{"schema_version":1,"research_id":"eoa_inverse_innovation_exp05_external_evaluation_20260803","source_assessment_id":"invariant_mode_decomposition_design__information_theory:P1:v0","cell_id":"invariant_mode_decomposition_design__information_theory","search_queries":["predictive coding error propagation packet loss intra refresh telemetry multiple description joint source channel coding official paper","mode selective refresh predictive coding eigenmodes error propagation packet loss multivariate telemetry","HEVC intra refresh error propagation reference picture resynchronization standard","control-aware communication rate distortion goal-oriented communication sensor telemetry modal","site:public.ccsds.org predictive telemetry compression error propagation packet loss sensor data CCSDS","site:ntrs.nasa.gov telemetry compression packet loss predictive coding error propagation multichannel","CCSDS 123 predictive multispectral compression standard predictor official PDF","NASA telemetry data compression onboard correlated sensor streams need bandwidth","bit allocation unstable modes networked control quantization eigenvalues sensor communication paper","rate allocation system eigenmodes quantized state estimation unstable modes communication","predictive coding error transition matrix eigenvalue error propagation refresh coding","modal transform coding error propagation feedback predictive quantization reset coefficient","joint direction refresh predictive codec error propagation eigenvector coefficient packet loss","transform domain intra refresh predictive coding packet loss eigenvectors","error resilient predictive coding periodic refresh DPCM packet loss multichannel paper","task based quantization linear estimation transform coding eigenvectors bit allocation","BLS software developers median pay May 2025 official","BLS computer and information research scientists median pay 2025 official","Karhunen Loeve transform coding bit allocation correlated sources primary paper PDF","transform coding correlated sources eigenvectors variance bit allocation IEEE classic paper","site:stanford.edu transform coding Karhunen Loeve bit allocation pdf","\"Convergence Rate of Quantization Error in Networked Control Systems\" PDF Fang Antsaklis","site:nd.edu \"Convergence Rate of Quantization Error\""],"sources":[{"source_id":"S1","title":"Chapter 11: Onboard Systems","publisher":"NASA Science","url":"https://science.nasa.gov/learn/basics-of-space-flight/chapter11-1/","source_class":"GOVERNMENT_OR_REGULATOR","publication_date":"n.d.; current page accessed 2026-08-03","accessed_at":"2026-08-03","claims_supported":["Spacecraft telemetry combines thousands of science and engineering measurements.","Onboard systems use compression to reduce transmitted bits and encoding to reduce data loss.","Telemetry content and format can be changed for mission phases, downlink rates, and anomaly diagnosis."]},{"source_id":"S2","title":"CCSDS Active Publications and Data Compression Working Group","publisher":"Consultative Committee for Space Data Systems","url":"https://ccsds.org/publications/allpubs/","source_class":"OFFICIAL_ORGANIZATION_DATA","publication_date":"n.d.; current page accessed 2026-08-03","accessed_at":"2026-08-03","claims_supported":["CCSDS states that onboard compression is needed because spacecraft storage and downlink capacity are limited.","The CCSDS data-compression activity develops low-complexity compression standards exploiting three-dimensional correlation.","The page identifies an organizationally credible adopter and standardization channel through the CCSDS Data Compression Working Group, with NASA/JPL and ESA contacts.","CCSDS also expresses demand for higher telemetry rates, better performance, lower cost, mass and power, and selectable reliability or delivered-data quality."]},{"source_id":"S3","title":"Consistent Estimation of Erased Data in a DPCM Based Multiple Description Coding System","publisher":"IBM Research","url":"https://research.ibm.com/publications/consistent-estimation-of-erased-data-in-a-dpcm-based-multiple-description-coding-system","source_class":"PRIMARY_RESEARCH","publication_date":"2003-05-01","accessed_at":"2026-08-03","claims_supported":["Packet losses cause error propagation in predictive coding environments.","Multiple-description coding and sequence-consistent estimation are established approaches for recovering lost DPCM data.","Error resilience can be added while retaining standard-codec-compatible predictive encoders."]},{"source_id":"S4","title":"VK_KHR_video_encode_intra_refresh","publisher":"Khronos Group, Vulkan Documentation Project","url":"https://docs.vulkan.org/features/latest/features/proposals/VK_KHR_video_encode_intra_refresh.html","source_class":"STANDARD","publication_date":"n.d.; current proposal accessed 2026-08-03","accessed_at":"2026-08-03","claims_supported":["Predictive-reference errors can propagate until a decoder refresh boundary.","Full intra frames stop propagation but cause bitrate spikes; gradual intra refresh amortizes the cost across frames.","Implementations must prevent clean refreshed regions from referencing dirty regions, and incorrect refresh controls can defeat error resilience.","Intra-refresh support and constraints vary by codec and implementation."]