P44 — AIEP — Deterministic Convergence and Drift Detection Engine (ts)

Applicant: Neil Grassby
Inventor: Neil Grassby
Classification: withheld — internal
Priority: Claims priority from GB2519711.2 filed 20 November 2025

Abstract

A deterministic execution drift detection and convergence enforcement system for distributed arbitration frameworks operating under an Architected Instruction & Evidence Protocol (AIEP). Canonical state artefacts are generated and hashed during execution. A schema-versioned exclusion map defines permitted differences. Deterministic structural diff classification identifies non-permitted drift, including structural divergence, schema violations, lineage inconsistency, and hash mismatch. Execution enablement is suppressed in a fail-closed manner when non-permitted drift is detected. The invention improves distributed reproducibility, convergence reliability, and execution certification capability.

Field of the Invention

[0001] The present invention relates to deterministic execution verification in distributed computing systems.

[0002] More particularly, the invention relates to a deterministic execution drift detection and convergence enforcement system for distributed arbitration frameworks operating under an Architected Instruction & Evidence Protocol (AIEP) system as defined in United Kingdom patent application GB2519711.2.

Background

[0003] Distributed deterministic systems rely upon invariant-bound state transitions to ensure reproducibility across nodes.

[0004] In practice, execution environments may diverge due to differences in configuration, dependency resolution, schema interpretation, or state mutation ordering.

[0005] Even where deterministic hashing mechanisms are employed, partial divergence may occur at intermediate execution stages without immediate detection.

[0006] Conventional systems detect divergence only after final output comparison, without identifying structural drift within execution pathways.

[0007] Known approaches to version pinning or hash comparison do not provide schema-governed exclusion maps combined with deterministic drift classification and fail-closed convergence enforcement.

[0008] There exists a need for a deterministic drift detection system capable of identifying execution divergence at structural, schema, and artefact levels, and enforcing convergence prior to execution enablement.

Summary of the Invention

[0009] The invention provides a computer-implemented method for deterministic execution drift detection and convergence enforcement.

[0010] The method comprises generating canonical state artefacts during execution cycles.

[0011] Each artefact is bound to a deterministic artefact hash.

[0012] A version-pinned exclusion map defines permitted non-material differences.

[0013] Structural comparison is performed between expected and observed artefacts using deterministic diff classification.

[0014] Divergence is classified into one or more categories comprising:

(a) permitted exclusion variance;

(b) structural drift;

(c) schema violation;

(d) lineage inconsistency;

(e) hash mismatch.

[0015] Where divergence exceeds permitted exclusion variance, execution enablement is suppressed in a fail-closed manner.

[0016] Convergence enforcement requires reconciliation to canonical state prior to reactivation of execution.

[0017] The technical effect is deterministic identification and containment of execution drift within distributed computing systems, improving reproducibility and certification reliability.

Brief Description of the Drawing

FIG. 1 — Temporal Evidence Gap Detection

   t0        t1        t2       [GAP]      t3        t4
    │         │         │                   │         │
    ●─────────●─────────●───────────────────●─────────●
    │         │         │                   │         │
   Ev(A)    Ev(B)     Ev(C)              Ev(D)     Ev(E)
                          │
              ┌───────────▼─────────────┐
              │    NegativeProofRecord  │
              │    period: [t2+1, t3-1] │
              │    hash: H(gap_spec)    │
              │    "absence is proven   │
              │     not merely noted"   │
              └─────────────────────────┘

Definitions

[0018] Drift: A deviation between expected canonical artefact state and observed artefact state during execution.

[0019] ExclusionMap: A schema-versioned definition of permitted non-material state differences.

[0020] DriftClassification: A deterministic categorisation of identified divergence.

[0021] ConvergenceState: A state in which observed artefacts match canonical artefacts after exclusion filtering.

[0022] DriftCertificate: A deterministic record documenting detected divergence and classification.

Brief Description of the Drawings

Figure 1 illustrates artefact hash generation during execution.
Figure 2 illustrates deterministic structural diff classification.
Figure 3 illustrates exclusion map filtering.
Figure 4 illustrates fail-closed convergence enforcement.

Detailed Description of Preferred Embodiments

Canonical Artefact Generation

[0023] During each execution cycle, state artefacts are serialised in canonical form.

[0024] Canonical serialisation enforces deterministic key ordering and structural representation.

[0025] An artefact hash is computed over the canonical serialisation.

Exclusion Map Governance

[0026] An ExclusionMap is bound to a specific schema version.

[0027] The ExclusionMap specifies artefact fields or structural paths that are permitted to vary without constituting drift.

[0028] Exclusion rules are deterministically evaluated prior to drift classification.

[0029] Exclusion rules are version-pinned and incorporated into hash-bound governance records.

Drift Detection

[0030] Observed artefacts are compared against canonical artefacts using deterministic pointer-based structural diff classification.

[0031] Differences are enumerated in canonical order.

[0032] Each difference is classified according to DriftClassification categories.

[0033] Differences falling within ExclusionMap definitions are marked as permitted.

[0034] Differences outside ExclusionMap definitions are classified as structural drift.

[0035] Schema violations are identified through schema validation failure.

[0036] Lineage inconsistencies are identified through ParentHash verification failure.

[0037] Hash mismatch is identified where canonical artefact hash differs from expected value.

Convergence Enforcement

[0038] Where structural drift, schema violation, lineage inconsistency, or hash mismatch is detected, execution enablement is suppressed.

[0039] A DriftCertificate is generated documenting classification and artefact identifiers.

[0040] Execution may resume only after reconciliation to ConvergenceState.

[0041] ConvergenceState is achieved when artefacts, after exclusion filtering, match canonical state representation.

Distributed Operation

[0042] Nodes operating over identical artefact sets, schema versions, and exclusion maps produce identical drift classifications.

[0043] Drift detection results are reproducible across distributed nodes.

[0044] Divergence between nodes is detectable through comparison of DriftCertificates.

Fail-Closed Behaviour

[0045] ExecutionEnablementSignal generation requires absence of non-permitted drift.

[0046] Any unclassified or unresolvable divergence results in fail-closed suppression.

[0047] No execution profile permits bypass of drift enforcement.

Technical Effect

[0048] The invention enables deterministic structural drift detection within distributed computing systems.

[0049] The invention prevents silent execution divergence.

[0050] The invention enforces convergence prior to execution continuation.

[0051] The invention improves reproducibility, cross-node consistency, and certification reliability.

[0052] The invention provides structured classification of divergence rather than binary output comparison.

Claims

1. A computer-implemented method for deterministic execution drift detection within a distributed arbitration system operating under an Architected Instruction & Evidence Protocol (AIEP), the method comprising:

   (a) generating canonical serialised state artefacts;

   (b) computing artefact hashes over the canonical serialisations;

   (c) applying a schema-versioned ExclusionMap defining permitted differences;

   (d) performing deterministic structural diff classification between expected and observed artefacts;

   (e) classifying detected differences into permitted variance or non-permitted drift; and

   (f) suppressing execution enablement in a fail-closed manner when non-permitted drift is detected.

2. The method of claim 1 wherein structural diff classification is pointer-based and evaluated in canonical order.

3. The method of claim 1 further comprising generating a DriftCertificate documenting divergence classification.

4. The method of claim 1 wherein convergence is required prior to resumption of execution.

5. The method of claim 1 wherein lineage inconsistency is detected through ParentHash verification.

6. A distributed computing system configured to perform the method of any preceding claim.

7. A non-transitory computer-readable medium storing instructions which, when executed, perform the method of any of claims 1–5.