Final Report Summary - INVEST (inVEST: Foundations for a Shift from Verification to Synthesis)
Reactive systems are computer systems that maintain a continuous interaction with the environment in which they execute. Examples of reactive systems are controllers embedded in cars or planes, system level software, device drivers, communication protocols, etc. On the one hand, those systems are notoriously difficult to develop correctly (because of characteristics like concurrency, real-time constraints, parallelism, etc). On the other hand, their correctness is often critical as they are used in contexts where safety is an issue, or because of economical reasons related to mass production.
To ensure reliability of reactive systems, advanced verification techniques have been developed. One particularly successful approach is model-checking. Nevertheless, model-checking is used to find bugs in designs but it does not support the design itself.
In this project, we have developed new theoretical foundations in order to allow the development of new algorithms and tools to support the automatic synthesis of modern reactive systems (instead of their verification a posteriori). Our work contributes to make possible a shift from verification to synthesis. This shift is based on new theoretical foundations that generalise transition systems and automata – models of computation in the classical approach to verification – by the more flexible, and mathematically deeper, game-theoretic framework.
Our work has mostly been of fundamental nature but we have also developed new algorithms and implemented them in prototypes of tools. In particular, we have made a large number of contributions to the following list of research subjects:
-new solution concepts for synthesis and non-zero sum games played on graphs (Assume admissible strategies, Regret minimization, Subgame perfection, combination of worst-case and expectation guarantees)
-quantitative games (a.o. multi-dimensional quantitative games, quantitative secure equilibria)
-games with imperfect information (a.o. quantitative games with imperfect information, admissibility and imperfect information)
-new synthesis algorithms based on abstraction and compositional reasoning
-new prototypes of tools (Absynthe for symbolic safety games, and Acacia+ for LTL games)
Those new foundations and prototypes will find their way in a new generation of computer-aided design tools that will support the automatic synthesis of modern reactive systems and ensure correctness by construction.
To ensure reliability of reactive systems, advanced verification techniques have been developed. One particularly successful approach is model-checking. Nevertheless, model-checking is used to find bugs in designs but it does not support the design itself.
In this project, we have developed new theoretical foundations in order to allow the development of new algorithms and tools to support the automatic synthesis of modern reactive systems (instead of their verification a posteriori). Our work contributes to make possible a shift from verification to synthesis. This shift is based on new theoretical foundations that generalise transition systems and automata – models of computation in the classical approach to verification – by the more flexible, and mathematically deeper, game-theoretic framework.
Our work has mostly been of fundamental nature but we have also developed new algorithms and implemented them in prototypes of tools. In particular, we have made a large number of contributions to the following list of research subjects:
-new solution concepts for synthesis and non-zero sum games played on graphs (Assume admissible strategies, Regret minimization, Subgame perfection, combination of worst-case and expectation guarantees)
-quantitative games (a.o. multi-dimensional quantitative games, quantitative secure equilibria)
-games with imperfect information (a.o. quantitative games with imperfect information, admissibility and imperfect information)
-new synthesis algorithms based on abstraction and compositional reasoning
-new prototypes of tools (Absynthe for symbolic safety games, and Acacia+ for LTL games)
Those new foundations and prototypes will find their way in a new generation of computer-aided design tools that will support the automatic synthesis of modern reactive systems and ensure correctness by construction.