},{"source_id":"S5","title":"Convergence Rate of Quantization Error in Networked Control Systems","publisher":"University of Notre Dame; presented at the 14th Mediterranean Conference on Control and Automation","url":"https://www3.nd.edu/~pantsakl/publications/357-MED06.pdf","source_class":"PRIMARY_RESEARCH","publication_date":"2006-06-28","accessed_at":"2026-08-03","claims_supported":["A known linear state matrix and fixed bit rate can be used to optimize quantization-error convergence.","Communication resources can be allocated dynamically among unstable system modes.","The work supplies a close mathematical analogue: eigenvalue-dependent, mode-level bit allocation under a fixed rate to suppress propagated error.","The analysis assumes an LTI model, known encoder-decoder state, and restrictive stability conditions, limiting direct transfer to a deployed nonlinear or drifting telemetry codec."]},{"source_id":"S6","title":"Hardware-Limited Task-Based Quantization","publisher":"IEEE Transactions on Signal Processing; author-hosted by Weizmann Institute of Science","url":"https://www.weizmann.ac.il/math/yonina/sites/math.yonina/files/Hardware-Limited_Task-Based_Quantization.pdf","source_class":"PRIMARY_RESEARCH","publication_date":"2019-08-19","accessed_at":"2026-08-03","claims_supported":["Quantizer design can target a downstream inference task instead of source-reconstruction error alone.","Linear pre-quantization combining, dimensionality reduction, and task-aware allocation under a finite bit budget are established research practices.","Task-aware quantization can outperform task-ignorant designs in studied settings, but the cited work does not test packet-loss recovery or predictive-codec error modes."]},{"source_id":"S7","title":"Multispectral Data Compression through Transform Coding and Block Quantization","publisher":"Purdue University Libraries, LARS Technical Reports","url":"https://docs.lib.purdue.edu/larstech/43/","source_class":"PRIMARY_RESEARCH","publication_date":"1972-01","accessed_at":"2026-08-03","claims_supported":["Transform coding of correlated multispectral data using joint coefficients is longstanding prior art.","Karhunen–Loève components, fixed-rate bit allocation, reconstruction MSE, and downstream classification performance were jointly evaluated.","The candidate differs from source-covariance transform coding only if it selects directions from error propagation dynamics and verifies task sensitivity rather than selecting source principal components."]},{"source_id":"S8","title":"Computer and Information Research Scientists: Occupational Outlook Handbook","publisher":"U.S. Bureau of Labor Statistics","url":"https://www.bls.gov/ooh/computer-and-information-technology/computer-and-information-research-scientists.htm","source_class":"GOVERNMENT_OR_REGULATOR","publication_date":"2025-08-28","accessed_at":"2026-08-03","claims_supported":["The May 2024 median wage was $140,910 annually or $67.74 hourly for computer and information research scientists.","Software developer, QA analyst, and tester median pay was $131,450 in May 2024.","These wage benchmarks support labor-based resource-equivalent estimates but do not include overhead, specialized flight assurance, hardware, or mission delay costs."]}],"problem_evidence":{"support":"MODERATE","rationale":"Predictive-coding packet loss is directly documented to propagate error, and standards practice treats decoder refresh as necessary to bound propagation. NASA and CCSDS document constrained telemetry resources and the use of compression and loss protection. However, no opened source demonstrates the candidate's narrower empirical premise: a stable, consequential cross-stream eigenmode hidden by acceptable per-stream distortion in an operational telemetry codec.","source_ids":["S1","S2","S3","S4"]},"stakeholder_evidence":{"support":"MODERATE","rationale":"CCSDS explicitly expresses the need for bandwidth-efficient, low-complexity onboard compression and identifies NASA/JPL and ESA participants who could sponsor standards-oriented evaluation. NASA describes codec and telemetry-content decisions within spacecraft data systems. This establishes a credible adopter class and organizational channel, but no source expresses demand for mode-targeted refresh specifically, and the mission-specific codec owner, network operator, downstream owner, and live-change authority remain unidentified.","source_ids":["S1","S2"]},"prior_art":{"proximity":"SUBSTANTIAL_COLLISION","closest_analogues":[{"name":"Dynamic bit assignment among unstable modes","similarity":"Fang and Antsaklis optimize a fixed communication rate across unstable modes of a known state matrix to improve quantization-error convergence, closely matching the proposal's eigenmode-based protection logic.","remaining_difference":"Their object is a quantized networked-control plant under restrictive LTI assumptions, not reconstruction-error dynamics inside a predictive telemetry codec; they do not add task-sensitivity ranking, packet-loss replay, residual governance, or protected modal refresh coefficients.","source_ids":["S5"]},{"name":"Gradual intra refresh and decoder refresh points","similarity":"Established codec practice spends bounded intra-coded capacity over successive frames to stop