An Approach to Validation of Combined Natural Language and Formal Requirements for Control Systems

Abstract

The paper presents a novel approach to validation of behavioral requirements for control systems. A requirement is specified by a natural language pattern and its expression in Linear Temporal Logic (LTL). This way flexibility and understandability of natural language is combined with advantages of formalization that is a basis for various stages of system development, testing and verification. Still, validity of the requirements remains a major challenge. The paper considers application of mutation analysis for capturing of correct behavioral requirements. Generation and exploration of mutants supports a better understanding of requirements, The novelty of the approach is that the suggested mutations are semantic-based, as opposed to the more common syntax-based mutation analysis. A significant advantage of the approach is that it allows to focus only on plausible potential faults in understanding of the required system behavior, and to avoid generation of a vast amount of mutants that are irrelevant to the intended meaning of the requirements. Moreover, in many cases the effect of semantic-based mutations just can not be achieved by usual syntax-based mutations of LTL formulas associated with requirements. The approach is illustrated using a rail cross control example.