{"id":"97455f2a-13f0-41ce-b4f9-4f54785b9738","arxiv_id":"2509.00286","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"A survey of intelligent dynamic spectrum management for satellite networks, covering regulations, cognitive radio techniques, AI/ML methods, and performance metrics for Cognitive Satellite (CogSat) systems.","lead":"Satellite networks often get fixed slices of radio spectrum, which creates artificial scarcity as more satellites launch. This paper surveys how dynamic sharing plus AI could let satellites use spectrum more flexibly, and maps the regulations, techniques, and open problems.","discovery_kind":"review","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The 'promising path' claim hinges on an unverified transferability of terrestrial CR timing and CSI assumptions to SatCom; Section VII-C-c admits CogSat is unrealized, and the sense-decide-act loop may exceed spectrum-hole lifetimes.","rationale":"The reader's weakest assumption correctly identifies the terrestrial-to-satellite transferability problem. I agree with the broad diagnosis, but I sharpen it to a specific, testable timing and CSI feasibility barrier: the sensing-decision-propagation loop in satellite networks may be too slow relative to spectrum-hole dynamics for OSA to yield the benefits claimed in the surveyed literature. This is a load-bearing concern because the paper's central claim is that DSM is a 'promising solution' to spectrum scarcity; if the primary DSM techniques cannot be realized within the physical timing constraints, the promise is unsupported. The paper's own limitation statements in Section VII-B-b and VII-C-c count as evidence that realization is open, but the concern here is stronger: not merely that realization is pending, but that the standard CR assumptions may not hold. A concrete simulation or analytical bound on the OSA loop latency in a representative GEO-LEO scenario would settle whether the concern lands. If the loop delay is small relative to hole duration, the concern is mitigated; if not, the central claim would need to be weakened to only CSA or database-driven approaches, or would require a new sensing/decision architecture. Since the reader's verdict is already CONDITIONAL, this concern does not move the verdict; it reinforces the condition and suggests the specific condition that must be verified.","tokens_in":43820,"tokens_out":4664,"duration_ms":67414,"concrete_test":"Take a concrete GEO-LEO coexistence scenario from [101] or [102], with a GEO FSS earth station as the primary user and a LEO satellite or ground terminal as the secondary user. Using realistic sensing duration from [168]/[169], processing/reconfiguration time, one-way propagation delay for the relevant orbit, and the LEO pass duration over the protected zone, compute the distribution of usable spectrum-hole duration versus the total sense-decide-act loop latency. If the loop latency exceeds the spectrum-hole duration for a substantial fraction of time, the OSA benefit vanishes, and the transferability assumption is refuted. Re-running the same comparison with the delay-free assumptions used in the simulations behind Tables II and IV would show the discrepancy directly.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that DSM through CogSat can relieve artificial spectrum scarcity rests on a transferability assumption the paper never defends: that OSA, CSA, spectrum sensing, and REM designs developed for terrestrial cognitive radio can operate with equivalent benefit in satellite networks. The paper itself concedes in Section VII-C-c that 'CogSat networks have not been realized,' and Section VII-B-b states that latency and delay 'can also affect real-time decision making and the responsiveness of CogSat networks.' The failure mode is more specific than generic delay. In OSA, the SU must sense a spectrum hole, decide, reconfigure, and transmit before the PU resumes or before LEO/GEO geometry changes the interference environment. GEO one-way propagation is on the order of 125 ms; LEO passes over a protection zone last seconds to minutes; Doppler and beam dynamics are continuous. If the total sense-decide-act loop is comparable to or longer than the stability period of a spectrum hole, the SU either misses the opportunity or interferes with the PU. Many surveyed algorithms, e.g., the DRL-based channel allocation schemes in [176]-[178], are evaluated in settings that do not include these delays, so they do not establish realizable gains. Similarly, CSA relies on an interference-temperature model with accurate PU CSI; in satellite networks the PU receiver may be an earth station served by a wide footprint, and obtaining timely CSI across international boundaries and long propagation delays is not demonstrated. Thus the surveyed techniques may not transfer, and the claim that DSM is a promising path is conditional on an unverified timing and CSI feasibility that the paper does not establish.