{"id":"9281a456-8b5c-4390-a297-0474ab8879da","arxiv_id":"2412.20524","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Ns3Sionna embeds Sionna's ray tracing channel model into ns-3, with coherence-time caching and parallelized point-to-multipoint computation for realistic Wi-Fi simulation.","lead":"A new open-source module, Ns3Sionna, connects the ns-3 network simulator to the Sionna ray tracing engine, replacing statistical channel models with physically computed radio paths in 3D scenes. It adds caching and parallelization so small and medium Wi-Fi simulations with realistic indoor and outdoor channels become practical.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 'more realistic' claim rests on unvalidated Sionna RT accuracy; free-space Friis agreement in Section V cannot support indoor/outdoor realism, so measurement-based validation is required.","rationale":"I agree with the reader's weakest-assumption identification. The paper's own validation is minimal and does not touch the scenarios where the claim is made. The concern is not internal inconsistency: the architecture and caching logic are coherent, and the free-space match in Section V is a useful sanity check. The issue is that the load-bearing empirical premise—that Sionna RT with user-supplied scenes and materials yields realistic indoor/outdoor channels—is assumed, not demonstrated. The validation in Fig. 4 only confirms equivalence in free space; it does not compare against ns-3's statistical models in a non-free-space setting. Thus the abstract's comparison claim is under-supported. I also note the paper does provide real engineering value: the point-to-multipoint batching, pre-caching with coherence-time TTL, and open-source release are concrete contributions. Since the reader already conditioned acceptance on measurement-based validation, my analysis supports that verdict rather than changing it.","tokens_in":11524,"tokens_out":2926,"duration_ms":31674,"concrete_test":"Run the two-room indoor scenario from Section VI-A at 5 GHz in Ns3Sionna and in ns-3 with HybridBuildingsPropagationLossModel or ITU-R 1238, and compare both against a measurement campaign (or a published measurement dataset with the same room geometry and frequencies) along an equivalent STA trajectory. Compute, for each model, the mean absolute error and the spatial/temporal autocorrelation of received power relative to the measurements. If Ns3Sionna is not closer to the measured data than the ns-3 models on these metrics, the 'more realistic' claim is unsupported. For the outdoor claim, repeat with a published urban microcell measurement set in a comparable street canyon.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim ('more realistic path and delay loss estimates ... particularly spatial and temporal correlation') depends on Sionna RT producing channels that are accurate for the indoor and outdoor scenes used in Section VI. The only validation provided (Section V, Fig. 4) is free-space agreement with the Friis formula and a constant-speed delay model. That check exercises only the integration plumbing: any distance-dependent model would pass it, and it does not test multipath, wall diffraction, material parameters, shadowing, or correlation structure. Sections VI-A and VI-B then present indoor and outdoor examples without any comparison to measured channels or to the ns-3 statistical models that the claim says are less realistic. Consequently, the distinctive part of the claim—that the ray-traced channels are more realistic in complex environments—has no empirical anchor in this paper. This is a missing-validation concern rather than an identified error: Sionna RT may well be accurate, but the manuscript does not supply the evidence needed to support the stated superiority.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents Ns3Sionna, a software framework that integrates the Sionna RT ray-tracing engine into the ns-3 network simulator. Two new ns-3 models (SionnaPropagationLossModel and SionnaPropagationDelayModel) offload channel computation to Sionna RT, and the framework adds a ray-tracing-based mobility model, caching of channel state information, channel reciprocity reuse, and parallel computation of point-to-multipoint channels. The authors claim that Ns3Sionna provides more realistic path and delay loss estimates than existing ns-3 propagation models, especially in spatial and temporal correlation, and they report example indoor and outdoor scenarios plus benchmark results showing large runtime overheads that are mitigated by caching and GPU-parallel predictive computation.","tokens_in":11710,"tokens_out":3717,"duration_ms":40707,"significance":"If the realism claim were properly supported, this would be a valuable open-source contribution: it would give the ns-3 community a way to run protocol-level simulations over ray-traced, scene-specific channels with spatial and temporal consistency, and the repository makes the implementation available. The paper also demonstrates useful engineering ideas, such as exploiting channel coherence time for caching and computing point-to-multipoint channels in one Sionna call to amortize ray-tracing cost. However, the central claim of the paper—that Ns3Sionna is more realistic than existing ns-3 models in indoor and outdoor environments—is not currently validated. The only quantitative validation is a free-space comparison against the Friis