Test Case Generation as an AI Planning

While Artiicial Intelligence techniques have been applied to a variety of software engineering applications, the area of automated software testing remains largely unexplored. Yet, test cases for certain types of systems (e.g., those with command language interfaces and transaction based systems) are similar to plans. We have exploited this similarity by constructing an automated test case generator with an AI planning system at its core. We compared the func-tionality and output of two systems, one based on Software Engineering techniques and the other on planning, for a real application: the StorageTek robot tape library command language. From this, we showed that AI planning is a viable technique for test case generation and that the two approaches are complementary in their capabilities. 1. Automated Test Case Generation Testing consumes a large amount of time and eeort in software development. Although critical for ensuring reliability and satisfaction, much of the process is tedious: constructing data sets and/or sequences of commands to probe for faults. Some of the test cases will uncover faults, but one expects that most will not. A variety of approaches have been put forth for automated test generation in Software Engineering (SE) (see section 2). Another approach is to acknowledge a similarity between some types of test cases (i.e., sequences of commands for software with command language interfaces) and plans in Artiicial Intelligence (AI). Both are sequences of commands to achieve some goal; both need to conform to syntactic requirements of the commands and the semantic interactions between commands. Considerable research in AI planning has addressed subgoal and operator interactions; thus, the mechanisms of planning seem ideally suited to test case generation. In this paper, we explore the similarities between test case and plan generation, propose another approach, as implemented in a prototype, to automated test case generation using an AI planner and compare the new approach to an existing SE approach.