{"id":"b4314ec9-a649-4cf3-b175-ee6337575a6b","arxiv_id":"2608.00524","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A stability-gated QUBO formulation routes CAVs into urban platoons, giving an 18.5% simulated fleet-energy saving and 14.2% hardware sampling of the exact optimum via shallow-chain QAOA.","lead":"Roadside cameras judge which city road segments have stable traffic, and a quantum-assisted dispatcher then groups self-driving cars into fuel-saving platoons only on those segments. A 24-hour simulation of Troy, NY reports 18.5% lower fleet energy demand, and a 25-qubit IBM processor sampled the best route with 14% probability.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"18.5% fleet energy saving lacks a physical mechanism: at the reported 29.5 km/h fleet average, even an upper-bound 20% drag cut yields ~2.3 kWh, not 66.3 kWh; an energy decomposition is missing.","rationale":"The reader's weakest_assumption identifies η_drag and exogenous traffic. I agree with the first concern only partially: the drag coefficient is uncertain, but the more serious issue is internal arithmetic. Even with generously large drag reductions, the stated physics in Eq. (1) cannot produce a 66.3 kWh saving at the reported urban average speed. This is not a disagreement with external consensus; it is a simple consistency check on the paper's own reported aggregate statistics. The QPU results, especially the 66.7% CNOT-depth reduction and the nonzero P_feas/P_opt on ibm_boston, are more credible and are supported by the hardware table, though shot counts and run-to-run variance are absent. The exogenous-traffic limitation is real and explicitly acknowledged in Section VII.A and the Conclusion; it is secondary to the unexplained magnitude of the energy saving. I would keep the reader's CONDITIONAL verdict, but the acceptance conditions should include an energy-budget decomposition that attributes the 66.3 kWh to drag reduction, route changes, and trajectory/CACC smoothing separately. Without that, the headline 18.5% result remains unverified even as a simulation claim.","tokens_in":11303,"tokens_out":13813,"duration_ms":167440,"concrete_test":"Run an energy-budget decomposition on the SUMO outputs: for each policy, split E_physical into rolling, inertial, and aerodynamic terms from recorded v(t), a(t) traces, and recompute the cooperative scenario with η_drag=0 to isolate the drag contribution. Independently compute the maximum possible drag saving as Σ over platooned vehicle pairs of 0.5ρ Cd Δη A v^3 Δt (Δη=0.20/0.05 as applicable) from the simulation traces, and compare it with the 66.3 kWh claimed. If this bound is below, say, 10 kWh, the 18.5% headline must be re-attributed to route/trajectory effects or the baseline must be re-specified.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Table III reports 358.4 kWh baseline vs 292.1 kWh cooperative (saving 66.3 kWh) with total fleet travel time 48.2 h and distance 1420.5 km, i.e. mean speed 8.19 m/s. With Eq. (1) parameters, aero power per vehicle is 0.5*1.225*0.30*2.4*v^3 ≈ 242 W at that speed; over 48.2 vehicle-hours the entire baseline aero dissipation is roughly 11.7 kWh. Even granting η_drag=0.20 to every platooned vehicle (the paper's actual follower/leader split averages 12.5%), the maximum drag-mediated saving is about 2.3 kWh, <0.7% of baseline. Producing 66.3 kWh of saving through drag alone would require the aero term to be ≥331 kWh, i.e. 92% of baseline tractive energy, which is inconsistent with urban speeds and with the same model's rolling/inertia terms. Thus the headline saving cannot be a consequence of the only platooning mechanism stated in Section III.B. It must be coming either from uncontrolled route/trajectory differences (cooperative distance and travel time are lower by 35 km and 6.1 h), from CACC smoothing not represented in Eq. (1), or from a baseline that is not the H_cost-minimizing policy claimed in Section VII.A. The paper reports no decomposition or sensitivity for these components, so the central energy result is not yet explained.