ON THE MECHANICAL DERIVATION OF LOOP INVARIANTS

We describe an iterative algorithm for mechanically deriving loop invariants for the purpose of proving the partial correctness of programs. The algorithm is based on resolution and a novel unskolemization technique for deriving logical consequences of first-order formulas. Our method is complete in the sense that if a loop invariant exists for a loop in a given first-order language relative to a given finite set of first-order axioms, then the algorithm produces a loop invariant for that loop which can be used for proving the partial correctness of the program. Existing techniques in the literature are not complete. © 1993, Academic Press Limited. All rights reserved.