Marker Consensus Calibration

Mechanism

A marker is a passive or semi-passive reference object placed at a known or estimable location. In the disclosed embodiment, each marker is associated with writable non-volatile memory or with a marker-bound consensus surface in the spatial-mesh substrate that accumulates credentialed observations addressed to the marker's region. When an operating unit passes the marker, it observes the marker through its own sensors, computes the marker's apparent position relative to the unit's claimed pose, and emits a credentialed observation that contains the marker identity, the apparent position, the unit's claimed pose, the unit's claimed capability class (and the credential under which the claim is made), and the freshness window in which the observation was valid.

The marker's consensus surface accumulates these contributions. The consensus computation is deterministic against a published consensus policy class that specifies the admissible contribution set, the cross-source quorum required for a refinement step, the weighting of contributions by capability class and freshness, and the drift-detection criteria by which the marker's position estimate is identified as having migrated rather than fluctuated. The output of the consensus is a credentialed observation of the marker's refined position, stamped with the consensus policy under which it was produced, the quorum class that was achieved, and the contributing-credential set summary. Receivers consume the refined-position observation through the same admissibility pipeline they apply to any other credentialed observation.

The same consensus surface produces a second output: a calibration observation about each contributing unit's claimed capability. If a unit's contributions consistently disagree with the consensus by more than the published tolerance for the unit's claimed capability class, the consensus surface emits a credentialed capability-disagreement observation that is consumed by the capability-awareness layer; the unit's capability claim is then either down-classified, suspended, or — under publication-credentialed authority — revoked. This is the bidirectional calibration that distinguishes the disclosed primitive from one-way reference-network publication: the marker calibrates the unit, and the unit's contributions calibrate the marker, and both calibrations are credentialed observations under the same substrate.

Drift detection is itself a credentialed primitive. A marker whose accumulated observations indicate position migration (ground settling, frost heave, vehicle impacts, deliberate displacement, or seismic motion) emits a marker-drift observation under the consensus surface's credential. The drift observation triggers either a marker-resurvey solicitation (which, if integrated with the forecast-uncertainty solicitation primitive, retasks credentialed survey-class units toward the marker) or a marker-deprecation observation that excludes the marker from contributing to capability calibration until its position is re-established. Each transition is observable, and the marker's history is replayable from its observation log.

Cross-source quorum is the structural protection against single-source capture of the consensus. A consensus refinement step requires contributions from a published minimum number of distinct authority classes and a published minimum number of distinct units within each class; a quorum-deficient set produces no refinement and emits a quorum-insufficient observation that is itself consumed by upstream solicitation primitives to direct additional contributions. Rate-limiting is the structural protection against contribution flooding: each contributing credential is bounded in contributions per marker per time window, and aggregate contributions to a single marker are bounded by the marker's consensus-policy parameters. The combination — credentialed admission, cross-source quorum, rate-limit — is what makes the consensus structurally bounded rather than ad-hoc averaging.

Operating Parameters

Four parameter classes govern the consensus. The admission class specifies which credential classes may contribute observations to a marker's consensus surface, the minimum capability class required for admission, and the freshness window in which contributions remain admissible to the consensus computation. The admission class binds the consensus to credentialed contributors and excludes uncredentialed or expired contributions structurally rather than by post-hoc filtering.

The quorum class specifies the cross-source quorum required for a consensus refinement step: minimum distinct authority classes, minimum distinct units per class, and the weighting by capability class. A high-stakes embodiment (precision-survey-grade markers in a transportation corridor) requires multi-class quorum with regulator-credentialed participation; a lower-stakes embodiment (yard-internal markers for fleet-internal logistics) may operate with single-class quorum across many same-class units. Quorum is published, observable, and auditable.

The rate-limit class bounds contribution density. Each contributing credential is permitted a maximum number of contributions per marker per time window, with back-off behavior when contributions are rejected for quorum, freshness, or admissibility reasons. Rate-limiting prevents a malfunctioning or compromised contributor from saturating the consensus surface, and prevents adversarial behavior in which a contributor would attempt to drive the consensus toward a false position. Aggregate rate-limits across all contributors to a single marker bound the consensus computation cost and protect the marker-bound substrate from contribution overload.