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a cyber-physical pipeline for platooning-aware vehicle routing: roadside-unit video perception extracts kinematic features, a stability gate selectively enables platooning rewards, the multi-vehicle VRP with cooperative platooning is written as a QUBO with native quadratic pair terms, and the QUBO is solved by LC-QAOA on IBM quantum hardware. The authors claim exact ground-state equivalence with MILP on small instances, a 66.7% CNOT-depth reduction on a 25-qubit ibm_boston benchmark with P_feas=38.6% and P_opt=14.2% at p=2, and an 18.5% (66.3 kWh) fleet tractive-energy reduction in a 24-hour SUMO simulation of Troy, NY.","tokens_in":11546,"tokens_out":10864,"duration_ms":132655,"significance":"The paper contains several genuine strengths: an exhaustive 2^25-bitstring verification against CPLEX, transparent decoupling of the optimization objective H_total from the physical energy metric E_physical, a real QPU execution with raw sampling statistics, and an explicit stability-gating mechanism that connects edge perception to the QUBO rewards. If the energy result were mechanically explained, the contribution would be a useful demonstration of an edge-assisted hybrid quantum dispatch loop. However, the headline 18.5% fleet energy saving is not supported by the stated physical mechanism, and the baseline ablation is internally suspicious. The QUBO/hardware portion is more defensible but is not the paper's headline claim.","major_comments":[{"comment":"The headline energy saving is not explained by the platooning drag mechanism. From Table III, mean fleet speed is 1420.5 km / 48.2 h = 8.19 m/s. Using Eq. (1), baseline aero power at this speed is about 0.5*1.225*0.30*2.4*8.19^3 ≈ 242 W per vehicle, giving about 11.7 kWh of aero dissipation over 48.2 vehicle-hours (≈13.3 kWh after the η_elec=0.88 in Eq. (2)). Even applying a 20% drag reduction to every vehicle yields at most ≈2.7 kWh, and the paper's actual follower/leader split averages 12.5%, giving ≈1.7 kWh. The reported saving is 66.3 kWh, an order of magnitude larger. Table III also shows 35.3 km less distance and 6.1 h less travel time in the cooperative case, so the saving must come from route/trajectory differences or an unreported mechanism. Please provide a decomposition into aero, rolling, and inertia contributions, and a sensitivity sweep over η_drag; without this the abstrac","section":"§VII.A, Table III; §III.B, Eq. (1)"},{"comment":"The baseline ablation appears internally inconsistent. Both baseline and cooperative policies are stated to use the same VRP model with the same H_cost (Eq. (6)), where c_ij is defined in §IV.C.a as per-vehicle energy. The platooning reward s_e,k,l in Eq. (7)/§IV.C.b is the aerodynamic drag saving, which at urban speeds is a few Wh per km per pair, far smaller than the per-vehicle energy cost (≈0.25 kWh/km implied by Table III). A perturbation of this size cannot change the optimal route by 35 km and 6.1 h; if anything, adding negative rewards on shared segments should make longer/shared routes more attractive, not shorter routes. Either the baseline is not the H_cost-minimizing solution of the same model, or the implemented reward weights are not the physical values in §IV.C.b. Please report the optimized route sets, the reward magnitudes, and a direct comparison of baseline and coopera","section":"§IV.B, §IV.C, §VII.A"},{"comment":"The claim that γ=50.0 'mathematically guarantees exact ground-state equivalence with classical MILP solvers' is stronger than the evidence presented. §V.A reports exhaustive search only on 4- and 5-node instances; a finite set of instances does not establish a general penalty bound. Please either provide a theorem showing the required γ in terms of N, K, c_ij, s_e, and the binary-encoding ranges, or revise the wording to 'verified on the reported instances.' This matters because exact ground-state equivalence is stated as a main contribution.","section":"§IV.C.c, §V.A"},{"comment":"The exogenous-traffic assumption is acknowledged in §VII.A and the Conclusion, but its load-bearing role for the energy result is underemphasized. Because candidate routes do not mutate SUMO background traffic, travel times and platoon feasibility windows are fixed, and the model cannot capture congestion shifts caused by dispatch decisions. The 18.5% figure should be presented as an estimate under fixed background traffic, with a discussion of the likely direction and magnitude of bias; otherwise readers may treat it as a deployment prediction rather than an upper-bound-style policy ablation.","section":"§VII.A, §VIII"}],"minor_comments":[{"comment":"The definition of c_ij contains a dimensional inconsistency: the term 'ma' appears as a force multiplied by distance, but in Eq. (1) ma is instantaneous power. If a is a