{"id":"302d67d3-55cd-46ba-ba76-e485c543bdfa","arxiv_id":"2507.04675","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"In full-wave simulation, a 1-bit hybrid reconfigurable intelligent surface with per-element slot coupling into a parallel-plate waveguide estimates angle of arrival to within 3 degrees while steering reflected beams without quantization lobes.","lead":"This paper designs, in simulation, a radio reflector that senses where a signal arrives from while also steering the reflected signal, using only two readout wires. It matters because such an integrated sensing-and-beamforming surface could lower the cost and complexity of future smart radio environments, but no hardware prototype is built yet.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The ±3° AoA accuracy is validated only against the same simplified dipole model used to build the sensing matrix; a full-wave end-to-end check is missing.","rationale":"The reader's weakest assumption identifies exactly the same load-bearing point: the sensing matrix and the test signals come from one simplified forward model, so the AoA accuracy is not yet shown to survive real electromagnetic behavior. My independent read of Section III and Section IV confirms that the element-level HFSS simulations are used only to extract reflection phases for a MATLAB dipole-array model, and that H is built from 'the difference between the voltages on the coaxial connectors'—with no full-wave array-level sensing simulation or measurement reported. The beamforming results do get full-wave HFSS validation at the array level, which supports the steering claim but not the sensing claim. The proposed check—full-wave array-level simulation of the coaxial-port voltages with the same masks and angles, then plugging those voltages into the existing H—directly breaks the circularity. Because this is a simulation paper and the authors have prior experimental work in this line, CONDITIONAL is the appropriate verdict: the design is plausible and internally coherent, but the headline AoA number needs this verification or a prototype before acceptance as a hardware-capable claim. I do not see a more fundamental flaw: the phase-randomization beamforming evidence is reasonable, and the sensing concept is not contradicted by any internal inconsistency.","tokens_in":7455,"tokens_out":1416,"duration_ms":18973,"concrete_test":"Build the full 19-element HFSS model of Fig. 3 including the PPWG, slots, and two coaxial ports. For a fixed slot-length distribution and a small set of on/off masks (e.g., the 16 masks used in Fig. 7), simulate plane-wave incidence from several angles (say 0°, 15°, 30°, -20°, 45°) at 5.6 GHz and record the actual coaxial-port voltage differences. Then, using the sensing matrix H constructed from the simplified dipole model as described in Section IV, estimate the AoA from these full-wave voltages with the same CGS/MLP pipeline. If the estimation error exceeds ±3° at equivalent SNR=20 dB, the claimed accuracy is an artifact of the circular model loop; if it stays within ±3°, the simplified forward model is adequate. This test isolates the load-bearing assumption without requiring hardware.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central quantitative claim, AoA estimation with ±3° accuracy at M=16 and SNR=20 dB (Section IV, Fig. 7), rests on a closed simulation loop. The sensing matrix H is 'constructed by collecting the difference between the voltages on the coaxial connectors' for each reference angle and mask, but the text and prior work indicate these voltages are computed from a simplified array model in which each element is a dipole with phases taken from isolated HFSS element simulations (Section III). The same model then generates the test signals g = H f + noise for the Monte Carlo trials. If the real array—with mutual coupling, finite-array edge effects, and PPWG propagation—produces coaxial-port voltages that differ from this dipole model, then both the sensing matrix and the test data shift together only in simulation; in hardware the mismatch would appear as an unmodeled distortion, and the reported ±3° accuracy would not transfer. This is a modeling-validity concern rather than an internal inconsistency: the paper's own Section III acknowledges using 'a simplified array-level simulation in MATLAB' after full-wave element simulation, and the likely effect of assembling 19 elements with slots and a shared PPWG is not quantified. The MLP results are subject to the same limitation because training and test data are generated from the same source model, despite the added synthetic noise. The beamforming full-wave results (Fig. 5) are independent evidence for steering, but they do not validate the sensing forward model.