The drift-detection class specifies the criteria by which the consensus surface declares a marker's position to have migrated. The criteria include statistical tests against the contribution distribution, comparison against neighboring markers' consensus, comparison against external reference-network observations where available, and explicit operator-credentialed override. Drift detection produces a credentialed marker-drift observation that triggers either resurvey solicitation or marker deprecation under the published policy. The drift-detection class is itself published, so a regulator can audit the detection sensitivity and the false-positive rate without inspecting source code.

Alternative Embodiments

The primitive composes across application domains. In on-road autonomy embodiments, markers may be retroreflective road furniture, lane markings with embedded identifiers, pole-mounted reference objects, or curb-embedded passive devices; contributing units include credentialed fleet vehicles and credentialed survey vehicles. In mining and quarry embodiments, markers are placed at known reference points within the operational area; contributing units include haul trucks, drills, and survey-class platforms operating under operator credentials. In agricultural embodiments, markers are field-perimeter or in-field reference objects; contributing units include tractors, harvesters, and aerial platforms operating under operator and equipment credentials.

In smart-port and smart-yard embodiments, markers are container-corner references, crane-base references, and gate references; contributing units include yard tractors, straddle carriers, and mobile cranes. In airfield and airspace embodiments, markers are runway and taxiway reference objects observed by ground vehicles, and additional credentialed observation classes include surveyor-credentialed periodic resurvey contributions. In defense and expeditionary embodiments, markers are deployed at the start of an operation and calibrated through fleet contributions in regions where centralized reference networks are unavailable; the consensus policy class is published under the operational authority and the quorum class enforces multi-credential composition.

Embodiments may also compose with existing centralized reference networks. Where CORS-style coverage exists, CORS observations enter the consensus surface as one credential class among others, with regulator-credentialed weighting; where CORS coverage is sparse or absent, fleet contributions provide the primary calibration signal under their own credentialed admission. The transition between CORS-anchored and fleet-anchored regimes is itself a credentialed observation, so the substrate's behavior is auditable across the transition. Progressive deployment is a direct consequence: regions add markers and contributing units incrementally, and the consensus precision improves with density without requiring complete coverage to provide value.

Embodiments may differ in marker physical realization (passive retroreflective, semi-passive RFID-bearing, semi-active with writable NV memory, fully virtual as a mesh-resident consensus surface) and in the addressing schema by which markers are bound to spatial-temporal regions. The primitive — credentialed contribution, cross-source quorum, rate-limit, drift detection, bidirectional calibration — is invariant across realizations.

Composition

The primitive composes with the broader cognition-patent architecture. The capability-awareness layer consumes the consensus's capability-disagreement observations to maintain calibrated capability claims across the fleet; units whose claims drift from consensus are down-classified or suspended structurally rather than empirically. The forecasting and uncertainty-solicitation primitives consume marker-position observations as credentialed reference inputs and may emit solicitation observations directed at marker-resurvey when the consensus surface flags drift. The spatial-mesh substrate carries every observation in the loop — contributions, refinements, drift events, calibration outputs — under the same credentialed-observation taxonomy.

The primitive also composes additively with the data-carries-authority inversion. The consensus surface is not a centralized authority; it is a credentialed issuer whose authority is published and whose output is admissibility-evaluated by receivers under the published policy class. Removing a consensus surface for a marker does not remove the marker's authority basis; it removes one issuer, and other contributors may continue to issue credentialed observations of the marker's position under their own credentials. The architecture is structurally robust to single-issuer compromise because no single issuer is privileged.

The primitive composes with regulatory regimes by treating regulator-credentialed survey contributions as a high-weight credential class in the consensus. A regulator may publish periodic survey observations that enter the consensus alongside fleet contributions; the consensus policy class binds the weighting, and the regulator can audit the consensus by inspecting the published parameters and the observation log. This is what makes the primitive admissible in regulated domains where positioning precision must be demonstrably bounded.

