Université Paris Sud – LRI 06 / 2010

This article presents optimizations of a randomized method that generates paths while ensuring a good coverage of the model, regardless its topology. The optimizations aim at diminishing the required memory, thus allowing the generation of longer paths. Pure random exploration generally leads to a bad coverage of the model. Methods, based on counting and uniform drawing in combinatorial structures, can ensure a good coverage of paths. Due to memory consumption, such methods can neither explore very large models nor generate very long paths. In this paper, we leverage the limitation of path lengths by using new algorithms with better space complexity. Experimental results show significant improvement over previous randomized approaches. This work opens new perspectives to efficiently explore models for simulation, random testing and model-checking purposes.