A Local Agentic AI Framework to Generate Edge Case Scenarios for Safety Assurance
A laptop-based local agentic framework generates and critiques edge-case scenarios for autonomous-system safety assurance.
Researchers propose a laptop-hosted agentic workflow that uses local language models to generate edge-case scenarios for autonomous systems. System requirements, system descriptions, and environment descriptions ground the prompts, while a second model critiques each scenario against domain criteria. Refined scenarios are simulated and scored, then similar edge cases are exploited or new scenarios explored. Clustering and thematic analysis of the descriptions produce traceable safety-assurance evidence.
- Local think-act-observe loops generate scenarios without cloud models.
- A second independent LM critiques and refines each scenario.
- Simulation metrics drive exploitation of edge cases or new exploration.
- Clustering and thematic analysis support a traceable safety case.
From the original article
1 min read at arxiv.orgGenerating edge case scenarios is crucial for training, testing and safety assuring autonomous systems. This identifies hazardous failures to increase system reliability and robustness, ensures the system can still function safely on the boundaries of its operational design domain and learns the expected system behaviour for these edge cases.
By Victoria J. Hodge · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.