Abstract

When people perform computations, they routinely monitor their results, and try to adapt and improve their algorithms when a need arises. The idea of self-adaptive software is to implement this common facility of human mind within the framework of the standard logical methods of software engineering. The ubiquitous practice of testing, debugging and improving programs at the design time should be automated, and established as a continuing run time routine. Technically, the task thus requires combining functionalities of automated software development tools and of runtime environments. Such combinations lead not just to challenging engineering problems, but also to novel theoretical questions. Formal methods are needed, and the standard techniques do not suffice. As a first contribution in this direction, we present a basic mathematical framework suitable for describing self-adaptive software at a high level of semantical abstraction. A static view leads to a structure akin...

Citations