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proposes a 1-bit hybrid reconfigurable intelligent surface (HRIS) in which resonant patch elements with PIN diodes both steer reflected beams and sense the angle of arrival (AoA). A portion of the incident signal is coupled through slots in the ground plane into a parallel-plate waveguide (PPWG), and two coaxial connectors collect the combined signal. AoA estimation is performed with compressive sensing (CGS) and with an MLP classifier. Beamforming with quantization-lobe suppression is demonstrated through HFSS full-wave array simulations using a randomized slot-length distribution, and AoA accuracy of ±3° is claimed for M=16 masks at SNR=20 dB, with degradation at lower SNR and improved robustness claimed for the MLP.","tokens_in":7651,"tokens_out":4310,"duration_ms":49788,"significance":"If the AoA accuracy and beamforming results hold in hardware, the proposed HRIS would be a valuable simplification over previous hybrid RIS designs: it uses only two coaxial readouts, 1-bit tuning, and no feedback loop, and it combines sensing and reflection in the same aperture. The full-wave HFSS element design and beamforming simulations are substantial strengths, as is the explicit comparison between CGS and MLP sensing approaches. However, the central sensing claim is currently supported only by a closed simulation loop in which the sensing matrix, the MLP training data, and the test signals all come from the same simplified dipole forward model; no full-wave end-to-end sensing simulation or measurement is provided. The significance of the paper is therefore conditional on additional validation or on careful reframing of the claimed accuracy.","major_comments":[{"comment":"The ±3° AoA accuracy is asserted for a simulation loop in which the sensing matrix H and the test signals g are both generated from the same simplified dipole-array model introduced in Section III. Section III states that each element is modeled as a dipole with phases extracted from isolated HFSS unit-cell simulations, and Section IV builds H by collecting the difference between coaxial-port voltages without a full-wave simulation of the full 19-element array with slots and PPWG. Any systematic error in that forward model—including mutual coupling, finite-array edge effects, PPWG propagation, and slot-coupling details—cancels between H and the synthetic test data, so the reported accuracy is an in-model consistency check rather than evidence for hardware-level AoA estimation. The central sensing claim needs either a full-wave end-to-end simulation of the coaxial-port voltages or a hardware measurement before the accuracy statement can stand.","section":"Section IV, Eq. (1) and Fig. 7"},{"comment":"The MLP is trained on the same 25 simulated samples (repeated 300 times with synthetic Gaussian noise) that define the dictionary used by CGS, and it is evaluated on the test portion of the same dataset with additional Gaussian noise. Because no independent physical model, measured data, or full-wave simulation is introduced at any stage, the MLP accuracy curve in Fig. 7(d) does not demonstrate robustness to model mismatch. The claim of robustness should be limited to additive noise within the assumed forward model, or the authors should provide validation on a different forward model.","section":"Section IV, MLP paragraph"},{"comment":"The slot-length distribution is chosen after simulating several random distributions and selecting the one with the lowest sidelobe levels, as the text acknowledges. This post-hoc selection means the HFSS beam patterns in Fig. 5 demonstrate that one particular realization works, but they do not establish a design rule or reliability of phase randomization with two slot lengths for general arrays. The beamforming claim should be framed as an existence demonstration, not as a validated design methodology.","section":"Section III, phase randomization"}],"minor_comments":[{"comment":"The sentence 'The ith element of f is zero if the incident AoA coincides with the ith reference angle and 0 otherwise' contains a typo; the intended meaning is presumably 'one if' and 'zero otherwise'.","section":"Section IV, Eq. (1)"},{"comment":"The text describes the element spacing as 'large element spacing (λ/2)', but with p=24 mm at 5.6 GHz the spacing is approximately 0.45λ, which is not conventionally considered large; please rephrase or