment":"The parameters used in the co-design objective—the cost matrix C, the scaling c0, and the prespecified scalars p_i in Theorem 3—are not reported in the simulation setup, so the topologies in Figs. 9 and 10 cannot be reproduced. The paper should list these values and ideally provide a sensitivity study for c0 and p_i.","section":"Section V.B and Theorem 3"},{"comment":"The claim that DCC-U provides the \"fastest convergence to a consensus\" is qualitative; no convergence-time or settling-time metric is defined or measured. A quantitative criterion would make the comparison more precise.","section":"Section V.D"},{"comment":"The symbol ρ_i is used both for the inventory decay rate in Section III.A and for the passivity index in Theorem 3; this notational conflict should be resolved (for example, by using different letters for the local passivity indices).","section":"Throughout"},{"comment":"There are minor typos: \"discrte-time\" in Corollary 1, \"are are\" in the proof of Proposition 3, and the caption of Fig. 13 says \"Average PCMAE\" while the text defines CAPMAE. These should be corrected.","section":"Corollary 1 and Proposition 3 proof"}],"recommendation":"major_revision","confidential_remarks":"The paper leans heavily on the authors' own prior framework in [46] and [47], one of which is listed as \"Submitted\" and the other as an arXiv preprint; the editors may wish to check whether the overlap with these unpublished works is adequately disclosed and whether the new supply-chain-specific contribution is substantial enough for a journal paper. The central technical gap I see is the dimension mismatch between the local dissipativity certificates (full-state output) and the global interconnection output (inventory error), which the authors should address with an output transformation argument or by revising the local design. The bullwhip/ripple issue is also a validation gap that is fixable with direct measurements in the existing simulator."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear Colleague,\n\nQuick take: the paper is a respectable application of the authors' own dissipativity-based co-design machinery to inventory consensus in parallel supply chains. The SCN modeling is new—delay states, inventory error dynamics, consensus output—and they ship a simulator. The LMI chain from Prop. 1 through Thm. 2 is internally consistent and the simulations show the DCC methods beat the baselines on consensus PMAE. That part holds up.\n\nThe soft spot is the bullwhip and ripple claim. Section III.E asserts that attenuating disturbances r(t) to the consensus error z(t) \"will minimize both the bullwhip effect and the ripple effect.\" But z(t) is a vector of pairwise inventory-error differences. A disturbance common to all chains cancels in z, so the L2-gain bound (33) can hold trivially even if orders or absolute inventories amplify demand. Bullwhip is within-chain order variance; ripple is disruption propagation through the network. Neither is a function of between-chain consensus error. No theorem or experiment in Section V measures bullwhip or ripple—only consensus PMAE/APMAE/CAPMAE. So the headline practical benefit is not entailed by the main result. That is a real gap, not a nitpick. The authors could fix it by either reframing the contribution as consensus co-design with disturbance attenuation on consensus error, or by adding direct bullwhip/ripple metrics to the experiments.\n\nOther issues are minor. Theorem 3's proof is omitted, though it follows from the preceding remarks; the Monte Carlo comparison lacks error bars despite 1000 realizations; and the core synthesis (Prop. 2) is carried over from the authors' platoon papers [46],[47]. That is legitimate extension, but it means the novelty is in the application and modeling, not the underlying LMI machinery.\n\nIn sum: the consensus co-design story is plausible and worth refereeing, but the bullwhip/ripple claim needs to be either demonstrated or withdrawn. I'd send it to review with a request for revision.","headline":"A solid consensus co-design paper whose advertised bullwhip/ripple benefit rests on an asserted proxy rather than a proved link.","tokens_in":24914,"tokens_out":2338,"would_cite":false,"duration_ms":21050,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["93A14","93D25","93C55","90B05"],"pacs":[],"model":"deepseek-v4-flash","headline":"Distributed supply chains can be synchronized by one convex co-design that picks both controller gains and communication links from per-chain dissipativity data alone.","keywords":["inventory consensus","supply chain networks","dissipativity-based control","equilibrium-independent dissipativity","topology co-design","linear matrix inequalities","bullwhip effect","ripple effect"],"falsifier":"In the paper's simulator, apply the DCC-U controller and compute, at each echelon, the bullwhip ratio (variance of orders placed divided by variance of customer demand) and a ripple metric such as the time to restore target inventory after the transportation and inventory failures injected at $t=240$ and $t=480$; if these metrics do not improve relative to the LSSC, LSFC, and GCC baselines while consensus PMAE does improve, the asserted link between consensus-error attenuation and bullwhip/ripple mitigation would be refuted.","tokens_in":23932,"feed_emoji":"📦","tokens_out":9809,"duration_ms":74128,"temperature":0.7,"pith_summary":"This paper seeks to establish that the inventory-consensus problem for parallel supply chains can be solved by