representative acceleration, clarify how it is obtained from the SUMO trajectory and why it is not integrated over the segment.","section":"§IV.C.a"},{"comment":"The column 'QUBO Energy Cost (kWh)' is confusing: CPLEX has QUBO value -2.333 and cost 16.8, while QPU rows with P_feas=0 report repaired candidates. Define the relation between the QUBO value, the decoded route cost, and the kWh units.","section":"Table IV"},{"comment":"The stability thresholds (CV_v ≤ 0.12, σ²_a ≤ 0.5 m/s²) are presented as fixed constants with no calibration or sensitivity analysis. A brief sensitivity table or a reference for these values would strengthen the claim that the gating mechanism is robust.","section":"§VI.A.3, Table II"},{"comment":"The curve 'Hourly Platooned Distance (%)' is not defined in the caption or text: is it the fraction of fleet vehicle-km in platoon, or the fraction of candidate platoonable distance occupied? Please define and state the scale.","section":"Fig. 3"}],"recommendation":"major_revision","confidential_remarks":"The paper has two separable contributions: the QUBO/LC-QAOA hardware demonstration, which is supportable and worth publishing if properly scoped, and the 18.5% fleet-energy claim, which is currently not mechanistically explained. I would ask the authors for an energy decomposition and a re-examination of the baseline ablation before further consideration. The route differences in Table III are the main red flag; if the baseline is not actually the H_cost-minimizing policy, the ablation is invalid and the headline number may be an artifact of route optimization rather than platooning."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague, here's my read on arXiv:2608.00524.\n\nThe paper does two things well. First, the exhaustive 2^25 bitstring verification is real, carefully described, and matches CPLEX. Second, the physical QPU runs on ibm_boston with LC-QAOA are genuine: 66.7% depth reduction and 14.2% exact-ground-state sampling at p=2 are concrete numbers. The stability-gating idea—using RSU-derived traffic metrics to zero out platooning rewards on turbulent segments—is a sensible way to make the QUBO reflect physical conditions.\n\nThat said, the paper's central claim doesn't hold up. The 18.5% fleet energy saving (66.3 kWh) is attributed to aerodynamic drag reduction from platooning. But the paper's own Eq. (1) and Table III contradict that. At the fleet average speed of 8.19 m/s, total baseline aero dissipation is about 11.7 kWh. Even a 20% drag cut for every vehicle—a generous upper bound—yields at most 2.3 kWh saved. The remaining ~64 kWh has to come from somewhere else, and the paper doesn't decompose it. The cooperative routes are 35 km shorter and take 6.1 fewer vehicle-hours; those differences alone would produce large energy changes through rolling resistance, acceleration, and travel-time reduction. So the headline saving is conflated with route-length and schedule improvements, not platooning physics.\n\nOther soft spots: the vision pipeline is never actually run; perception robustness is tested by injecting Gaussian noise into feature vectors. The CPLEX objective appears as 16.80 in one section and -2.333 in Table IV without reconciliation. The 'mathematical guarantee' of ground-state equivalence is only demonstrated for two tiny instances, not a general proof.\n\nIs it worth refereeing? Yes, but for the right reasons. The hardware and formulation work deserve expert scrutiny. The authors need to add an energy breakdown by mechanism, sensitivity on eta_drag and the stability thresholds, and either correct or remove the unwarranted energy claim. A serious referee should ask for those before publication.\n\nNet: this is a solid experimental paper about QUBO/QAOA for platooning problems, currently overselling its energy results. I'd cite it for the QPU data, not for the energy saving.","headline":"Real QPU evidence and a clean small-instance verification, but the headline 18.5% energy saving is not physically supported: the drag mechanism in Eq. (1) can account for only a few kWh of the claimed 66.3 kWh.","tokens_in":12202,"tokens_out":2937,"would_cite":false,"duration_ms":34298,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A stability-gated QUBO formulation of cooperative platoon routing cuts simulated urban fleet energy demand by 18.5% and can be sampled on shallow quantum circuits.","keywords":["connected autonomous vehicles","platooning","vehicle