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper is a survey of dynamic spectrum management (DSM) for satellite communications, organized around the proposition that cognitive satellite (CogSat) networks, empowered by AI/ML, can relieve artificial spectrum scarcity. It reviews enablers (satellite/terrestrial integration, cognitive radio, SDR/SDN/NFV, edge computing, blockchain), regulatory and standardization bodies (IEEE, ITU, 3GPP, ETSI, national authorities), DSM techniques (OSA, CSA, spectrum sensing, REM/databases, frequency reuse, power allocation, beam pointing/hopping/forming), ML methods for sensing, allocation, interference mitigation, and resource management, as well as training, resilience, performance metrics, and open challenges. It provides taxonomy tables (Tables I–IV) and identifies future research directions.","tokens_in":44053,"tokens_out":6358,"duration_ms":74503,"significance":"As a survey, the paper's value lies in its breadth and organization: it compiles a large corpus of recent work and structures it along multiple dimensions, including regulation, architecture, and KPIs. This is useful as a reference, particularly for readers entering the CogSat area. The paper also gives concrete attention to recent industry integrations (e.g., 5G NTN trials, GEO–LEO operators) and to performance metrics that are rarely collected in one place. If the central promise is treated as a research agenda, the survey is credible. However, the paper is mostly descriptive; it offers no machine-checked results, no quantitative feasibility analysis, and its central claim that CogSat provides 'feasible solutions' rests on an unexamined transfer of terrestrial cognitive-radio assumptions to satellite dynamics.","major_comments":[{"comment":"Reference [105] is listed under DSM Technique = 'Frequency Reuse' with the contribution 'Improved algorithms for frequency reuse in satellite communication.' However, the cited paper is titled 'A novel cognitive scheme based on dual-threshold energy detection in satellite systems,' and it is correctly cited in Section IV-C for energy detection. This is a factual misclassification in the central taxonomy. The authors should correct this row and audit the other rows in Table II for similar mismatches.","section":"Table II, [105] row"},{"comment":"The survey asserts that OSA/CSA techniques developed for terrestrial cognitive radio can be transplanted to SatCom, but it does not address the time-scale mismatch that is load-bearing for the feasibility claim made in the Conclusion. GEO one-way propagation is on the order of 125 ms; LEO passes over a protection zone last seconds to minutes; and the full sense-decide-act loop must complete before the spectrum hole closes or the geometry changes. The DRL channel-allocation results surveyed in [176]–[178] are evaluated without these delays, as the paper itself hints in Section VII-B-b. Since Section VIII calls CogSat 'feasible solutions,' the authors should either provide a quantitative or analytical discussion of this constraint, or soften the claim and frame the surveyed results as an open research hypothesis.","section":"Sections IV-A, IV-B, V-B, VII-B-b, and VIII"},{"comment":"Section VII-C-c states that 'CogSat networks have not been realized, thus adding the requirement of converting/REModeling data captured in current satellite deployments.' This admission conflicts with the absolute wording of the Conclusion, which says intelligent DSM approaches 'provide feasible solutions.' It also means the ML-based methods reviewed in Sections V-B and V-D cannot yet be trained on real CogSat data, and the simulation-only evaluations (e.g., [176]–[178]) provide limited evidence for the central claim. The authors should clarify the epistemological status of the claim: either present CogSat DSM explicitly as an open research agenda with illustrative results, or supply a separate analysis of what is needed to close the data and deployment gap.","section":"Section VII-C-c and Section VIII"}],"minor_comments":[{"comment":"Reference [183] is the same paper as [177] (Hu, Liu, Chen, Wang, Wang, 'A Deep Reinforcement Learning-Based Framework for Dynamic Resource Allocation in Multibeam Satellite Systems,' IEEE Commun. Lett. 2018). The duplicate row should be removed or replaced with the intended distinct reference.","section":"Table IV, [183] row"},{"comment":"The sentence 'The formal model of ML is the Markov Decision Process' is technically incorrect: the MDP is a formal model for reinforcement learning, not for ML in general. Please rephrase, e.g., 'RL problems are often formalized as Markov Decision Processes.'","section":"Section II-C3"},{"comment":"Minor typos: 'Very High Throughput Satellites (VITS)' should be 'VHTS'; 'Rount Trip Time' should be 'Round Trip Time'; 'Sectrum Sensing' in the Table II caption should be 'Spectrum Sensing'; 'two-way-really aided' in Table II should likely be 'two-way relay aided'; 'PU precedencetechniques' in Table III needs a space.","section":"Sections V-D and VI-G"},{"comment":"The phrase 'based on the spectrum presidence' contains a typo: 'presidence' should be 'precedence' or 'priority.'","section":"Section IV-E1"},{"comment":"Several references are inconsistently formatted or incomplete: [29] has no author list, [171] appears as 'IEEE TIEEE Commun. Mag.' which is malformed, and many URL entries lack access dates. A careful reference cleanup is needed.