justify.","section":"Introduction"},{"comment":"The caption phrase 'for the case of 1 incident angle and 5 different random masks' is unclear; please specify what is plotted in each panel and the exact quantity shown.","section":"Fig. 6 caption"},{"comment":"The 'simplified array-level simulation in MATLAB' is important for reproducibility, but the equations for the array factor and element-pattern model are not given; please include the array factor expression, the element pattern used, and how the extracted HFSS phases are combined.","section":"Section III"}],"recommendation":"major_revision","confidential_remarks":"The referee report focuses on the closed-loop validation issue. The paper's stated contribution is the integrated sensing and beamforming concept, but the sensing performance is not validated against any independent forward model. A full-wave sensing simulation or a hardware prototype would be needed to accept the ±3° accuracy claim as stated."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a credible, incremental design study, and the beamforming half of it is in good shape; the sensing numbers need independent validation before you trust them. The new piece is the specific architecture—1-bit PIN-diode patches, per-element slots coupling into a parallel-plate waveguide, two coaxial readouts, randomized slot lengths for phase randomization, and CS/MLP AoA estimation from the two-port voltage differences. That combination does not appear in the prior SIW-based hybrid RIS work, and it is plausibly simpler to manufacture and needs fewer RF chains.\n\nCredit where due: the element-level HFSS design is careful, the array-level beamforming is full-wave and shows steering at several angles with suppressed quantization lobes, and the authors are transparent that they picked the slot layout post hoc from a small set of random distributions. The sensing geometry is explained clearly, and the use of masks to build a better-conditioned inverse problem is a reasonable direction.\n\nThe soft spot is the one the stress-test note flags, and it is real. The ±3° AoA accuracy and the MLP curves in Fig. 7 are generated by a closed loop: H, the MLP training samples, and the test signals all come from the same simplified dipole-array model whose element phases are taken from isolated HFSS simulations. That makes the accuracy numbers a self-consistency check, not a prediction. A full-wave end-to-end sensing simulation, or hardware, would close the gap; without one, the central quantitative claim is conditional. The MLP evaluation has the same limitation, and the robustness-to-SNR claim is weaker because training and test data are synthetic expansions of the same 25 angles. The post hoc slot-layout selection means the beam patterns are best-case among tried layouts; that is not dishonest, but it should be stated as a design heuristic. The \"simple, low-cost\" claim is also untested, though not unreasonable. The citation pattern is fine—the authors build on their own SIW prototype line, including an experimental demonstration, which is the right prior work to lean on.\n\nBottom line: this deserves a serious referee. I would send it out, with the request that the authors either add a full-wave sensing validation or clearly label the AoA results as model-based demonstrations and share the code. It is useful for the applied EM/RIS community as a design proposal; it is not yet evidence for the hardware claim.","headline":"A credible, incremental design study with solid full-wave beamforming evidence; the AoA accuracy claims are only validated against the same simplified model used to build the sensing matrix.","tokens_in":8249,"tokens_out":2663,"would_cite":true,"duration_ms":30372,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A 1-bit hybrid reconfigurable surface can both sense angle of arrival and steer reflected beams using two coaxial readout ports.","keywords":["reconfigurable intelligent surface","1-bit metasurface","angle-of-arrival estimation","compressive sensing","parallel-plate waveguide","PIN diode","phase randomization","hybrid RIS"],"falsifier":"Assemble or simulate the full 19-element array in a full-wave solver that includes mutual coupling, edge effects, and parallel-plate waveguide propagation, illuminate it from known angles, and compare the two-port voltage differences with the simplified dipole model's predictions; if the measured sensing matrix differs substantially, the reported ±3° accuracy will not transfer to hardware.","tokens_in":7192,"feed_emoji":"📡","tokens_out":4383,"duration_ms":50112,"temperature":0.7,"pith_summary":"The paper tries