routing problem","QUBO","QAOA","edge computing","traffic stability gating","quantum hardware benchmark"],"falsifier":"Re-run the 24-hour microscopic simulation with drag-reduction coefficients varied over a physically plausible range (follower 10-30%, leader 0-10%) and with background traffic that reacts to the dispatched routes; if the fleet energy saving drops below a meaningful threshold under any realistic setting, the 18.5% claim fails.","tokens_in":11015,"feed_emoji":"🚗","tokens_out":8341,"duration_ms":87291,"temperature":0.7,"pith_summary":"The paper sets out to show that platooning-aware vehicle routing can be made both tractable and physically safe by writing it directly as a QUBO model, where pairwise platooning savings are native quadratic terms, and by letting roadside perception gate those rewards to road segments with stable traffic flow. In a 24-hour microscopic simulation of an urban network, the resulting cooperative routes reduce fleet tractive-energy demand by 18.5% relative to a non-cooperative baseline. On a real 25-qubit quantum processor, Linear-Chain QAOA cuts two-qubit CNOT depth by 66.7% versus dense QAOA and samples the exact classical optimum with a feasible-sampling probability of 38.6%. If correct, the work gives a concrete closed-loop architecture in which roadside sensing, stability gating, and shallow quantum optimization cooperate for CAV dispatch.","feed_headline":"Platoon routing cuts simulated fleet energy 18.5 percent","feed_subtitle":"Stability-gated rewards and a 25-qubit quantum circuit beat non-cooperative dispatch in a 24-hour urban test.","key_machinery":"The central object is the QUBO Hamiltonian H_total = H_cost - H_platoon + gamma H_constraints. Platooning savings are carried by quadratic terms s_{e,k,l} x_{e,k} x_{e,l}, which map natively onto Ising ZZ couplings; the coefficient s is nonzero only when a roadside unit reports stable flow (gate y_e=1) and the two vehicles' predicted segment-entry times differ by at most 3 seconds. Constraints are quadratic penalties, with Miller-Tucker-Zemlin order variables binary encoded, and gamma=50 is set to guarantee exact ground-state equivalence to the classical MILP solution.","core_discovery":"The central claim is that multi-vehicle routing with cooperative platooning can be formulated directly as a QUBO Hamiltonian: every pairwise platooning interaction becomes an off-diagonal Ising term, eliminating the auxiliary binary variables that classical MILP linearizations require. Route feasibility is enforced through quadratic penalties, including binary-encoded Miller-Tucker-Zemlin subtour constraints, and the penalty scale is chosen so the QUBO ground state exactly matches a classical exact solver. The paper validates the equivalence by exhaustive enumeration of all 25-bit states on small instances, then executes the model on a physical 25-qubit processor. There, Linear-Chain QAOA at","pith_inferences":["Editorial inference: the 18.5% saving is likely driven mostly by the stability-gated reward structure rather than by the quantum solver; the marginal value of quantum execution would be isolated by solving the same gated QUBO with a classical optimizer.","Editorial inference: because the simulation treats background traffic as exogenous, the result is a one-shot policy response; a reactive-traffic model could erode platoon windows and shrink the saving.","Editorial inference: the drag-reduction coefficients come from highway heavy-truck platooning, while the simulated fleet consists of passenger EVs at 1-2 m urban gaps; field measurements of urban EV platoon aerodynamics would be the natural test of the energy model.","Editorial inference: the same stability-gated QUBO pattern could transfer to other spatiotemporal coordination problems, such as signal-priority transit or depot-based truck platooning, wherever a perception layer can gate the reward."],"forward_implications":["If the ground-state equivalence holds beyond the small validated instances, platoon routing can be solved without the O(|E| K^2) auxiliary variables of MILP linearization, shrinking the constraint matrix and enabling faster dispatch.","Stability gating means platooning rewards are only active on laminar road segments, so turbulent traffic automatically suppresses close-gap coordination and reduces unsafe disengagements.","Depth-compressed Linear-Chain QAOA provides a hardware-viable path: on 25 qubits it samples the exact optimum