Automatic Dimensional Analysis of Cyber-Physical Systems

The first step in building a cyber-physical system is the construction of a faithful model that captures the relevant behaviors. Dimensional consistency provides the first check on the correctness of such models and the physical quantities represented in it. Though manual analysis of dimensions is used in physical sciences to find errors in formulas, this approach does not scale to complex cyber-physical systems with many interacting components. We present DimSim, a tool to automatically check the dimensional consistency of a cyber-physical system modeled in Simulink. DimSim generates a set of constraints from the Simulink model for each subsystem in a modular way, and solves them using the Gauss-Jordan elimination method. The tool depends on userprovided dimension annotations, and it can detect both inconsistency and underspecification in the given dimensional constraints. In case of a dimensional inconsistency, DimSim can provide a minimal set of constraints that captures the cause of the inconsistency. We have applied DimSim to numerous examples from different embedded system domains. Experimental results show that the dimensional analysis in DimSim is scalable and is capable of uncovering critical errors in models of cyberphysical systems.