predictive-reference error propagation without a full-frame bitrate spike.","remaining_difference":"Refresh targets image regions or codec partitions rather than joint cross-stream directions selected from an estimated error-transition operator.","source_ids":["S4"]},{"name":"Task-based quantization","similarity":"Existing research allocates finite quantization resources after linear combination according to downstream task performance rather than raw signal fidelity alone.","remaining_difference":"It does not select weakly damped error-propagation modes or evaluate disturbance recovery in a synchronized predictive encoder-decoder.","source_ids":["S6"]},{"name":"Karhunen–Loève transform coding with fixed-rate allocation","similarity":"Longstanding transform coding projects correlated sensor or image data onto joint eigen-directions, allocates bits among coefficients, and can evaluate both reconstruction and downstream classification.","remaining_difference":"KLT directions arise from source covariance and variance, whereas the surviving distinction requires directions derived from the codec's reconstruction-error transition and prioritized by propagation plus task consequence.","source_ids":["S7"]},{"name":"Multiple-description recovery for erased DPCM data","similarity":"The work directly addresses packet-loss-induced error propagation in predictive coding and adds redundant information to recover missing samples.","remaining_difference":"It estimates erased DPCM data through multiple descriptions and consistency search rather than resetting selected joint modal error coordinates.","source_ids":["S3"]}],"distinctive_claim_remaining":"Under an identical total refresh/protection-bit budget and unchanged mandatory per-stream coverage, transmitting protected coefficients for joint directions selected jointly by held-out error-propagation gain and downstream decision sensitivity will reduce predeclared held-out task distortion and post-disturbance recovery time more than periodic full/keyframe refresh, per-stream adaptive refresh, source-PCA allocation, and direct rate-matched allocation, without increasing worst-stream fidelity violations or residual-structure failures. This is a contrastive performance claim, not a world-novelty claim.","confidence":"HIGH"},"implementation_evidence":{"support":"MODERATE","rationale":"Every major computational primitive is feasible with established methods: fitting a local linear operator, eigendecomposition or a more robust Schur/SVD treatment, transform coefficients, fixed-budget bit allocation, deterministic replay, task-based evaluation, and refresh synchronization. Prior work supports mode-level rate allocation and task-aware linear combining, while codec standards demonstrate practical refresh mechanisms. Unverified issues are decisive: operator stationarity and identifiability, non-normal or defective matrices, coefficient signaling and protection overhead, encoder-decoder state synchronization, nonlinear cross-mode excitation, compliance with mandatory stream protection, and integration with a specific codec and channel stack. Offline replay presents low legal and safety risk if data access is authorized and no live transmission changes occur; live use requires mission-specific approval and assurance.","source_ids":["S3","S4","S5","S6","S7"]},"scores":{"meaningful_impact":{"score":3,"rationale":"If the hypothesized persistent joint errors occur, faster recovery at a fixed bit rate could protect decision-relevant telemetry fidelity. Prevalence and realized downstream harm are not externally measured.","source_ids":["S1","S2","S3","S4"]},"stakeholder_pull":{"score":3,"rationale":"NASA/CCSDS actors visibly need bandwidth-efficient, reliable telemetry compression, but no identified actor requests this allocation rule or commits data, staff, or funding.","source_ids":["S1","S2"]},"incremental_advantage":{"score":2,"rationale":"The proposal combines known unstable-mode allocation, task-aware quantization, transform coding, and refresh. Advantage over direct end-to-end rate-distortion optimization is plausible but unsupported until replay.","source_ids":["S4","S5","S6","S7"]},"distinctiveness_plausibility":{"score":2,"rationale":"A narrow codec-error-dynamics implementation remains distinguishable, but the central idea of assigning limited bits to dynamically unstable modes substantially collides with published prior art.","source_ids":["S5","S6","S7"]},"technical_implementability":{"score":4,"rationale":"An offline implementation is technically straightforward given an instrumented deterministic codec and synchronized trace. Robust spectral treatment, overhead, and regime drift make operational deployment materially harder.","source_ids":["S4","S5","S6","S7"]},"adoption_authority_feasibility":{"score":3,"rationale":"CCSDS and NASA/JPL provide credible institutional routes, and codec owners can authorize offline replay. Mission-specific authority for live allocation and safety minima is not evidenced.","source_ids":["S1","S2"]},"evidence_readiness":{"score":3,"rationale":"The next test is well bounded and can use archived data, but the necessary trace, codec implementation, loss model, task metric, and comparator code are proprietary or partner-controlled and were not located