Section: Research Program

Proof search and automated decision procedures

Interactive proof assistants provide a very expressive logical formalism, rich enough to allow extremely precise descriptions of complex objects like the meta theory of a programming language, a model of C compiler, or the proof of the Four Color Theorem. This description includes logical statements of the properties required by the objects of interest but also their formal proofs, checked by the merciless proof-checker of the system, which should be a small hence trusted piece of code. These systems provide the highest formal guarantee, for instance, of the correctness with respect to the mathematical specification of a code.

Proof-search is a central issue in such a formalization of mathematics. It is also a common aspect of automated reasoning and high-level programming paradigms such as Logic programming. However specific applications commonly involve specific logics or theories, like for instance linear arithmetic. Whether or not such a logical framework can express these at all, it is unlikely that its generic proof-search mechanisms can replace the methods that are specific to a logic or theory. Either because this specific domain lies outside the reach of generic proof-search or simply because generic proof-search is less efficient therein than a purpose-made procedure (typically a decision procedure).

But to enlarge the scope where a specific method applies, one can combine both generic proof search mechanisms with specific methods. We hence investigate how to craft formal proof producing decision procedures in the context of an interactive proof assistant. This activity includes understanding the impact of proof-search mechanism (polarization, focusing, etc.), the implementation of efficient connections between domain specific automated decision procedures (SMT solvers, polynomial optimization tools, etc.) with a proof assistant, and the combination of these two aspects in the design a unique logical framework where a generic notion of proof-search could serve each of the above purposes.