Research On Multi-agent System Communication Testing Technology

Since Multi-Agent system has been applied to more and more application fields,such as government, business and service area, the research on reliability ofMulti-Agent system has become a hot studying topic. Due to the characters of Agent,there are many differences between Agent-oriented software testing and traditionalsoftware testing. It is obvious that Agent system testing will become a new studyingarea.At present, most foreign researchers have focused on the studying to use TestingAgent as a test tool to test target Agent. Some even have presented the related testingframework and testing Agent architecture. However, the testing efficiency anderror-locating ability of the presented testing Agent still need to be improved.In order to increase the testing efficiency of Agent communication and the testresult reliability, this paper presents an improved testing Agent model. With theinstruction of the constructed model, testing process send sessions in unit instead ofmessages. The testing Agent can detect and mark errors according to thecommunication content. When testers analyze the testing documents, they can locatethe errors existing in the program by judging the communication status of the markederrors. Testing Agent not noly test the target objects but also test itself. By doing that,testing Agent can figure out if the errors are generated by the tested objects or byitself. Since one session involves more than one object in practice, this paper adoptsthe several-for-one method to implement the experiment and the experimental resultshows the reported testing Agent can conduct the communication testing for singleobject efficiently.Agent UML Sequence diagram is used to describe the communication sequencebetween Agents. This paper reports a transformation rule to transform the Agent UMLSequence diagram into information flow diagram. Based on that, this paper proposesan test case generation method of Agent communication testing by using extendedsequence diagram. According to the transformation rules, we first obtain theinformation flow diagrams of each Agent object and take advantage of the diagrams togenerate test cases to meet path coverage criteria. This paper reports an improved pathcoverage algorithm. This algorithm can resovle the loop problem in information flowdiagram, then obtain all the independent paths and finally use the acquired independent paths to generate relative test cases effectively. The case study shows itis feasible to generate test cases by using Agent UML Sequence diagram.