A Subjective Logic-based method for runtime confide... | AI Research

Key Takeaways

A Subjective Logic-based method for runtime confidence updates in safety arguments This paper introduces a new approach to dynamic quantitative assurance, de...
We present a method for dynamic quantitative assurance that enhances static safety cases with continuous, runtime-driven confidence updates.
At runtime, SPI evidence is continuously evaluated, and targeted claims are updated using a rule that increases confidence in the absence of violations and imposes prompt penalties when violations occur.
This design prioritizes safety-relevant responsiveness over exact classical Bayesian posterior updates.
# A Subjective Logic-based method for runtime confidence updates in safety arguments

Paper AbstractExpand

We present a method for dynamic quantitative assurance that enhances static safety cases with continuous, runtime-driven confidence updates. The method quantifies and propagates confidence across the development lifecycle by integrating design-time evidence and windowed runtime Safety Performance Indicators (SPIs) within a single Subjective Logic (SL)-based assurance case. At runtime, SPI evidence is continuously evaluated, and targeted claims are updated using a rule that increases confidence in the absence of violations and imposes prompt penalties when violations occur. This design prioritizes safety-relevant responsiveness over exact classical Bayesian posterior updates. We demonstrate the method using a simulation-based construction zone assist function, focusing on an ML-based construction cone detection component, and show how confidence evolves as SPI evidence is observed in operation.

A Subjective Logic-based method for runtime confidence updates in safety arguments

This paper introduces a new approach to dynamic quantitative assurance, designed to bridge the gap between static safety documentation and real-world performance. Traditionally, safety cases are static documents created during the design phase. This research proposes a method to continuously update these safety arguments by integrating design-time evidence with real-time performance data, ensuring that confidence in a system’s safety evolves as it operates.

Integrating Runtime Evidence

The researchers utilize Subjective Logic (SL) to create a unified framework that combines initial design-time evidence with ongoing operational data. By using windowed Safety Performance Indicators (SPIs), the system can continuously evaluate how well a component is performing against its safety requirements. This allows the assurance case to remain relevant long after the development phase has concluded, providing a more accurate picture of safety in dynamic environments.

Prioritizing Responsiveness

A key feature of this method is its specific approach to updating confidence levels. Rather than relying on traditional, complex Bayesian posterior updates, the authors implemented a rule-based system that prioritizes responsiveness. When the system operates without violations, confidence in the safety claim gradually increases. Conversely, if a violation is detected, the system imposes an immediate penalty on the confidence score. This design ensures that the safety argument reacts quickly to potential hazards, which is critical for maintaining safety-relevant responsiveness.

Demonstrating Safety in Practice

To validate the method, the authors applied it to a simulation-based construction zone assist function. The study specifically focused on an ML-based component responsible for detecting construction cones. By observing the system in operation, the researchers demonstrated how the confidence in the cone detection component fluctuates in real-time based on the SPI evidence gathered. This practical application highlights how the method can effectively track and propagate confidence levels throughout the lifecycle of an autonomous or semi-autonomous system.