Prior-Art Distinction

Centralized RTK reference networks (CORS operated by NGS and equivalents internationally, NTRIP per RTCM, commercial RTK services like Trimble VRS Now, Hexagon SmartNet, Topcon TopNET) maintain precision positioning through dedicated reference stations operated by survey-credentialed authorities. The architecture works for the geographies where it has been deployed; it doesn't extend to deployments where reference-station maintenance is impractical (mining, agriculture in remote areas, expeditionary deployment, deeply rural autonomy), and it does not provide the bidirectional capability calibration the disclosed primitive provides. The disclosed primitive incorporates CORS-style observations as one credential class where available; it does not require them.

Crowdsourced-mapping systems (Mobileye REM, TomTom, HERE, Wejo) accumulate fleet observations into centralized map services and republish refined map products. The map service is the authority; the contributing fleet is anonymous or pseudonymous; the calibration is one-way (fleet contributes, service publishes). The disclosed primitive is bidirectional (the consensus calibrates both the marker and the contributors), credentialed (contributors are not anonymous; they participate under their published authority class), and structurally bounded by cross-source quorum rather than by the service's internal heuristics. The substrate primitives — addressable region, credential class, freshness, policy — are present in the disclosed substrate and absent from crowdsourced-mapping architectures.

Distributed-consensus literature (Paxos, Raft, Byzantine fault tolerance, blockchain consensus per Bitcoin and successors) provides protocols for agreement among distributed participants. The disclosed primitive incorporates consensus mechanisms by reference where applicable but addresses a different problem: spatial-temporal-region-scoped consensus across credentialed observation contributors with cross-source quorum and rate-limit, where the artifact under consensus is a physical reference position and the contributors' authority basis is itself an output of the consensus. Generic distributed-consensus primitives do not address spatial-temporal scoping or bidirectional capability calibration.

SLAM and pose-graph optimization (Davison 2003, Kümmerle et al., Cartographer, ORB-SLAM) provide mathematical apparatus for jointly estimating pose and landmark position from observation tracks. The disclosed primitive incorporates pose-graph apparatus by reference inside the consensus computation; the contribution of the disclosure is not the pose-graph mathematics but the architectural layer at which the apparatus is exposed as a credentialed cross-authority bidirectional calibration loop with structural quorum and rate-limit governance. SLAM literature does not address authority class, credentialed admission, cross-source quorum, or drift-detection-as-credentialed-observation.

Disclosure Scope

The disclosure covers the credentialed contribution observation, the marker-bound consensus surface, the admission/quorum/rate-limit/drift-detection parameter classes, the bidirectional calibration outputs (refined-position observation and capability-disagreement observation), and the drift-event and marker-deprecation observations. It covers embodiments across on-road autonomy, mining, agriculture, smart-port, smart-yard, airfield, defense, and expeditionary deployments. It covers compositions with existing centralized reference networks under non-privileged-issuer integration, and it covers progressive-density deployment where consensus precision improves incrementally with marker density and contributor density.

The disclosure covers marker physical realizations from passive retroreflective objects through semi-active writable-NV-memory devices to fully virtual mesh-resident consensus surfaces. It covers regulator-credentialed survey contributions as a high-weight credential class where regulatory regimes require it. It explicitly disclaims reliance on a single authority class for consensus; the quorum class is taxonomic and supports multi-authority composition. It explicitly disclaims reliance on a centralized reference network; the primitive operates standalone where centralized references are absent, and composes additively where they exist.

The disclosure is positioned at the architectural layer where capability calibration becomes a structurally credentialed property of fleet operation rather than a centralized survey-network expense. The primitive replaces neither the mathematical apparatus of pose-graph estimation nor the regulatory function of survey authorities; it provides the architectural layer at which both compose under bounded, credentialed, cross-source consensus governed by published parameters.