to establish that a single reconfigurable surface can do two jobs at once: detect the direction an incoming wireless signal arrives from, and reflect that signal toward a chosen direction, using only simple 1-bit (on/off) tuning elements and two output ports. The authors argue this is possible without external channel-state feedback, because a small fraction of every element's received signal is tapped through slots into a parallel-plate waveguide and read out by two coaxial connectors. They report angle-of-arrival estimates accurate to ±3° at 20 dB SNR with 16 random on/off masks, and beam patterns with suppressed quantization lobes in full-wave simulations. If these simulations carry over to hardware, the design offers a low-cost path to intelligent wireless power transfer, communication, and sensing in one device.","feed_headline":"1-bit surface senses signal angle to ±3 degrees","feed_subtitle":"Slot-coupled waveguide readout lets one low-cost RIS detect arrival direction and redirect beams with no feedback loop.","key_machinery":"The load-bearing object is the unit cell: a 12 mm square patch on a Rogers 4003 substrate, loaded with a PIN diode (0.15 pF off, 0.1 Ω on), with a rectangular slot in the shared ground plane coupling a fraction of the incident wave into a parallel-plate waveguide. Two slot lengths, 2 mm and 6 mm, produce four reflection phases roughly 90° apart, which supplies the phase randomization needed to suppress quantization lobes of a 1-bit array; varying slot sizes randomly across the array decorrelates the periodic phase error that causes grating lobes. The sensing mechanism is the set of voltages induced on the two coaxial connectors; random on/off element states act as measurement masks, building the sensing matrix H whose columns are the port voltage differences for each reference angle, and both CGS and a four-mask MLP invert this mapping.","core_discovery":"The central claim is that a 1-bit hybrid RIS can estimate the angle of arrival of an incident wave and simultaneously redirect the reflected beam, with the sensing signal gathered by only two coaxial ports. The signal coupled into the parallel-plate waveguide is a superposition of contributions from all elements, and each random on/off pattern ('mask') provides a new measurement; with enough masks, the linear system $\\mathbf{g} = \\mathbf{H}\\mathbf{f}$ is inverted by conjugate gradient squared to recover which reference angle is present, and a four-mask multilayer perceptron does the same classification directly from port voltages. Full-wave HFSS simulations show beams steered to several target angles without grating lobes when slot lengths are randomly varied, and Monte Carlo simulations put estimation accuracy at ±3° for M = 16 masks at 20 dB SNR, with accuracy degrading gracefully at lower SNR but recovering with more masks.","pith_inferences":["The same slot-length randomization trick could be transferred to other low-bit reflective metasurfaces as a cheap grating-lobe fix, since it only changes the passive slot geometry rather than the element count or control circuitry.","If the simplified dipole forward model were replaced by a calibration matrix measured over the air, the sensing approach could hold up against mutual coupling and fabrication tolerances; the paper itself does not run that experiment.","The two-port readout is inherently one-dimensional in this work, but adding a second waveguide or extra coaxial port per axis would plausibly extend the same scheme to two-dimensional angle-of-arrival estimation.","Because the MLP trains on synthetically noised versions of the same 25 reference angles, its reported robustness partly reflects simulator consistency; testing on out-of-distribution angles or on hardware data would verify generalization beyond the training grid."],"forward_implications":["A single aperture can take over both sensing and reflecting tasks, removing the need for separate feedback hardware to learn the channel.","Beam steering with 1-bit elements and large spacing becomes practical because random slot sizes suppress quantization lobes without adding tuning bits.","The two-port compressive-sensing readout keeps RF chain count and cost low, which could make dense RIS deployments more economical.","The MLP variant shows that four masks suffice for classification once trained, cutting the measurement overhead for near-real-time angle tracking.","Accuracy degrades gracefully as SNR drops and improves as the number of masks grows, giving a tunable performance-versus-overhead trade-off."],"supporting_citations":[{"why":"Supplies