Robotic process automation (RPA) has received increasing attention in recent years. It enables task automation by software components, which interact with user interfaces in a similar way to that of humans. An RPA project life cycle is closely resembling a software project one. However, in certain contexts (e.g., business process outsourcing), a testing environment is not always available. Thus, deploying the robots in the production environment entails high risk. To mitigate it, an innovative approach to automatically generate a testing environment and a test suite for an RPA project is presented. The activities of the humans whose processes are to be robotized are monitored and a UI log is confirmed. On one side, the test environment is generated as a fake application, which mimics the real environment by leveraging the UI log information. The control flow of the application is governed by an invisible control layer that decides which image to show depending on the interface actions that it receives. On the other side, the test case checks whether the robot can reproduce the behaviour of the UI log. Promising results were obtained and a number of limitations were identified such that it may be applied in more realistic domains.

The Robotic Process Automation (RPA) paradigm has received increasing attention in recent years. It enables task automation by software components which interact with user interfaces in a similar way to that of humans. An RPA project follows a similar lifecycle as a software project. However, in certain contexts (e.g., business process outsourcing, BPO), a testing environment is not always available. Thus, deploying the robots in the production environment entails high risk. To mitigate this risk, an innovative approach to automatically generate a testing environment and a test case for an RPA project are described. The activities of the humans whose processes are to be robotized are monitorized and an UI log (i.e., a sequence of screen captures, mouse and key actions) is confirmed. On the one hand, the test environment is generated as a fake application, which mimics the real enviroment by leveraging the UI log information. To this end, the control flow of the application is governed by an invisible control layer which decides which image to show depending on the interface actions that it receives. On the other hand, the test case checks whether the robot can reproduce the behaviour of the UI log. A prototype has been constructed and tested in a controlled scenario. Promising results have been obtained and a number of limitations to be addressed have been identified such that it may be applied in more realistic domains.