Runtime Trustworthiness Under Uncertainty: A Fuzzy Cause Tree and Subjective Logic Framework for Autonomous Systems
Year: 2026
Type: conference-paper
Abstract: Autonomous systems operating in open environments face gradual, context-dependent degradation that threshold-based monitors cannot adequately characterize. Building on our prior work on dynamic safety assessment, this paper presents a three-layer runtime trustworthiness assurance framework: (1) a Fuzzy Cause Tree (FCT) for interpretable degradation modeling across navigation, task execution, and supervision/fallback pillars; (2) a Subjective Logic (SL) layer that quantifies belief, disbelief, and uncertainty with a multiplicative evidence quality metric Q = C • F • K over evidence coverage C, freshness F , and consistency K; and (3) a data-driven decision policy mapping fused opinions to continue, fallback, or supervised responses, with boundaries derived via coverageguided fuzzing. We use an autonomous outdoor cleaning robot as a running example to instantiate the framework, though the approach applies to autonomous systems more broadly.
Source: HAL
Language: English
Subfield: Safety, Risk, Reliability and Quality
Field: Engineering
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7220757359