Improving Predictability, Efficiency and Trust of Model-Based Proof Activity

Abstract

We report on our industrial experience in using formal methods for the analysis of safety-critical systems developed in a model-based design framework. We first highlight the formal proof workflow devised for the verification and validation of embedded systems developed in Matlab/Simulink. In particular, we show that there is a need to: determine the compatibility of the model to be analysed with the proof engine, establish whether the model facilitates proof convergence or when optimisation is required, and avoid over-specification when specifying the hypotheses constraining the inputs of the model during analysis. We also stress on the importance of having a certain harness over the proof activity and present a set of tools we developed to achieve this purpose. Finally, we give a list of best practices, methods and any necessary tools aiming at guaranteeing the validity of the verification results obtained.