SOTIF‐Oriented Perception Uncertainty Analysis and Risk‐Aware Sensor Layout Optimization for Autonomous Navigation System

ABSTRACT Maritime autonomous surface ships (MASS) rely on multi‐sensory perception systems to detect navigation obstacles, but design and performance limitations may cause safety of the intended functionality (SOTIF) issues even when no component failure occurs. This study proposes a SOTIF‐oriented framework integrating perception uncertainty analysis, collision‐risk assessment, and sensor layout optimization. First, a scenario‐function‐architecture framework is established to identify perception‐related SOTIF issues and formulate quantitative requirements. Second, sensor detection probabilities and state‐estimation uncertainties are modeled and evaluated at the multi‐sensor architecture level. Third, an ellipse‐based deformable safety envelope provides a spatial reference for multi‐ship collision risk assessment, with fused range, bearing, and relative‐speed uncertainties incorporated into the risk indicator. Sensor positions and orientations are optimized to maximize coverage within a fixed reference region under detection performance and measurement error constraints for daytime and nighttime conditions. A synthetic encounter scenario involving ten ships illustrates the envelope evolution and the prescribed response of the risk indicator to perception uncertainty. Across five independent optimization runs, the mean combined coverage was 68.49%, with a sample standard deviation of 0.02% points, compared with 67.89% for a fixed heuristic layout. The selected configuration achieved daytime and nighttime coverage of 74.45% and 62.56%, respectively. These results illustrate the application of the framework to sensor‐configuration evaluation under the adopted SOTIF‐oriented perception constraints.