co-designing the distributed controllers and the communication topology together, in a single convex optimization. The scheme needs only each supply chain's dissipativity data — a quantitative input–output 'energy' bookkeeping of the chain — rather than an exact global model, which makes the design scalable and compositional. The guarantee is that the closed-loop network attenuates demand and waste disturbances to a prespecified level, expressed as an $L_2$-gain bound from the disturbance signal $r(t)$ to the consensus error $z(t)$. If the claim holds, supply chain operators get a computationally efficient recipe that simultaneously synchronizes inventories, decides which communication links are worth keeping, and positions the network against the bullwhip and ripple effects.","feed_headline":"One convex program co-designs supply-chain control and topology","feed_subtitle":"Per-chain dissipativity data alone yield inventory consensus with fewer links than full connectivity.","key_machinery":"The central object is equilibrium-independent dissipativity (EID): a system property asserting that the change in a positive storage function is bounded by a quadratic supply rate evaluated on deviations from any equilibrium, not just the origin. Each supply chain's error dynamics (24) are made EID with an input-feedforward–output-feedback passivity (IF-OFP) supply rate by the local LMI of Theorem 1. The idea that carries the global step is the compositionality of dissipativity: the network storage function is a weighted sum of the chains' storage functions, so the network-level supply rate follows from the per-chain supply rates together with the interconnection matrix $M$ in (31). Theorem 2 exploits this to assemble the consensus gains $K_{ij}$ from per-chain data, using the change of variables $L_{\\eta y}^{ij} = X_i^{11}B_i\\bar{K}_{ij}$ to keep every block of the LMI (37) linear in the decision variables; the co-design objective stays convex, which is what makes the joint controller-and-topology optimization computationally tractable.","core_discovery":"The paper's central claim is Theorem 2: after each supply chain's error dynamics are made dissipative (equilibrium-independent, with an IF-OFP passivity supply rate) by the local state-feedback controllers of Theorem 1, the closed-loop supply chain network (29)–(31) satisfies the robustness constraint $\\|z\\|_2^2/\\|r\\|_2^2 \\le \\gamma^2$ with $\\gamma^2 \\le \\bar{\\gamma}$, where $r(t)$ collects the demand and waste disturbances and $z(t)$ is the consensus error across the chains. The controller $K$ that achieves this is obtained from the LMI problem (36)–(38), whose objective jointly minimizes the $L_2$-gain $\\gamma^2$ and the $\\ell^1$-norm of the coupling gains weighted by communication costs, thereby selecting the topology by driving unnecessary links to zero. The synthesis requires only the per-chain dissipativity indices, not the full network model, and the numerical study reports that the resulting distributed designs outperform steady-state, local-feedback, and fully-connected global consensus baselines in consensus error over 1000 Monte Carlo realizations that include injected transportation and inventory failures.","pith_inferences":["A direct testable extension is to measure the bullwhip ratio (variance of echelon orders over variance of end demand) and a ripple metric (inventory recovery time after the injected failures at $t=240$ and $t=480$) in the same simulator; the paper's link between consensus-error attenuation and these supply-chain phenomena is checkable without new theory.","The same dissipativity-based interconnection synthesis should transfer to other consensus problems with delay dynamics and per-agent local controllers — production–inventory networks with returns, fleets of storage devices, or water-distribution systems — wherever each agent can be made IF-OFP by a local LMI.","Reading the sparsity objective as feature selection, the co-design effectively chooses which inventory-state differences matter most for synchronization; the cost matrix $C$ lets a designer encode link prices directly, an interpretation the paper leaves implicit.","The method currently takes mean demand and mean waste levels as known inputs for the steady-state component (18); an adaptive loop that estimates and updates these means online, which the authors list as future work, would make the scheme self-tuning under shifting demand regimes."],"forward_implications":["Inventory consensus can be guaranteed to a prespecified disturbance-to-error gain $\\bar{\\gamma}$ while the optimization simultaneously drops communication links that do not earn their cost, so the network ends up sparser than the fully-connected baseline.","The design is compositional and local: each chain's local controller is designed from its own passivity indices, and the global problem is a single LMI whose dimension scales with the number of chains without needing the global plant model.","The optimal topologies found in the simulations prefer non-adjacent links over adjacent ones and place more information flow at corner inventories, giving concrete, non-obvious rules for where to put communication in a supply chain network.","Because the method enforces $L_2$-gain attenuation from the demand and waste disturbances to the consensus error, it is designed to damp order amplification upstream (bullwhip effect) and disruption propagation in all directions (ripple effect) in the control-theoretic sense of constraint (33).","The method