Online adaptation of disturbance taxonomies and weight templates

Develop online updates for the disturbance taxonomy and Runtime Criticality Score weight templates so that the framework can adapt after deployment without changing its coordination interface.

Background

The Runtime Criticality Score uses a configurable but deployment-time-fixed taxonomy of disturbances, such as network congestion, thermal throttling, and memory pressure, together with class-specific weight templates estimated during calibration. The paper identifies the absence of post-deployment updating as an unresolved limitation and indicates that online updates could extend the framework’s adaptation capability without changing its interface.

References

Two limitations remain open. The disturbance taxonomy and weight templates are fixed at deployment; online updates would extend adaptation without changing the interface.

— Infrastructure-Aware Runtime Protection of Co-Deployed AI Services at the Industrial Edge  (2610.18313 - Sathyavageeswaran et al., 13 Oct 2026) in Discussion and Conclusion