Research And Realization Of Embedded Software Testing System Based On Multi-Agent

As the growing of embedded system applications, the complexity and the scale of system software also grow increasingly. The testing and verification are the guarantee of the validity and safety of embedded system software. Current software testing technologies are mainly based on the traditional coverage testing and most of them are manual, one problem is that these technologies cannot perfom effectively or guarantee the quality of software, besides, the coupling between embedded software and hardware is very hard, so there is an urgent need for embedded software testing tools which can analyse automaticly and intelligently. This thesis brings the idea of Agent into software testing area, designs and realizes an intelligent embedded software testing system called AgEast with using the property of Agent. This system can realize these function:Firstly, it realizes the black-box functional test based on strictly analyzes the testing tools of embedded software domestic and overseas. Through delivering the black-box test cases to target machine using serial comunications and receiving the output at the host, and then making analysis and process with these signals, we can decide whether the function is realized or not.It realizes the white-box coverage test. It presents a multi-agent model for embedded software testing. we take lexical analysis and syntax analysis with sourse program in pretreatment Agent, and uses the static-instrumentation method to complete instrumentation for the source code of embedded software in program instrumentation Agent, then we translate and edit the program and transplant it to the target machine, so that we can compute the statement coverage and branch coverage according to interrelated arithmetic through inputing test cases.It realizes the automatic generation of the test case in two aspects, we build the BDI model TC_Agent, realize the automatic generation of the black-box test case according to the restriction relationship of belief,desire and intention; we build the GA_Agent model which based on genetic algorithm, realize the automatic generation of the white-box test case through designing the genetic manipulation which consists of coding, fitness function, selection, crossing and variationFinally, the thesis use the AgEast testing system which we design to have a test on the program of robot palletizer, and the result tells us that both of the function and coverage rate reach expected requirements, so that we primely prove the practicability of this system.