Automatic Generation Of Embedded Software Test Cases Based On Improved Particle Swarm Optimization

With the increasing demand and scale of application software,automated testing has become the main trend of software testing.The traditional method of manual generation of test cases is time-consuming and labor-intensive,produces fewer cases,and requires experienced testers to ensure code coverage.In the fast updating process of software,manual method often affects the progress of software testing.However,due to the diversity of test procedures and the uncertainty of test cases,it is difficult to automatically generate reduced test sets with high coverage.And because the generation of test cases is undecidable,many search-based algorithms have inherent shortcomings.Particle Swarm Optimization(PSO)is fast,robust,and easy to use.However,the traditional PSO has the drawback of easily falling into local extremum and premature convergence or stagnation.In this paper,the PSO is improved from four aspects:particle swarm initialization,neighborhood topology,parameter setting and blending strategy,according to the data stream dependency in the program and the proposed evolutionary structure testing system.The main improvements to the algorithm include the use of non-uniform random distribution of the way to initialize the particle group,combined with comprehensive learning PSO to the neighborhood topology,the use of shrink factors rather than the inertia of the speed update strategy,and the algorithm of the relevant parameters Value settings.Finally,we compare the improved PSO with the random testing algorithm as the benchmark,and prove that the algorithm of this paper has some improvement on the generation efficiency and the simplification of use cases set.Then,the test system designed in this paper is applied to a comprehensive testing and verification platform of our laboratory.The test case of the CAN communication program of a certain DSP is automatically generated in combination with the scene-based test method,and the coverage is high and the test case set is more compact.