Contracts-Based Control Integration into Software Systems

Abstract

Among the different techniques that are used to design selfadaptive software systems, control theory allows one to design an adaptation policy whose properties, such as stability and accuracy, can be formally guaranteed under certain assumptions. However, in the case of software systems, the integration of these controllers to build complete feedback control loops remains manual. More importantly, it requires an extensive handcrafting of non-trivial implementation code. This may lead to inconsistencies and instabilities as no systematic and automated assurance can be obtained on the fact that the initial assumptions for the designed controller still hold in the resulting system. In this chapter, we rely on the principles of design-by-contract to ensure the correction and robustness of a self-adaptive software system built using feedback control loops. Our solution raises the level of abstraction upon which the loops are specified by allowing one to define and automatically verify system-level properties organized in contracts. They cover behavioral, structural and temporal architectural constraints as well as explicit interaction. These contracts are complemented by a first-class support for systematic fault handling. As a result, assumptions about the system operation conditions become more explicit and verifiable in a systematic way.

Towards Mapping Control Theory and Software Engineering Properties using Specification Patterns

“…Still the generated code, as we learned from the Ferrari case, needs to be further optimized and customized before it can be integrated into a larger system. This requires extensive manual modification of the code [33]-an error-prone activity that may introduce bugs that are hard to spot without proper verification techniques. In addition, when the controller becomes part of a larger system, it is also influenced by it, further questioning whether the necessary CT properties still hold.…”