also absorbs transportation and inventory failures better than steady-state, local-feedback, and fully-connected consensus strategies, as seen in the cumulative average consensus-error metric."],"supporting_citations":[{"why":"Supplies the equilibrium-independent dissipativity definition and the compositionality principle that allows network-level dissipativity to be assembled from per-chain dissipativity data.","marker":"[40]"},{"why":"Provides the LMI lemma (50) that converts the dissipativity inequality into an equivalent linear matrix inequality, used throughout the proofs of the paper.","marker":"[44]"},{"why":"Contributes the interconnection-matrix synthesis LMI (Proposition 2) for dissipativity-based co-design of distributed controllers and topologies, which the paper adapts from vehicular platoons to supply chain networks.","marker":"[46], [47]"},{"why":"Supplies the block-wise equivalence (BEW) lemma used in Remark 8 to derive the local necessary conditions that improve the local controller design.","marker":"[49]"},{"why":"Defines the H-infinity consensus problem for supply chain systems under switching topologies and uncertain demands that this paper extends by co-designing the communication topology.","marker":"[19]"},{"why":"Provides the classic consensus formulation and the fully-connected global consensus baseline (GCC) used as a comparison strategy in the simulations.","marker":"[50]"},{"why":"Establishes the relationships among passivity, sector-boundedness, and dissipativity that justify the IF-OFP supply-rate choice used in the local designs.","marker":"[39]"}],"fun_headline_variants":["Convex co-design syncs supply chains with fewer links","Dissipativity-based co-design trims supply-chain links","Per-chain data alone yield robust inventory consensus","One LMI chooses control and topology for supply chains","Co-design cuts links while keeping inventory consensus"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The practical payoff rests on equating the bullwhip and ripple effects with the $L_2$-gain bound from disturbances to consensus error stated in (33), an identification the simulations never measure directly, since they report only consensus error rather than any bullwhip or ripple metric.","fun_headline_variants_meta":{"raw":{"variants":["Convex co-design syncs supply chains with fewer links","Dissipativity-based co-design trims supply-chain links","Per-chain data alone yield robust inventory consensus","One LMI chooses control and topology for supply chains","Co-design cuts links while keeping inventory consensus"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00026,"raw_usage":{"total_tokens":1612,"prompt_tokens":993,"completion_tokens":619,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":609,"completion_tokens_details":{"reasoning_tokens":545}},"tokens_in":609,"tokens_out":619,"duration_ms":6042,"temperature":1.0,"reasoning_tokens":545,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-08T14:58:12.250085+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"In the paper's simulator, apply the DCC-U controller and compute, at each echelon, the bullwhip ratio (variance of orders placed divided by variance of customer demand) and a ripple metric such as the time to restore target inventory after the transportation and inventory failures injected at $t=240$ and $t=480$; if these metrics do not improve relative to the LSSC, LSFC, and GCC baselines while consensus PMAE does improve, the asserted link between consensus-error attenuation and bullwhip/ripple mitigation would be refuted.","supporting_citations":[{"cited_title":"Measuring and Eliminating the Bullwhip in Closed Loop Supply Chains Using Control Theory and Internet of Things,","cited_arxiv_id":null,"evidence_quote":"Supplies the equilibrium-independent dissipativity definition and the compositionality principle that allows network-level dissipativity to be assembled from per-chain dissipativity data."},{"cited_title":"Dynamical Investigation and Distributed Consensus Tracking Control of a Variable-Order Fractional Supply Chain Network Using a Multi-Agent Neural Network-Based Control Method,","cited_arxiv_id":null,"evidence_quote":"Provides the LMI lemma (50) that converts the dissipativity inequality into an equivalent linear matrix inequality, used throughout the proofs of the paper."},{"cited_title":"Compositional Design and Verification of Large-Scale Systems Using Dissipativity Theory,","cited_arxiv_id":null,"evidence_quote":"Supplies the block-wise equivalence (BEW) lemma used in Remark 8 to derive the local necessary conditions that improve the local controller design."},{"cited_title":"Supply-Chain Networks: A Complex Adaptive Systems Perspective,","cited_arxiv_id":null,"evidence_quote":"Defines the H-infinity consensus problem for supply chain systems under switching topologies and uncertain demands that this paper extends by co-designing the communication topology."},{"cited_title":"Y ALMIP : A Toolbox for Modeling and Optimization in MATLAB,","cited_arxiv_id":null,"evidence_quote":"Provides the classic consensus formulation and the fully-connected global consensus baseline (GCC) used as a comparison strategy in the simulations."},{"cited_title":"An H∞ Control Method of the Bullwhip Effect for a Class of Supply Chain System,","cited_arxiv_id":null,"evidence_quote":"Establishes the relationships among passivity, sector-boundedness, and dissipativity that justify the IF-OFP supply-rate choice used in the local designs."}],"review_version":1}