Agents Learn from Human Experts: An Approach to Test Reconfigurable Systems

摘要

Faulty software is costly and possibly life threatening as software products permeate our daily life. Therefore, the test process formulates an indispensable component of the development cycle; yet it is a formidable task. In an effort to alleviate its challenges, this contribution outlines a novel approach to enrich traditional test techniques with intuition-based test strategies learned by observing skilled human testers during various test sessions. Consequently, the strategies learned would be verified, combined, and generalized to be further applied in similar test situations. Hence, a reasonable portion of the workload done by human testers would be shifted to the test system itself. This leads to a significant reduction in the development time and cost; yet the test efficiency is not sacrificed.