Constructing multiple unique input/output sequences using metaheuristic optimisation techniques

Abstract

Multiple unique input/output sequences (UIOs) are often used to generate robust and compact test sequences in finite state machine (FSM) based testing. However, computing UIOs is NP-hard. Metaheuristic optimisation techniques (MOTs) such as genetic algorithms (GAs) and simulated annealing (SA) are effective in providing good solutions for some NP-hard problems. In the paper, the authors investigate the construction of UIOs by using MOTs. They define a fitness function to guide the search for potential UIOs and use sharing techniques to encourage MOTs to locate UIOs that are calculated as local optima in a search domain. They also compare the performance of GA and SA for UIO construction. Experimental results suggest that, after using a sharing technique, both GA and SA can find a majority of UIOs from the models under test.