Android application automatic test method and system based on reinforcement learning

Abstract

The invention discloses an Android application automatic test method and system based on reinforcement learning. The method comprises the steps of: in the testing process, automatically operating an Android APK by the automatic test tool, and obtaining a current interface control layout situation and speculating an executable interaction event; and adopting a Q-learning algorithm, when the interaction event is explored for the first time, obtaining an initial value, and selecting and executing the interaction event by the automatic test tool according to the value of the interaction event. A reward is generated after the interaction event is executed each time to update the value of the interaction event. The reward giving mainly takes the difference between the new state and the past state as a judgment standard. The neural network is introduced to compare the states, and whether the two states are in the same functional scene or not can be judged. Based on the reward given by the neural network judgment result, the automatic test tool can be guided to explore each scene in the Android application preferentially, so that the test efficiency is improved, and meanwhile, more defectsexisting in codes are discovered.

Description

technical field [0001] The invention relates to Android application automatic testing, in particular to an Android application automatic testing method and system based on reinforcement learning. Background technique [0002] With the popularity of smartphones, mobile applications have brought great changes to our lives. However, due to problems such as a large state space, it takes a lot of manpower and time to conduct a relatively comprehensive test on mobile applications. This demand is difficult to meet when major Internet companies are rapidly launching rapid version iterations of applications. Therefore, automated testing tools for mobile, especially efficient automated testing tools, are of great value. Among them, the Android system accounts for more than 70% of the mobile operating systems, so it is particularly important to automatically test Android applications. [0003] At present, there are some automatic testing methods for Android applications, but practice...

AI Technical Summary

Problems solved by technology

This type of method is limited by techniques such as symbolic execution, and performs well in scenarios where specific code is covered, but is not suitable for scenarios that require overall code coverage

Images