Mutation Testing for Jason Agents

Most multi-agent system (MAS) testing techniques lack empirical evidence of their effectiveness. Since finding tests that can reveal a large proportion of possible faults are a key goal in testing, we need techniques to assess the fault detection ability of test sets for MAS. Mutation testing offers a direct and powerful way to do this: it generates faulty versions of the program following mutation operators then checks if the tests can distinguish the original program from those versions. In this paper, we propose a set of mutation operators for the Jason agent-oriented programming language, and then introduce a mutation testing system for individual Jason agents that implements some of our proposed mutation operators. We use our implemented mutation operators to assess the tests for a small Jason system, and show that the tests that meet a combination of existing coverage criteria do not kill all mutants.