Executor-Side Progressive Risk-Gated Actuation for Agentic AI in Wireless Supervisory Control

2026-05-04

AI Summary

This work addresses the challenge in wireless supervisory control where existing AI agents lack explicit semantic mechanisms to assess risk factors such as stale telemetry, concurrent policies, and deadline violations, hindering safe and efficient decision-making. To overcome this, the authors propose the PRGA execution-side contract, which structures agent intent into a triplet (C0/C1/C2) and employs a two-stage deterministic policy to evaluate risk. A progressive risk-gating mechanism is introduced, retrieving C1 evidence only when resources permit, thereby decoupling risk judgment from evidence acquisition. This approach simultaneously ensures safety margins and significantly improves efficiency: on 3GPP energy-efficiency and network slicing SLA benchmarks, it achieves a 23.3–27.4% reduction in safe-action latency, cuts control-plane overhead by 52.7–54.2%, rejects 100% of out-of-bound stale inputs, and matches static-threshold baselines in performance with no more than a 0.5-percentage-point degradation.

Abstract

Agentic artificial intelligence (AI) shows promise for automating O-RAN wireless supervisory control, but translated intents still require an executor-side decision before live network actuation. Existing control flows lack explicit semantics for whether an intent should commit, gate for evidence, or reject under stale telemetry, concurrent policies, deadline and bandwidth limits, and rollback constraints. We propose Progressive Risk-Gated Actuation (PRGA), an executor-side contract for risk-gated wireless intent execution. PRGA structures each intent into executable local triage (C0), on-demand coordination evidence (C1), and post-hoc provenance support (C2), with C2 kept off the online safety path. A deterministic two-stage policy checks expiry, freshness, rollback-handle validity, local conflict, blocking preconditions, and planner-executor risk divergence from C0, then retrieves C1 only for gated intents when deadline and bandwidth budgets allow; evidence-mandatory gates reject when required C1 is unavailable. On two 3GPP-parameterized energy-saving and slice-SLA benchmarks, PRGA reduces time-to-first-safe-action by 23.3-27.4% and per-commit control-plane bytes by 52.7-54.2% against a decision-identical eager full-evidence cost-overlay comparator, thereby isolating retrieval-cost accounting; remains non-inferior within a pre-declared 0.5 percentage-point unsafe-action margin against an invariant-respecting static-threshold comparator; and rejects 100% of injected over-threshold stale inputs in the stale-state fault campaign. On these benchmarks, PRGA improves supervisory responsiveness and control-plane efficiency within the evaluated unsafe-action boundary.

Problem

• Agentic AI
• Wireless Supervisory Control
• Risk-Gated Actuation
• Intent Execution
• Executor-Side Decision

Innovation

• Progressive Risk-Gated Actuation
• Agentic AI
• Wireless Supervisory Control
• Executor-Side Decision
• Intent Execution