Enhancing Robotic Mission Analysis via Process Mining and Visual Analytics

Robotic systems are increasingly used to automate complex tasks in dynamic environments, making mission analysis essential to improve reliability and performance. Among the existing techniques for analyzing robots’ behavior, process mining is particularly suitable since it enables the extraction of high-level behavioral patterns from event logs, providing actionable insights into the overall system behavior. However, the large volume and the low-level of data generated by robotic systems pose challenges for the effective use of process mining, which mainly relies on high-level representations of system activities. To address this, robotic mission execution should be recorded at a higher level of abstraction to enable its discovery and reasoning in terms of control-flow. At the same time, low-level contextual data, as spatial occupancy, energy consumption, and interactions, must be integrated to provide a comprehensive understanding of the mission. This integration can be achieved by combining process mining with visual analytics, thus enhancing the automated mission discovery with interactive visualizations of contextual data and enabling mission analysis from multiple perspectives. This paper presents a methodology for preparing and analyzing robotic mission data via process mining and visual analytics. The methodology guides robotic experts in preparing event logs compliant with process mining standards and supports mission analysis through the ROBOTRACE tool, which enables process discovery and contextual visualizations. The methodology is assessed through a multi-robot case study and an evaluation with domain experts to show its feasibility, interpretability, usability, and utility.