Learning of cooperative behavior for multi-agent software testing.

摘要

In this thesis work we present a new model for testing complex systems. Our model is based upon using learning of cooperative behavior for testing complex systems. Specifically, we concentrate on the testing of multi-agent systems for high level behavior. Our model includes a multi-agent system that learns the required coordination needed between a group of tester agents to cause a desired behavior to be exhibited in the targeted system. As a test application, we use the Agent Rescue Emergency Simulator (ARES) system. The ARES system presents a simple post urban disaster zone where different agents must work together to save survivors. The ARES system is used to show our model in action as it relates to the testing problems presented by the system.