model, which is a self-consistency check of the integration rather than evidence for multipath, shadowing, or correlation accuracy. The indoor and outdoor examples are qualitative demonstrations without comparison to measured channels or to the ns-3 statistical models that the paper claims are less realistic. The performance study is useful and honestly reports large overheads, but it does not compensate for the missing validation of physical realism.","major_comments":[{"comment":"The validation section compares Ns3Sionna only with ns-3's Friis propagation loss model and constant-speed delay model in free space. This checks that the integration plumbing works, but it does not validate the paper's central claim of more realistic path and delay loss estimates in indoor and outdoor environments. A distance-dependent Friis-style model would pass this same test in free space, so the test does not exercise multipath, reflections, diffraction, wall penetration, material parameters, shadowing, or any correlation structure. The authors should add validation against measured channel data or against established ray-tracing benchmarks in representative indoor and outdoor scenes, including metrics such as path loss, delay spread, and spatial/temporal correlation.","section":"Section V, Figure 4"},{"comment":"The indoor and outdoor example scenarios are presented as demonstrations of the framework's output, but no quantitative comparison is made against either measured channels or the existing ns-3 propagation models mentioned in the introduction (e.g., LogDistance, ITU-R P.1238, or HybridBuildingsPropagationLossModel). As a result, the abstract's claim that Ns3Sionna 'provides more realistic path and delay loss estimates for both indoor and outdoor environments than existing ns-3 propagation models' is not supported by the evidence in the paper. The authors should run the same scenarios with the relevant ns-3 baseline models and report quantitative differences in path loss, delay, and correlation metrics.","section":"Section VI, Figures 6 and 8"},{"comment":"The propagation delay is defined as the delay of the shortest path among all ray-traced paths. In the free-space validation this reduces exactly to the constant-speed delay d/c, so the delay comparison in Figure 4 is a tautology for the testing environment. The delay model's realism in multipath environments is not assessed at all; the paper reports no delay spread or first-arrival-path statistics for the indoor or outdoor scenarios. Please add delay-related metrics (e.g., RMS delay spread or excess delay) to the scenario evaluation, since delay estimation is part of the central claim.","section":"Section IV-B and Section V"}],"minor_comments":[{"comment":"The text says 'The first two points are addressed by caching' and then 'The second point is addressed as follows,' where the second sentence appears to refer to the third design point (avoiding far-away receivers). Please correct the numbering to avoid confusion.","section":"Section IV-A"},{"comment":"There is a grammatical error in 'message passing as it gives are the possibility to run ns-3 and Sionna on different computers.' Please revise this sentence.","section":"Section IV-C"},{"comment":"The y-axis labels in Figure 9 appear to be missing an exponent (the text mentions differences of four orders of magnitude, but the axis shows '100' and '105'). Please format the axis with explicit powers of ten.","section":"Section VIII-A, Figure 9"},{"comment":"Figure 11 has no y-axis label and the caption does not state what the speedup is relative to. Please clarify the baseline for the speedup computation.","section":"Section VIII-B, Figure 11"},{"comment":"The term 'PHZ abstraction models' appears to be a typo for 'PHY abstraction models.' Please fix this and similar typos throughout the discussion.","section":"Section VIII-C"},{"comment":"The URLs in references [15] and [17] are missing the '://' separator (e.g., 'https:www.nsnam.org'). Please correct them.","section":"References [15] and [17]"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is better positioned as a systems/tools paper than as a validation study. The core engineering contribution is real and the open-source release is a plus, but the advertised realism result is not demonstrated within the paper. The missing validation is the load-bearing gap: without comparison to measurements or to the ns-3 baseline models, the comparative claim is unsupported. Adding such a comparison in the same scenarios is feasible within the manuscript's scope and should be required before publication. The performance evaluation is honest about overheads, but the paper should also temper or carefully qualify the abstract's claim if the validation remains limited."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear [Name],\n\nThe short version: this is a serious engineering contribution—the first real integration of Sionna RT into ns-3, with a sensible caching and pre-computation scheme that makes ray-traced channels usable at packet level—but the paper overstates what it has demonstrated. The headline claim about \"more realistic\" indoor and outdoor channels is not backed by the validation section. Section V only compares free-space path loss against ns-3's Friis model. That check exercises the integration plumbing, not the physics. Any distance-dependent model would pass it.