Using Automatic Failure Detection for Cognitive Agents in Eclipse: (Demonstration)

In order to reduce debugging effort and enable automated failure detection, we proposed an automated testing framework for cognitive agent programs that provides support for detecting frequently occurring failure types in [14]. Automated testing yields a reduction in the effort needed to detect a failure and is more effective than manual code inspection methods [16]. A failure is an event in which a system does not perform a required function within specified limits [10]. Failures thus are manifestations of undesired behaviour. They are caused by a fault, an incorrect step, process, or data definition in a program [10] or mistake in a program [17]. Upon detecting a failure, a programmer needs to locate and correct the fault that causes the failure. We introduced a test language based on two basic temporal operators, and use this language to specify test templates for detecting failure types. These test templates refine a failure taxonomy introduced previously in [17]. A test approach has also been specified that explains how to instantiate test templates and derive test conditions for specific failure types. The main steps of this approach are (i) to define success in terms of functional requirements, (ii) to test cognitive state updating, and (iii) to classify failures that concern actions and goals.