Test Agents: The Next Generation of Test Cases

Abstract

Growth of software size, lack of resources to perform regression testing, and failure to detect bugs faster have seen increased reliance on continuous integration and test automation. Even with greater hardware and software resources dedicated to test automation, software testing is faced with enormous challenges, resulting in increased dependence on centralized and complex mechanisms for automated test case selection as part of continuous integration. These mechanisms are currently using static entities called test cases that are concretely realized as executable scripts. Our key vision is to provide test cases with more reasoning, adaptive behavior and learning capabilities by using the concepts of software agents. We refer to such test cases as test agents. The model that underlie a test agent is capable of flexible and autonomous actions in order to meet overall testing objectives. Our goal is to increase the decentralization of regression testing by letting test agents to know for themselves when they should be executing, how they should update their purpose, and when they should interact with each other. In this paper, we envision test agents that display such adaptive autonomous behavior. Existing and emerging developments and challenges regarding the use of test agents are explored-in particular, new research that seeks to use adaptive autonomous agents in software testing.

An Agency-Directed Approach to Test Generation for Simulation-based Autonomous Vehicle Verification

“…Such intelligent, agent-based test generation has first been applied in the human-robot interaction domain [30], where a coverage-driven test generation approach was supplemented with reinforcement learning to improve the efficiency and effectiveness of testing the critical part of a robot-to-human handover task within a collaborative manufacturing scenario. Test agents have also been proposed by Enoiu et al [31] for regression testing, although they are used more for test selection from a library of tests, rather than for test generation. These agents use inter-agent messaging to decide what tests to execute and with what prioritisation.…”