Hardness of Deriving Invertible Sequences from Finite State Machines

Abstract

Many test generation algorithms use unique input/output sequences (UIOs) that identify states of the finite state machine specification M. However, it is known that UIO checking the existence of UIO sequences is PSPACE-complete. As a result, some UIO generation algorithms utilise what are called invertible sequences; these allow one to construct additional UIOs once a UIO has been found. We consider three optimisation problems associated with invertible sequences: deciding whether there is a (proper) invertible sequence of length at least K; deciding whether there is a set of invertible sequences for state set S' that contains at most K input sequences; and deciding whether there is a single input sequence that defines invertible sequences that take state set S to state set S'. We prove that the first two problems are NP complete and the third is PSPACE-complete. These results imply that we should investigate heuristics for these problems.