\n\nWhat is genuinely new and good: the point-to-multipoint pre-caching, where one Sionna call computes channels from a sender to all potential receivers, is a smart use of the engine's parallelism. The virtual-node idea—simulating future node positions so their channels are computed speculatively—is also original relative to the ns-2 ray-tracing work they cite. The reciprocity cache keyed by coherence time is the right way to avoid redundant computation, and the paper honestly reports the costs: ray tracing is one to four orders of magnitude slower than ns-3's statistical models, and the authors say so plainly. The code is open source, the architecture is clearly described, and the related work is fair, with appropriate citations to Sionna and to Dricot and De Doncker without any self-citation inflation.\n\nSoft spots, in proportion: the missing measurement validation is the main one. The indoor and outdoor example scenarios (Section VI) show plausible-looking spatial and temporal correlation, but there is no comparison to measured channels or even to ns-3's statistical models, so the stated superiority of the ray-traced channels over those models is an assertion, not a result. This is a missing-validation issue rather than an identified error—Sionna RT may well be accurate—but the manuscript does not supply the evidence. A secondary, minor issue: the benchmark numbers in Table I and Figures 9–11 are reported as single values with no error bars, which is fine when the gaps are orders of magnitude, but the GPU speedup factors (14.2548×, 7.9023×) imply a precision that likely isn't there. I'd also ask for a commit hash or dependency pins in the paper, though the repo may already provide that.\n\nWho this is for: anyone doing system-level Wi-Fi or 6G protocol simulation who needs spatially and temporally consistent channels and is willing to pay a big runtime cost. It is a tools paper, and the tool looks real and usable.\n\nRecommendation: send to peer review. The engineering is sound and the caching ideas are worth publishing, but the realism claim needs either a measurement-based comparison or softer language. A serious referee should ask for one of those.","headline":"A solid, useful ns-3/Sionna integration with a real caching design, but the 'more realistic' claim outruns the validation, which only checks free-space Friis agreement.","tokens_in":12229,"tokens_out":1314,"would_cite":true,"duration_ms":15444,"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":"Ns3Sionna replaces ns-3's statistical radio models with ray-traced channels that are spatially and temporally correlated.","keywords":["ray tracing","ns-3","Sionna RT","channel model","spatial correlation","temporal correlation","network simulation","coherence time"],"falsifier":"Run a measurement campaign in a real two-room indoor layout with an open door, recording received power and delay along a 1 m/s walking path; the central claim fails if Ns3Sionna does not reproduce the measured spatial correlation of shadowing and the sharp wall-induced attenuation more closely than ns-3's independent log-normal shadowing model.","tokens_in":11361,"feed_emoji":"📡","tokens_out":8023,"duration_ms":66834,"temperature":0.7,"pith_summary":"Ns3Sionna is a software module that lets the ns-3 network simulator compute wireless channels by ray tracing instead of by the simplified statistical and stochastic models it normally uses. The paper claims that for a given 3D scene, the module produces path loss and propagation delay estimates that are more realistic than existing ns-3 models, especially in how the channel evolves smoothly across space and time. Because multipath, shadowing, and frequency-selective fading are computed from the geometry and materials of the environment, the channel state information also becomes available for sensing applications. The authors back this with a free-space validation against the Friis formula and with indoor and outdoor example scenarios, and they show that pre-caching based on coherence time and point-to-multipoint ray tracing make the approach computationally feasible for small to medium networks.","feed_headline":"Ray-traced channels give ns-3 realistic space-time correlation","feed_subtitle":"New Ns3Sionna module couples Sionna RT with ns-3 to capture multipath and correlated shadowing.","key_machinery":"The load-bearing mechanism is Sionna RT, a GPU-accelerated ray tracer that computes radio propagation paths through a user-provided 3D scene with material properties, returning a channel impulse response for each transmitter-receiver pair. Ns3Sionna wraps this into two ns-3 models, SionnaPropagationLossModel and SionnaPropagationDelayModel, and adds a cache keyed by coherence time and channel reciprocity so that channels are not recomputed unnecessarily. To exploit parallelism, the framework asks Sionna to compute the full point-to-multipoint channel from a sender to all potential receivers, including virtual nodes at predicted future positions, in one call, which makes the expensive ray tracing affordable for simulations with a small to medium number of mobile nodes.","core_discovery":"The central claim is that coupling Sionna RT's physically based ray tracing with ns-3 yields channel realizations, expressed as path loss, delay, and channel frequency response, that reflect the actual propagation environment rather than a probability distribution. In particular, the paper argues that Ns3Sionna's channel