publicly.","source_ids":["S1","S2","S5"]},"safety_net_benefit":{"score":2,"rationale":"The method might preserve low-bandwidth or safety-relevant telemetry under disturbance, but no evidence identifies a disadvantaged population or verifies that safety-critical streams benefit rather than lose protection.","source_ids":["S1","S2"]},"scalability":{"score":3,"rationale":"The mathematics and replay workflow transfer across multistream codecs, but every predictor, source mix, channel regime, and downstream task requires refitting, validation, and drift governance.","source_ids":["S4","S5","S6"]}},"score_confidence":"MODERATE","costs":{"first_evidence":{"band_2026_usd":"10K_TO_50K","scope":"One bounded archived-trace replay: instrument the codec, inject predeclared loss and quantization disturbances, fit the transition model on training data, implement four rate-matched comparators, and report held-out metrics.","confidence":"MODERATE","assumptions":["A usable archived trace and deterministic codec replay harness already exist.","Roughly four to eight engineer/researcher weeks are required.","No new hardware, live transmissions, licensing purchase, or security accreditation is required.","The BLS wage benchmark is converted to a 2026 resource-equivalent with ordinary employer overhead; this is not a vendor quote."],"source_ids":["S8"]},"initial_deployment_startup":{"band_2026_usd":"50K_TO_250K","scope":"Prototype shadow-mode integration for one codec and one telemetry configuration, including coefficient signaling, synchronization checks, dashboards, test automation, and design review.","confidence":"LOW","assumptions":["Approximately three to nine person-months across codec, data, and validation engineering.","Existing compute, telemetry storage, channel simulator, and observability infrastructure can be reused.","No flight-certified or safety-critical live actuation is included.","Partner data access and security review do not create major delays."],"source_ids":["S4","S8"]},"operational_launch":{"band_2026_usd":"250K_TO_1M","scope":"Production or mission launch for one operational telemetry stack, including hardware-in-the-loop or equivalent testing, fault injection, safety and coverage verification, rollback, documentation, training, and change approval.","confidence":"LOW","assumptions":["One to three engineer-years plus assurance, operations, and program-management effort.","The existing predictor and transmission budget remain unchanged except for the approved refresh allocation.","No spacecraft hardware redesign or new radio certification is needed.","Mission-assurance burden can vary enough to exceed this band; no public mission-specific estimate was found."],"source_ids":["S1","S2","S4","S8"]},"annual_recurring":{"band_2026_usd":"50K_TO_250K","scope":"Monitoring, scheduled mode re-estimation, regression replay after source/predictor/channel changes, incident review, threshold maintenance, and annual reauthorization for one deployment.","confidence":"LOW","assumptions":["Approximately 0.25 to 1.0 specialist FTE plus shared compute and operations support.","The number of codec configurations and missions is small.","No major redesign is required during the year.","Costs are labor-resource equivalents derived from BLS benchmarks, not observed program expenditures."],"source_ids":["S8"]}},"verified_pipeline_gates":{"externally_supported_problem":{"status":"YES","reason":"Primary research directly documents packet-loss error propagation in predictive DPCM, while a codec standard explains why decoder refresh is required to stop predictive-reference errors. The specific hidden cross-stream mode remains unverified but the broader problem is externally supported.","source_ids":["S3","S4"]},"externally_credible_adopter_or_authorizer":{"status":"YES","reason":"CCSDS explicitly identifies onboard compression and link reliability as needs and names NASA/JPL and ESA participants in the relevant compression and coding activities. This is a credible research and standards adoption channel, although live authority is mission-specific.","source_ids":["S1","S2"]},"distinct_testable_incremental_claim":{"status":"YES","reason":"The candidate can be compared at exactly equal bit rate against periodic keyframes, per-stream adaptive refresh, source-PCA allocation, and direct allocation using predeclared held-out task distortion, recovery, minimum-fidelity, and residual criteria.","source_ids":["S4","S5","S6","S7"]},"bounded_next_evidence_step":{"status":"YES","reason":"A single archived trace, deterministic replay, fixed disturbance set, training/held-out split, fixed budget, four comparators, and explicit falsifiers bound the experiment and prevent live operational exposure.","source_ids":["S3","S4","S5","S6"]},"no_unresolved_safety_or_authority_stop":{"status":"YES","reason":"The proposed next step is offline and reversible, preserves all mandatory coverage, and authorizes at most shadow mode. Live reallocation remains prohibited without codec-owner, network-operator, downstream-owner, and mission-safety approval.","source_ids":["S1","S2","S4"]},"credible_cost_scope_and_range":{"status":"YES","reason":"All four estimates state a bounded scope, labor assumptions, exclusions, confidence, and broad resource-equivalent bands anchored to an