the sparse-sensing HRIS approach and the sensing-matrix/estimation procedure that this paper adapts to a PPWG geometry.","marker":"[26]"},{"why":"Introduces the hybrid RIS concept with integrated sensing capability that the proposed design simplifies and makes easier to manufacture.","marker":"[9]"},{"why":"Provides the earlier experimental demonstration of sensing with hybrid RISs, the hardware baseline this design aims to strengthen with a stronger coupled signal.","marker":"[27]"},{"why":"Gives the phase-randomization method used here to mitigate quantization lobes in low-bit reconfigurable reflective surfaces.","marker":"[28]"},{"why":"Supplies the generalized reflection-law phase-gradient beam steering that determines the on/off distribution for each target direction.","marker":"[1]"},{"why":"The earlier conference paper reporting preliminary results of this specific 1-bit hybrid RIS design, extended here with full beamforming and sensing studies.","marker":"[29]"}],"fun_headline_variants":["1-bit RIS senses signal angle and steers beams","Two-port waveguide readout gives RIS sensing and steering","Hybrid RIS: slot-coupled sensing for angle, no feedback loop","Compressive sensing measures AoA on 1-bit hybrid RIS","1-bit surface: detect angle of arrival with two ports, no loops"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The angle-of-arrival results rest on a forward model in which each element's response is taken from isolated HFSS simulations and represented as a dipole whose phase depends only on its on/off state, and the same model generates both the sensing matrix and the test signals.","fun_headline_variants_meta":{"raw":{"variants":["1-bit RIS senses signal angle and steers beams","Two-port waveguide readout gives RIS sensing and steering","Hybrid RIS: slot-coupled sensing for angle, no feedback loop","Compressive sensing measures AoA on 1-bit hybrid RIS","1-bit surface: detect angle of arrival with two ports, no loops"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000368,"raw_usage":{"total_tokens":1930,"prompt_tokens":857,"completion_tokens":1073,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":473,"completion_tokens_details":{"reasoning_tokens":988}},"tokens_in":473,"tokens_out":1073,"duration_ms":10693,"temperature":1.0,"reasoning_tokens":988,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T19:43:08.691416+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Assemble or simulate the full 19-element array in a full-wave solver that includes mutual coupling, edge effects, and parallel-plate waveguide propagation, illuminate it from known angles, and compare the two-port voltage differences with the simplified dipole model's predictions; if the measured sensing matrix differs substantially, the reported ±3° accuracy will not transfer to hardware.","supporting_citations":[{"cited_title":"Sensing and reconfigurable reflection of electromagnetic waves from a metasurface with sparse sensing elements,","cited_arxiv_id":null,"evidence_quote":"Supplies the sparse-sensing HRIS approach and the sensing-matrix/estimation procedure that this paper adapts to a PPWG geometry."},{"cited_title":"A reconfigurable intelligent surface with integrated sensing capability,","cited_arxiv_id":null,"evidence_quote":"Introduces the hybrid RIS concept with integrated sensing capability that the proposed design simplifies and makes easier to manufacture."},{"cited_title":"Experimental demonstration of sensing using hybrid reconfigurable intelligent surfaces,","cited_arxiv_id":null,"evidence_quote":"Provides the earlier experimental demonstration of sensing with hybrid RISs, the hardware baseline this design aims to strengthen with a stronger coupled signal."},{"cited_title":"Mitigating quantization lobes in mmwave low-bit reconfigurable reflec- tive surfaces,","cited_arxiv_id":null,"evidence_quote":"Gives the phase-randomization method used here to mitigate quantization lobes in low-bit reconfigurable reflective surfaces."},{"cited_title":"Light propagation with phase discontinuities: generalized laws of reflection and refraction,","cited_arxiv_id":null,"evidence_quote":"Supplies the generalized reflection-law phase-gradient beam steering that determines the on/off distribution for each target direction."},{"cited_title":"Novel 1-bit hybrid reconfigurable intelligent surface with mitigated quantization lobe,","cited_arxiv_id":null,"evidence_quote":"The earlier conference paper reporting preliminary results of this specific 1-bit hybrid RIS design, extended here with full beamforming and sensing studies."}],"review_version":1}