Mechanism
The cascade-propagation primitive operates on a governance-credentialed topology graph whose nodes represent elements of a physical-world domain and whose edges represent propagation channels between those nodes. The graph is maintained by one or more governance authorities with domain responsibility. A per-edge propagation function defines how a disruption at a source node projects to connected nodes with governance-policy-defined transit, attenuation, transformation, or amplification characteristics, and a per-node aggregation function defines how multiple incoming propagation contributions combine at a receiving node.
The topology-learning and adaptive-refinement mechanism is the means by which this graph evolves beyond its initial declared state. It updates the topology graph, the propagation functions, and the aggregation functions from observed propagation outcomes. The mechanism draws on the training-governance primitive of the disclosure, applying experience-based refinement to the cascade model so that the topology becomes a more accurate predictor of future cascades over the operational lifetime of the deployment.
Refinements produced by the mechanism are themselves governance-credentialed rather than applied as unattested edits. Each topology update is recorded in the governance-chain lineage field alongside the topology references, propagation computations, directive emissions, mitigations, halting events, and refusals that the cascade-propagation primitive records. This preserves a reconstructable account of how the graph reached its current state.
Refusal outcomes contribute to the topology-learning mechanism. When a downstream receiving agent refuses a proposed mitigation, the refusal is emitted as a first-class governed observation; upstream coordinators may adapt their cascade-response strategy in subsequent propagation computations, and the refusal outcomes feed back into the refinement of the topology graph and its propagation and aggregation functions.
Governance Properties
The behavior of the topology-learning mechanism is governed rather than free-running. The topology graph is maintained by one or more governance authorities with domain responsibility, and updates to the graph, the propagation functions, and the aggregation functions enter through the same governance-credentialed custody that applies to the rest of the cascade-propagation primitive.
Because each topology update is recorded in the governance-chain lineage, the provenance of every refinement remains reconstructable: an auditor can trace how a given edge, propagation characteristic, or aggregation rule entered its current form, and which observed propagation outcomes contributed to it. The disclosure does not recite specific numeric thresholds, observation intervals, or admission rates for this mechanism; these are governance-policy-defined parameters of a given deployment rather than fixed properties of the architecture.
The mechanism applies experience-based refinement without requiring an explicit training-mode transition. Refinement proceeds continuously from observed propagation outcomes, including refusal outcomes contributed by downstream agents, so the cascade model is progressively reconciled with operational reality under governance oversight.
Composition With Cross-Domain Cascade
The cascade-propagation primitive supports a cross-domain cascade composition mechanism that combines cascade propagation across two or more topology domains, producing composite cascade determinations. Topology learning operates within this setting: refinements observed in one domain can sharpen the propagation and aggregation functions used when cascades cross domain boundaries, so that cascade-of-cascade determinations rest on a topology that has been reconciled with operational outcomes.
Where a topology spans multiple governance authorities, the primitive provides a cascade-authority resolution mechanism that resolves responsibility across the governing authorities of the affected segments. Topology updates produced by the learning mechanism remain bound to the authority with domain responsibility for the affected nodes and edges, so that no refinement displaces the custody structure of the graph.
The learning mechanism also draws on the training-governance primitive of the disclosure, which extracts refinements from governed execution experience. Applied to the cascade model, this produces a closed-loop refinement in which observed propagation outcomes, including refusals, feed back into the topology graph and its propagation and aggregation functions.
Composition With Other Primitives
Topology learning composes directly with the governance-credentialed topology graph: it is the principal mechanism by which that graph, together with its propagation and aggregation functions, evolves beyond its initial declared state. It composes with the governance-chain lineage, since each topology update is recorded as a lineage entry, and with the training-governance primitive, from which it draws experience-based refinement of the cascade model.
Learning composes with cascade-propagation evaluation by feeding observed propagation outcomes back into the topology, producing a closed-loop refinement in which the topology graph becomes a more accurate predictor of future cascades over the operational lifetime of the deployment. It composes with the refusal and upstream-coordination mechanism, whose refusal outcomes contribute to refinement. Through the lineage record it supports post-hoc reconstruction of why any given edge, propagation characteristic, or aggregation rule reached its current form.
Distinction From Prior Art
The cascade-propagation primitive is structurally distinguished from prior cascade-modeling architectures, including power-grid SCADA cascade-analysis, traffic simulation, epidemic modeling, supply-chain disruption modeling, and structural-failure modeling. Prior architectures operate on centrally-maintained models with ad hoc trust assumptions, whereas the present primitive operates on governance-credentialed topologies with authority-chained custody, and its topology-learning mechanism updates that credentialed graph rather than an untrusted central model.
Prior architectures produce unstructured alerts or central dashboards, whereas the present primitive produces governance-chain-preserving cascade-propagation observations, and the refinements it learns are themselves recorded in the governance-chain lineage. Prior architectures are also narrowly scoped to a single cascade domain and do not support cross-domain cascade composition, whereas the present primitive operates across domains through a shared architectural mechanism and composes cascades across topology domains.
The disclosure states that the present primitive supports governance-chain-preserving topology learning, providing adaptive topology refinement where prior cascade-modeling architectures do not. The distinguishing property is that topology updates are governance-credentialed and lineage-bound, rather than applied as unattested edits to a central model.
Disclosure Scope
This disclosure, U.S. Provisional Application No. 64/049,409, encompasses any governed spatial mesh in which a cascade-propagation primitive includes a topology-learning and adaptive-refinement mechanism that updates a governance-credentialed topology graph, its per-edge propagation functions, and its per-node aggregation functions from observed propagation outcomes, with each topology update recorded in the governance-chain lineage. The cascade-propagation primitive operates across power, transportation, fluid, thermal, structural, biological, communication, logistics, economic, and cyber-physical topologies, and admits extension to any future topology class through governance-policy-defined topology registration without architectural modification. Specific topology classes and refinement settings are recited as illustrative embodiments and do not limit the claims.
