Extracting Test Sequences from a Markov Software Usage Model by ACO

The aim of the paper is to investigate methods for deriving a suitable set of test paths for a software system. The design and the possible uses of the software system are modelled by a Markov Usage Model which reflects the operational distribution of the software system and is enriched by estimates of failure probabilities, losses in case of failure and testing costs. Exploiting this information, we consider the tradeoff between coverage and testing costs and try to find an optimal compromise between both. For that purpose, we use a heuristic optimization procedure inspired by nature, Ant Colony Optimization, which seems to fit very well to the problem structure under consideration. A real world software system is studied to demonstrate the applicability of our approach and to obtain first experimental results.