Modelling Hybrid Programs with Event-B

Abstract

Hybrid systems are one of the most common mathematical models for Cyber-Physical Systems (CPSs). They combine discrete dynamics represented by state machines or finite automata with continuous behaviors represented by differential equations. The measurement of continuous behaviors is performed by sensors. When these sensors have a continuous access to these measurements, we call such model an Event-Triggered model. The properties of this model are easier to prove, while its implementation is difficult in practice. Therefore, it is preferable to introduce a more realistic model, called Time-Triggered model, where the sensors take periodic measurements. Contrary to Event-Triggered models, Time-Triggered models are much easier to implement, but much more difficult to verify. Based on the differential refinement logic (dRmathcal {L}), a dynamic logic for refinement relations on hybrid systems, it is possible to prove that a Time-Triggered model refines an Event-Triggered model. The major limitation of such logic is that it is not supported by any prover. In this paper, we propose a correct-by-construction approach that implements the reasoning on hybrid programs particularly the reasoning of dRmathcal {L} in Event-B to take advantage of its associated tools.