Testing of autonomous agents: A critical analysis

Autonomous agents perform on behalf of the user to achieve defined goals or objectives. Testing of autonomous agents ensures their reliability which is very important to ensure user confidence. Testing of autonomous agents is a challenging task due to their autonomous and proactive behavior. Testing to build confidence in their operations is crucial. In this paper, the existing techniques for testing of autonomous software agents are surveyed and evaluated. Our main focus is on specification and model based testing of autonomous agents. Our analysis concludes that the existing techniques do not cover all aspects of autonomous agents, which may not guarantee the user trust as agents may exhibit unexpected behavior even after being tested.