Code and Data for Higher-Order Relational Assurance for Modular Autonomous System Architectures

This repository contains the code and numerical data supporting the manuscript “Higher-Order Relational Assurance for Modular Autonomous System Architectures” by Juan J. Segura. The study investigates how much relational information modular system interfaces must preserve in order to reconstruct an integrated dynamic-admissibility relation. The reproducibility package contains Python source code, benchmark configuration files, and CSV datasets and outputs used for the principal computational analyses. The package supports two complementary model layers: (1) a reduced continuous-control model used for exhaustive higher-order project-join analysis and robustness studies, and (2) a synthetic seven-state road-relative vehicle model used as a nonlinear dynamic falsification layer. The principal reproducible results include: complete pairwise reconstruction providing no improvement over unary reconstruction; persistence of this pairwise insufficiency across 27 combinations of numerical resolution, superelevation, and stopping horizon; and a sparse ternary architecture comprising the relations {R, mu, v}, {G, mu, v}, and {mu, v, tau} reproducing the complete third-order reconstruction in all 27 robustness cases. The vehicle parameters are synthetic and are not presented as OEM-calibrated haul-truck data. All admissibility projections are exact only with respect to the computed model relation. The reported discordance measures should not be interpreted as accident probabilities or empirical risk frequencies.