official occupational wage benchmark. Operational estimates remain low-confidence because no mission-specific integration quote exists.","source_ids":["S8"]}},"next_evidence_step":"With an authorized archived multistream trace and deterministic version-pinned codec, preregister a disturbance matrix covering bounded packet-loss bursts and quantization perturbations; a downstream distortion or decision-loss metric; mandatory per-stream fidelity floors; coefficient-overhead accounting; and thresholds for operator-fit residuals, mode conditioning, spectral separation, and basis drift. Fit the error-transition operator only on training segments. On held-out segments, compare (A) unchanged baseline allocation, (B) rate-matched periodic full/keyframe refresh, (C) rate-matched per-stream adaptive refresh, (D) source-covariance PCA/KLT coefficient protection, (E) direct end-to-end rate-matched allocation, and (F) the proposed propagation-plus-task-selected modal refresh. Use identical total transmitted bits including coefficient identifiers, quantization, protection, and synchronization overhead. Falsify the problem premise if an independent or memoryless error model matches the coupled model and no stable consequential joint residual remains. Falsify the intervention if F does not improve both predeclared held-out task distortion and recovery time over the strongest comparator, or if it increases fidelity-floor violations, structured residuals, synchronization failures, or conditioning/drift breaches. A positive result authorizes only a shadow-mode hardware-in-the-loop or operational replay, not live control.","blocking_evidence":["No opened source measures the prevalence, persistence, or downstream consequence of hidden cross-stream error modes in an operational predictive telemetry codec.","No authorized archived trace, codec implementation, packetization/channel model, or synchronized encoder-decoder state was available for independent replay.","Coefficient quantization, signaling, error protection, compute latency, and synchronization overhead have not been measured under the fixed bit budget.","No mission-specific downstream loss function or mandatory safety-critical stream coverage table has been supplied.","No live codec owner, network operator, downstream application owner, or mission-safety authority has committed to the experiment or approved live changes.","Head-to-head performance against direct end-to-end rate-distortion allocation and the other preregistered comparators requires proprietary data and testing.","Operational integration and recurring cost estimates lack a mission-specific engineering estimate or vendor/program quote."],"research_disposition":"PARTNERED_RESEARCH_PROGRAM","world_novelty_boundary":"World novelty, patentability, freedom to operate, market size, and realized impact were not measured. The search establishes substantial collision at the level of component ideas—predictive-codec refresh, transform-domain coefficient allocation, task-aware quantization, and unstable-mode bit allocation—but does not establish whether the exact combination of a fitted predictive-codec error-transition operator, downstream-sensitivity ranking, and protected modal reset coefficients has previously been disclosed or deployed.","arm":"COMPLETE_PROPOSAL_PORTFOLIO","candidate_version":0,"controller_recommendation":{"action":"STOP_EMPIRICAL_RESEARCH_NEEDED","repairable":false,"material_progress_observed":true,"progress_targets":["Secure an authorized, version-pinned predictive-telemetry codec, synchronized archived multistream trace, and deterministic replay harness.","Preregister the disturbance set, downstream loss, per-stream safety floors, all bit-overhead accounting, comparator implementations, and statistical decision rule.","Demonstrate that a coupled local error-transition model predicts held-out recovery better than memoryless and stream-independent models and leaves no consequential structured residual.","Run the equal-budget six-arm comparison and falsify the proposal unless modal refresh beats the strongest comparator on both task distortion and recovery without new safety-floor violations.","Measure mode conditioning, non-normal transient growth, spectral-gap stability, basis drift, cross-mode excitation, coefficient protection cost, latency, and encoder-decoder synchronization failures.","Obtain written mission-specific authority for shadow testing and document that mandatory refreshes and safety-critical stream protection cannot be reallocated.","Produce a mission-specific engineering estimate for integration, verification, launch, and annual monitoring costs."],"reason":"Bounded web research established that the broad problem is real, credible institutional adopters exist, and the implementation is mathematically feasible, but it also found substantial prior-art collision. The remaining incremental claim is empirical and cannot be resolved by further public-web search: it requires an authorized proprietary codec and trace, deterministic disturbance replay, equal-budget comparators, and mission-specific safety and authority review. Under the required controller rule, this fieldwork/proprietary-data dependency requires STOP_EMPIRICAL_RESEARCH_NEEDED with repairable set to false."},"proposal_index":1}