at p=2, while dense QAOA fails under gate noise.","The 18.5% fleet-energy reduction, if reproduced in physical operation, implies that dispatch software watching traffic stability and coordinating routes can deliver substantial operational savings without changing the vehicles themselves.","The architecture supports a closed loop in which roadside observations update stability flags, rewards, and re-optimization as traffic evolves, with optimized routes fed back into the traffic simulation."],"supporting_citations":[{"why":"supplies the 20% follower drag-reduction coefficient used to compute platooning rewards.","marker":"[1]"},{"why":"supplies the 5% leader drag-reduction coefficient used in the platoon energy model.","marker":"[2]"},{"why":"supplies the Miller-Tucker-Zemlin subtour constraints that the QUBO encodes in binary.","marker":"[3]"},{"why":"defines the Quantum Approximate Optimization Algorithm that the paper runs on hardware.","marker":"[10]"},{"why":"provides the Linear-Chain QAOA depth-compression technique behind the hardware results.","marker":"[18]"},{"why":"gives the prior QUBO highway-platooning formulations this paper extends to direct quadratic VRP-CP modeling.","marker":"[20], [21]"},{"why":"supplies the object detector used by roadside units to extract kinematic state.","marker":"[27]"},{"why":"provides the microscopic traffic simulation environment for the 24-hour fleet evaluation.","marker":"[33]"}],"fun_headline_variants":["Quantum platoon routing cuts fleet energy 18.5%","Edge-assisted quantum dispatch saves 18.5% energy","Hybrid quantum platooning yields 18.5% energy drop","QUBO platoon routing: 18.5% energy savings","CAV platooning with quantum optimization: 18.5% savings"],"cache_read_input_tokens":2688,"weakest_assumption_plain":"The headline saving assumes highway-truck platoon drag-reduction coefficients (20% follower, 5% leader) transfer to passenger electric vehicles at 1-2 m urban gaps, and that background traffic does not react to the dispatcher's routes.","fun_headline_variants_meta":{"raw":{"variants":["Quantum platoon routing cuts fleet energy 18.5%","Edge-assisted quantum dispatch saves 18.5% energy","Hybrid quantum platooning yields 18.5% energy drop","QUBO platoon routing: 18.5% energy savings","CAV platooning with quantum optimization: 18.5% savings"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000739,"raw_usage":{"total_tokens":3173,"prompt_tokens":813,"completion_tokens":2360,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":557,"completion_tokens_details":{"reasoning_tokens":2268}},"tokens_in":557,"tokens_out":2360,"duration_ms":17387,"temperature":1.0,"reasoning_tokens":2268,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T00:45:59.794075+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Re-run the 24-hour microscopic simulation with drag-reduction coefficients varied over a physically plausible range (follower 10-30%, leader 0-10%) and with background traffic that reacts to the dispatched routes; if the fleet energy saving drops below a meaningful threshold under any realistic setting, the 18.5% claim fails.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"supplies the 20% follower drag-reduction coefficient used to compute platooning rewards."},{"cited_title":"An experimental study on the fuel reduction potential of heavy duty ve- hicle platooning","cited_arxiv_id":null,"evidence_quote":"supplies the 5% leader drag-reduction coefficient used in the platoon energy model."},{"cited_title":"Integer programming formulation of traveling salesman problems.Journal of the ACM (JACM), 7(4):326–329, 1960","cited_arxiv_id":null,"evidence_quote":"supplies the Miller-Tucker-Zemlin subtour constraints that the QUBO encodes in binary."},{"cited_title":"Shallow and robust qaoa: Improving fea- sibility and hardware performance via linear-chain and ramp schedules,","cited_arxiv_id":null,"evidence_quote":"provides the Linear-Chain QAOA depth-compression technique behind the hardware results."},{"cited_title":"Ultralytics YOLO11: State-of-the-art object detection and tracking","cited_arxiv_id":null,"evidence_quote":"supplies the object detector used by roadside units to extract kinematic state."},{"cited_title":"Recent development and applications of sumo-simulation of urban mobility","cited_arxiv_id":null,"evidence_quote":"provides the microscopic traffic simulation environment for the 24-hour fleet evaluation."}],"review_version":1}