Governed Capability Evolution: Lifecycle-Time Compatibility Checking and Rollback for AI-Component-Based Systems, with a Proof-of-Concept Evaluation on Embodied Agents
Systems built from versioned AI components need lifecycle-time governance over how a new module version is admitted, monitored, and withdrawn. Established deployment patterns (canary, blue-green, feature flags, MLOps) monitor aggregate service-level signals of stateless services, not the stateful, policy-constrained runtimes that drive AI components. We formulate governed capability evolution as a software-lifecycle problem and ask whether staged upgrade governance prevents unsafe activations where practice-adapted canary and blue-green deployment do not, and at what cost. We propose four compatibility checks (interface, policy, behavioral, recovery) and a seven-stage pipeline (validation...
| Authors | Xue Qin, Simin Luan, John See et al. |
|---|---|
| Primary category | cs.RO |
| Published | 9 April 2026 |
| Last updated at source | 29 September 2026 |
Silytics shows public metadata from arXiv and links to the original. Information may be incomplete or out of date. Last refreshed 17 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.