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The survey is broad and likely useful as a reference, but the central feasibility claim needs stronger critical framing. The transferability/time-scale issue is my main technical concern; the authors should be asked to address it explicitly rather than merely append a caveat. The duplicate and misclassified entries in the taxonomy tables also suggest the need for a systematic data audit before publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague, here’s my take on arXiv:2509.00286. It’s a survey, not a discovery, and it does what a good survey should: pulls together regulation, DSM techniques, AI/ML methods, and KPIs for satellite spectrum management into one place. The taxonomy tables (Table II and IV) are the real contribution – they let a newcomer see at a glance what has been tried and where the gaps are. The coverage is genuinely broad, with citations to most of the relevant recent work. I also appreciate that the paper is honest about the status of CogSat: it says plainly in Section VII-C-c that these networks have not been realized and lists latency, data scarcity, and regulatory friction as open challenges. That honesty is worth a lot in this literature.\n\nThe soft spots are real but manageable. There are citation mismatches in the tables: [105] is described as frequency reuse but the actual paper is about dual-threshold energy detection, and [183] duplicates [177]. Those are the kind of errors an editor should ask the authors to fix before publication. There is also a general tendency to assert that terrestrial CR techniques 'can be' applied to satellite networks without much argument. The stress-test note about timing and CSI transferability is a valid concern, but it is not a flaw in the survey’s logic. The survey’s job is to map the field, not to prove feasibility; it explicitly flags these as open challenges. So I would not hold that against it.\n\nThe paper could also use a copyedit – some phrasing is rough and there are a few unsupported claims, like the one about DRL achieving 'resolutions beyond human capabilities' (Section II-C3), which reads more like marketing than analysis.\n\nOverall: this is a useful synthesis for researchers entering the area and for industry people trying to understand the regulatory and technical landscape. It deserves a serious referee turn, with a request to fix the cataloging errors and tone down some of the promotional language. I would not cite it in my own work within the next year, but I would put it on the reading list for a student starting in this area.","headline":"Broad, honest survey of satellite spectrum management; fixable cataloging errors, real but explicitly flagged transferability concerns.","tokens_in":44665,"tokens_out":2299,"would_cite":false,"duration_ms":26125,"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":"Cognitive satellite networks—satellites that sense and reuse idle licensed spectrum—can relieve the satellite spectrum squeeze, provided regulators and standardization bodies catch up.","keywords":["cognitive satellite","dynamic spectrum management","satellite communications","spectrum sensing","opportunistic spectrum access","machine learning","spectrum regulation","LEO constellations"],"falsifier":"A concrete test: equip a LEO constellation to sense and reuse a GEO downlink band at scale and measure interference against the ITU's 20-25% noise-margin budget. If secondary transmissions push GEO link outage beyond that margin under realistic traffic—or if a WRC session explicitly rejects secondary access in those bands—the paper's central claim fails.","tokens_in":1267,"feed_emoji":"🛰️","tokens_out":1586,"duration_ms":59325,"temperature":0.7,"pith_summary":"Satellite operators are running out of usable spectrum, not because the bands are full but because exclusive licensing leaves much of them idle at any given time and place. This survey argues that cognitive satellite (CogSat) networks—satellite systems that sense, learn, and adaptively reuse underutilized spectrum—are the workable remedy. Drawing on cognitive-radio techniques developed for terrestrial networks, the paper maps how opportunistic and concurrent spectrum access, spectrum sensing, radio environment maps, and AI/ML-based allocation could be transplanted into multi-orbit satellite systems. It shows that the technical pieces exist in scattered form; what is missing is a regulatory and standardization framework that permits dynamic sharing. If realized, the payoff is higher spectral efficiency and lower entry barriers for new satellite operators.","feed_headline":"AI-driven sharing could end the satellite spectrum squeeze","feed_subtitle":"A survey maps how cognitive satellites would reuse idle licensed spectrum—and why regulation still blocks them.","key_machinery":"The CogSat network concept: the integration of cognitive radio into satellite communications, carried by the primary-user/secondary-user hierarchy and the DSM toolbox—opportunistic spectrum access (reuse of spectrum holes), concurrent spectrum access (underlay sharing under an interference temperature), spectrum sensing, and radio-environment-map databases. AI/ML supplies