estimates are more realistic than ns-3's built-in models in terms of spatial and temporal correlation: a moving station sees smooth, correlated changes in received power as it moves through the 3D scene, and the channel remains stable over the coherence time. The module also provides a ray-tracing-based mobility model that keeps nodes inside the 3D geometry, and it exploits channel reciprocity, coherence time, and parallel point-to-multipoint computation to keep runtime practical.","pith_inferences":["Not stated in the paper: the realism claim inherits all of Sionna RT's error; the free-space validation only checks agreement with Friis, so the indoor and outdoor advantage over ns-3 remains an assumption until a measurement campaign compares the two.","Not stated in the paper: Ns3Sionna could serve as a data generator for machine-learning indoor positioning, since its spatially correlated shadowing is a more realistic training distribution than independent log-normal draws.","Not stated in the paper: the pre-caching pattern, computing a point-to-multipoint channel and predicting future positions, is a general recipe for coupling expensive physical models into any packet-level simulator."],"forward_implications":["Wi-Fi protocol simulations in indoor environments can now exhibit multipath fading, wall shadowing, and frequency-selective channels that standard ns-3 statistical models cannot produce.","Because channels are spatially consistent, positioning, sensing, and fine-grained channel state information studies can be done inside ns-3 without a separate propagation tool.","Coherence-time caching and point-to-multipoint computation keep the ray-tracing overhead practical for small to medium networks, enabling parameter sweeps that would otherwise take years.","The ray-tracing mobility model keeps nodes inside the 3D geometry and makes them bounce off obstacles, removing the unrealistic bounding-box random walk limitation.","The same design extends to LTE via the SpectrumChannel, to MIMO, and to reconfigurable intelligent surfaces, since Sionna RT already provides the needed primitives."],"supporting_citations":[{"why":"Supplies Sionna RT, the differentiable ray-tracing engine that computes propagation paths and channel impulse responses from 3D scenes.","marker":"[9]"},{"why":"Describes the Sionna library that includes the ray-tracing extension and provides the underlying link-level simulation capabilities.","marker":"[11]"},{"why":"Provides the foundational review of ray tracing for radio propagation modeling that justifies the technique's use for accurate channel prediction.","marker":"[7]"},{"why":"Documents the ns-3 simulator architecture and its existing propagation models, which Ns3Sionna extends and compares against.","marker":"[15]"},{"why":"Gives the coherence-time formula used to set the cache time-to-live and decide when a channel can be reused without recalculation.","marker":"[13]"}],"fun_headline_variants":["Ns3Sionna adds ray-traced realism to ns-3 simulations","Ray-traced channels make ns-3 simulations more realistic","ns-3 meets Sionna: physically accurate channel simulation","GPU ray tracing brings real-world channels to ns-3","Ns3Sionna: realistic channels and mobility for ns-3"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The entire approach assumes that Sionna RT, given a user-created 3D scene and material parameters, predicts channels accurately enough to represent real wireless propagation; the paper validates this only in free space against the Friis formula and offers no measurement campaign for its indoor and outdoor realism claims.","fun_headline_variants_meta":{"raw":{"variants":["Ns3Sionna adds ray-traced realism to ns-3 simulations","Ray-traced channels make ns-3 simulations more realistic","ns-3 meets Sionna: physically accurate channel simulation","GPU ray tracing brings real-world channels to ns-3","Ns3Sionna: realistic channels and mobility for ns-3"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000201,"raw_usage":{"total_tokens":1393,"prompt_tokens":973,"completion_tokens":420,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":589,"completion_tokens_details":{"reasoning_tokens":331}},"tokens_in":589,"tokens_out":420,"duration_ms":3817,"temperature":1.0,"reasoning_tokens":331,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T23:18:41.145238+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run a measurement campaign in a real two-room indoor layout with an open door, recording received power and delay along a 1 m/s walking path; the central claim fails if Ns3Sionna does not reproduce the measured spatial correlation of shadowing and the sharp wall-induced attenuation more closely than ns-3's independent log-normal shadowing model.","supporting_citations":[{"cited_title":"Ray Tracing for Radio Propagation Modeling: Principles and Applications,","cited_arxiv_id":null,"evidence_quote":"Provides the foundational review of ray tracing for radio propagation modeling that justifies the technique's use for accurate channel prediction."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Documents the ns-3 simulator architecture and its existing propagation models, which Ns3Sionna extends and compares against."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the coherence-time formula used to set the cache time-to-live and decide when a channel can be reused without recalculation."}],"review_version":1}