Dependability Evaluation of Industrial Networks by Using Monte Carlo With Importance Sampling

Mechanisms for detecting communication errors are crucial in industrial networks where reliability is a primary requirement. Even if the Cyclic Redundancy Check (CRC) polynomial generator is given for each transmission protocol, an application designer could choose a specific protocol according to its robustness, or additional error detection mechanisms can be freely added at the application level (e.g., nested CRC). Therefore, a methodology is desired to evaluate the residual error probability of a given protocol, i.e., considering its packet structure and error detection mechanism. Symbolic approaches proposed in the literature are not scalable for usual packet sizes. Monte Carlo simulation can be a valid alternative to inject errors and see what happens. However, the brute force generation of error patterns takes too much simulation time if the channel error probability is very low, as in realistic scenarios. This paper presents a Monte Carlo approach enriched with Importance Sampling, implemented in a tool11https://github.com/guarinopaolo/residual-error-probability-simulator. The framework takes the packet structure and the error detection mechanism as input, thus being independent of them. Experimental results validate the approach with respect to state-of-the-art approaches and show its effectiveness in exploring protocol alternatives.