Human-machine Collaborative Guidance Technology For Crowdsourcing Testing Of Mobile Applications

With the development of mobile phone operating systems,the Android operating system occupies 71.93%of the mobile phone market and the applications developed on the basis of the Android system have alse soared year by year.The large number of Android mobile applications and rapid iteration of applications have promoted the development of Android mobile application testing.The mainstream testing methods include mobile application automated testing and crowdsourcing testing.mobile application automated testing can test applications quickly and efficiently at low cost,but due to the serious fragmentation of the Android system,its test results cannot meet the ideal requirements.Crowdsourced testing is conducted with the help of crowdsourced workers,which enables applications to be tested in different models and different system versions.However,due to the different professional abilities of crowdsourced workers,the test results are quite different.In this context,this thesis proposes a human-machine collaborative guidance technology for mobile application crowdsourcing testing,which is compatible with the advantages of automated testing and crowdsourcing testing.It uses automated test results to guide crowdsourcing workers to conduct tests and improve their testing capabilities.At the same time,it can test uncovered windows by leveraging the power of crowdsourced workers to improve window coverage.This system includes automated test data processing module,Android static analysis module,Android boot module and server module.Firstly,the automated test data processing module can analyze the automated test results of mobile applications,and extract the information of non-repetitive exceptions and path tested on different models as the basis for the recurrence of the exceptions.Secondly,The Android static analysis module conducts static analysis on mobile applications by using static analysis tools to obtain the jump paths of all Windows of applications.By comparing it with the automated test results to find out the windows that are not covered in the automated test,and guide crowdsourcing workers to test them.The Android boot module is embedded in the application as a plug-in.During the test,the crowdsourced workers select the event to be tested by clicking the exception recommend button or uncovered recommend button,and use the path guidance button to quickly locate the location of exception or uncovered windows.The crowdsourcing workers click the buttons to request services.The server uses the Spring Boot framework and Mybatis to realize the functions of user login,task recommendations,path guidance,and test results reporting,which makes the system more scalable.The system has been tested on the Mooctest platform.This system selected 10 open source mobile applications in GitHub,and conducted a 20-minute integration acceptance test on 10 Android devices by setting up 4 sets of experiments.Analysis of the results submitted by users shows that the system has significantly improved the efficiency of workers' testing of exceptions and the number of exceptions found.In terms of window coverage,the system achieves full coverage of window paths,effectively making up for the lack of automated testing.