BPO Process Models for Robot Script Generation

This repository contains the process models associated with the study “MetaRPA: A Model-Driven Framework for Robotic Process Automation.” These models are used as case studies in the experimental evaluation of the proposed Model-Driven Engineering (MDE) pipeline for software robot generation. The dataset includes BPMN representations of three business processes derived from real-world Business Process Outsourcing (BPO) scenarios identified by domain experts from a collaborating organization. The process models provided in this repository correspond to the process-level representations derived from enriched UI logs. Within the proposed approach, these models serve as the baseline input for the model-to-model (M2M) transformations in the MDE pipeline. A detailed description of each case study is provided below. C1 - Invoice Management This case study represents the management of customer invoices within the Odoo enterprise platform. The process reflects a typical back-office financial operation where employees interact with the system’s user interface to create and register invoices. The workflow begins with the entry of customer information, followed by the creation of invoice lines and the specification of invoice details. Once the required information has been completed, the process concludes with the confirmation and storage of the invoice in the system. This case corresponds to a simple linear workflow composed of four tasks and a single execution path. The absence of decision points makes it the simplest scenario in the dataset. This case serves as a baseline process for validating the transformation pipeline and verifying that operational actions captured from user interactions can be consistently represented and later transformed into robot scripts. C2 - Application Evaluation The second case study models the evaluation of candidate applications within the Odoo platform. This workflow represents a simplified recruitment process in which candidate information is registered and subsequently assessed before determining the final outcome of the application. The process begins with the registration of candidate data in the system. After this information is recorded, evaluators perform assessment activities to analyze the candidate profile and determine whether the application should proceed or be rejected. Structurally, the process consists of six tasks and introduces a decision point that produces two possible execution variants depending on the evaluation outcome. This case allows the evaluation of how the transformation pipeline handles conditional control-flow structures when generating software robots from process models. C3 - Unsubscription Process The third case study represents the management of user unsubscription requests across multiple software platforms. This scenario models a service management workflow that integrates several applications commonly used in operational environments. The process begins when a user request for unsubscription is received via email. The request is subsequently registered as an incident in the Zendesk support platform, after which customer data stored in the Odoo system is retrieved and verified. Based on the retrieved information and the operational conditions associated with the request, the process determines the appropriate handling procedure for the unsubscription. This workflow involves interactions with multiple systems, including Gmail, Zendesk, and Odoo. Structurally, the process consists of seven tasks and includes decision logic that generates four possible execution variants. Due to the presence of cross-platform interactions and multiple possible outcomes, this case represents the most complex scenario in the dataset and provides a realistic test case for evaluating the transformation pipeline under more demanding conditions.