An Improved Test Tree Generation Algorithm from a Graphical Model

It is a widely recognized practice for test generation from the graphical model of the system using traversal algorithms in industry and academia. However, because traversal algorithms of the graph are not designed for test generation, some inexecutable test paths are usually generated from the graphical model of the system when the system has complex software behaviors. This problem will not only lead to the failure of software testing, but also greatly increase test cost of software. The paper discusses the problem of inexecutable test paths in model-based testing. Then, an improved algorithm is intended to generate a test tree from the graphical model so that test paths generated from the test tree satisfy transition constraints in the model. Next, we conduct an experiment on four systems to analyze the problem of inexecutable test paths. Experimental result shows that 1) our algorithm is more efficient than two traditional algorithms for constructing the test tree of the system, and 2) there are still some challenges that need to be overcome in order to obtain more reliable test cases in test generation from graphs.