Testing Multi-agent Systems for Deadlock Detection Based on UML Models

There is a growing demand for Multi-Agent Systems (MAS) in the software industry. The autonomous agent interaction in a dynamic software environment can potentially lead to runtime behavioral failures including deadlock. In order to bring MAS to the main stream of commercial software development, the behavior of MAS must be tested and monitored against the risk of unwanted emergent behaviors including deadlocks. In this paper, (1) we introduce a method for preparing test requirements for testing MAS; and (2) we deploy a MAS monitoring method for deadlock detection in MAS under test. The first method helps create test requirements using a resource requirement table from the MAS analysis and design. The second method monitors the MAS behavior to detect deadlocks at the run-time. Also, as model based software techniques such as Multi-agent Software Engineering (MaSE) are gaining more popularity; these model based approaches can help MAS developers to reduce the risk of having unwanted emergent behaviors such as deadlock in MAS.