Stochastic Contracts for Runtime Checking of Component-based Real-time Systems
This addresses reliability issues for developers using off-the-shelf components in real-time systems, though it is incremental as it builds on existing contract-based methods.
The paper tackles the problem of verifying real-time properties in component-based systems by extending Design by Contract with statistical inference to estimate execution times, showing in experiments that it integrates smoothly with acceptable overhead (less than 10%).
This paper introduces a new technique for dynamic verification of component-based real-time systems based on statistical inference. Verifying such systems requires checking two types of properties: functional and real-time. For functional properties, a standard approach for ensuring correctness is Design by Contract: annotating programs with executable pre- and postconditions. We extend contracts for specifying real-time properties. In the industry, components are often bought from vendors and meant to be used off-the-shelf which makes it very difficult to determine their execution times and express related properties. We present a solution to this problem by using statistical inference for estimating the properties. The contract framework allows application developers to express contracts like "the execution time of component $X$ lies within $γ$ standard deviations from the mean execution time". Experiments based on industrial case studies show that this framework can be smoothly integrated into existing control applications, thereby increasing their reliability while having an acceptable execution time overhead (less than 10%).