Selective Agentic Recovery for UAV Autonomy with a Persistent Mission Runtime

2026-06-12

AI Summary

This work addresses the challenge of unmanned aerial vehicles (UAVs) failing in complex or ambiguous missions due to path blockages or stalled progress, where frequent reliance on remote agents for recovery—though effective—introduces latency, resource overhead, and security risks. To mitigate these issues, the authors propose the Persistent Mission Runtime (PMR) framework, which ensures safety-critical operations through local closed-loop execution and incorporates a learned Cognitive Value of Invocation (learned-CVI) mechanism as a lightweight gating policy to intelligently determine when invoking a remote agent yields maximal benefit. Integrating a predefined recovery skill library, a safety filtering module, and a call-decision model, PMR enables on-demand, selective recovery. Evaluated across 400 Gazebo/PX4 benchmark trials, the approach boosts mission success rates in difficult scenarios from 5.0% to 95.0%, substantially outperforming baselines while reducing remote invocations by 16.7% and log token usage by 29.2%.

Abstract

Agentic AI can support unmanned aerial vehicle (UAV) autonomy by providing high-level recovery reasoning when local waypoint- or setpoint-based execution encounters blocked passages, repeated no-progress behavior, or mission-level ambiguity. On physical UAVs, however, remote reasoning is most useful when it is invoked selectively, since each call introduces latency, resource cost, backend uncertainty, and a need to validate the returned decision. This paper presents Persistent Mission Runtime (PMR), a UAV recovery framework that keeps the mission loop and safety-critical execution local while using an external agentic reasoner only as an on-demand recovery module. The reasoner selects from predefined recovery skills, and each returned decision is parsed, verified, safety-filtered, and mapped to local executor actions before it can affect flight. PMR introduces learned Cognitive Value of Invocation (learned-CVI), a compact admission gate that estimates when remote agentic reasoning is likely to improve near-term mission progress enough to justify its operational cost. Across a fixed 400-run Gazebo/PX4 benchmark with eight scenarios, learned-CVI raises hard/ambiguous-regime success from 5.0% under local-only autonomy to 95.0%, outperforms one-shot and periodic reasoning baselines by 20.0 and 32.5 percentage points, and reduces remote-agent calls by 16.7% and logged tokens by 29.2% relative to a manually tuned rule-based invocation baseline.

Problem

• UAV autonomy
• agentic recovery
• selective invocation
• mission runtime
• reasoning latency

Innovation

• Selective Agentic Recovery
• Persistent Mission Runtime
• learned-CVI
• UAV Autonomy
• On-demand Reasoning