Department of Computer, Control and Management Engineering

Robotic Process Automation (RPA) is an emerging technology in the field of Business Process Management (BPM) that enables the automation of intensive repetitive tasks (or simply routines). RPA solutions access the user interface (UI) layer of software (SW) applications and provide a virtual workforce of SW robots that are able to mimic human keyboard and mouse interactions with a UI as if a real person was doing them. To take full advantage of this technology, organizations leverage the support of skilled human experts that preliminarily observe how routines are executed on the UI of the involved SW applications, and then implement the executable RPA scripts required to automate the routines enactment by SW robots on a target computer system. However, the current practice is time-consuming and error-prone, as it strongly relies on the ability of the human experts to correctly interpret the routines (and their variants) to automate. In this paper, to tackle this issue, we use a design science research method to develop an approach, called SmartRPA, which is able to interpret the UI logs keeping track of many routine executions, and to automatically synthesize SW robots that emulate the most suitable routine variant for any specific intermediate user input that is required during the routine execution. The approach is implemented as an open-source tool and evaluated with four non-functional requirements employing both syntectic and real-world data.