Automatic synthesis of code using genetic programming

Abstract

Correct-by-design automatic system construction can relieve both programmers and quality engineers from part of their tasks. Classical program synthesis involves a series of transformations, starting with the given formal specification. However, this approach is often prohibitively intractable, and in some cases undecidable. Model-checkingbased genetic programming provides a method for software synthesis; it uses randomization, together with model checking, to heuristically search for code that satisfies the given specification. We present model checking based genetic programming as an alternative to classical transformational synthesis and study its weakness and strengths.