the learning layer; SDR, SDN, and NFV supply the reconfigurability. The paper's taxonomy of these techniques organizes the entire survey and its reading of the literature.","core_discovery":"The paper's central claim is that the spectrum scarcity constraining satellite communications is substantially artificial—caused by rigid, long-term, exclusive allocations—and that dynamic spectrum management through cognitive satellite networks can relieve it. CogSat networks adapt cognitive radio to space: secondary users detect and reuse temporarily unused licensed spectrum without harmful interference to primary users. The survey organizes the full enabling stack—network integration across orbits and with terrestrial 5G, SDR/SDN/NFV virtualizations, AI/ML for sensing, allocation, and interference mitigation, and database/REM approaches—and reads each against the regulatory landscape. It","pith_inferences":["The terrestrial TV-white-space experience suggests a plausible near-term path: start with database-driven sharing in one band, then layer sensing once interference behavior is understood.","The paper's own admission that no CogSat system has been built implies the strongest evidence would be a live multi-orbit testbed rather than further simulations.","Federated/distributed learning, which the paper flags as an open gap, may be the natural fit for constellations because it keeps training data grounded and shares only model updates.","If regulatory consensus fails, the 'artificial scarcity' may persist regardless of technical progress—making standardization the true control variable."],"forward_implications":["If CogSat works, LEO constellations can reuse GEO-allocated bands by sensing and avoiding inline interference, multiplying effective capacity without new spectrum.","AI/ML-based sensing can push detection into the low-SINR regime (near -20 dB) that long satellite links require, making opportunistic access practical.","KPIs such as spectrum utilization factor, detection/false-alarm probabilities, and service retainability would become the standard yardsticks for spectrum-sharing arrangements.","Regulatory bodies (ITU, 3GPP, ETSI) would need flexible licensing, machine-readable policies, and defined interference margins before any of this deploys.","The choice between onboard, on-ground, and hybrid ML training determines whether cognitive decisions can keep up with LEO dynamics and delay."],"supporting_citations":[{"why":"Supplies the DSM/CR foundation and the OSA-versus-CSA contrast the survey builds on.","marker":"[2]"},{"why":"Defines the dual-CogSat architecture and GEO/non-GEO coexistence scenarios.","marker":"[6]"},{"why":"Baseline survey of CR and CogSat application scenarios that the paper extends.","marker":"[10]"},{"why":"Early demonstration that cognitive-radio techniques can be applied to satellite communication.","marker":"[32]"},{"why":"Supplies the CogSat network architecture classification used in the survey.","marker":"[122]"},{"why":"Grounds the interference models and spectrum-sensing technique selection.","marker":"[123]"},{"why":"Sets the low-SNR sensing threshold (IEEE 802.22) the paper cites for CogSat.","marker":"[130]"},{"why":"The reference on AI/ML integration and onboard-computation limits that frames the ML discussion.","marker":"[9]"}],"fun_headline_variants":["Cognitive satellites reuse idle spectrum—regulators are the bottleneck","AI could let satellites share spectrum, but policy lags","Satellite spectrum scarcity eased by dynamic sharing","Survey: cognitive satellites can end artificial spectrum shortage","Cutting through spectrum gridlock with AI and sharing"],"cache_read_input_tokens":46336,"weakest_assumption_plain":"Terrestrial cognitive-radio techniques—sensing, opportunistic access, underlay sharing—transfer to satellite networks with equivalent benefit, although no cognitive satellite network has actually been built and satellite channels differ in delay, dynamics, and regulation.","fun_headline_variants_meta":{"raw":{"variants":["Cognitive satellites reuse idle spectrum—regulators are the bottleneck","AI could let satellites share spectrum, but policy lags","Satellite spectrum scarcity eased by dynamic sharing","Survey: cognitive satellites can end artificial spectrum shortage","Cutting through spectrum gridlock with AI and sharing"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000196,"raw_usage":{"total_tokens":1184,"prompt_tokens":719,"completion_tokens":465,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":463,"completion_tokens_details":{"reasoning_tokens":390}},"tokens_in":463,"tokens_out":465,"duration_ms":5987,"temperature":1.0,"reasoning_tokens":390,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T13:44:50.881444+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A concrete test: equip a LEO constellation to sense and reuse a GEO downlink band at scale and measure interference against the ITU's 20-25% noise-margin budget. If secondary transmissions push GEO link outage beyond that margin under realistic traffic—or if a WRC session explicitly rejects secondary access in those bands—the paper's central claim fails.","supporting_citations":[],"review_version":1}