Research on automatic software test case generation and path coverage optimization methods incorporating deep reinforcement learning

Software testing plays a crucial role in guaranteeing the quality and reliability of software, and the efficiency of test case generation and path coverage ability directly affect the effectiveness of testing. Traditional testing methods often face the problems of low path coverage and high testing cost under complex software structures. To this end, this paper proposes a software test case automatic generation and path coverage optimization method that integrates Deep Reinforcement Learning (DRL) mechanism. The method builds a reinforcement learning environment targeting program paths, models the testing process as the exploration behavior of an intelligent body in the program state space, and guides the intelligent body to automatically generate test sequences with high path coverage. In the test case generation process, a reward mechanism based on the dynamic adjustment of path sparsity is introduced, which effectively improves the exploration ability of rare paths. Experimental results show that the method has significant advantages in path coverage, testing efficiency and defect discovery capability compared with traditional methods on several open source software projects. The research in this paper provides new ideas and technical support for the intelligence and automation of complex software testing.