Aleksandar Marjanovic

Topic: Dynamic Modeling for Data-Driven Test Prioritization in Automotive Validation Development of a data-driven system for the analysis and prioritization of test cases in complex technical systems Results & Impact: - Reduced manual test planning effort through data-driven prioritization. - Improved quality of data-driven decision-making. - Increased scalability and efficiency within the testing process.

- Compared modern forecasting models (RNN, Transformer, Koopman operator) for modeling dynamical systems. - Developed and trained models on synthetic datasets (Van der Pol, Lorenz) and real IoT room-temperature measurements. - Evaluated model accuracy, robustness to noise, training time, and implementation complexity. - Identified strengths and limitations of each method for deterministic, stochastic, and chaotic system behavior.

Automation and analysis of large-scale technical communication datasets Results & Impact: - Significantly improved efficiency in data analysis. - Enhanced data quality for downstream analyses. - Enabled early detection of system instabilities and communication faults.

- Developed and validated an ECU-TEST extension for qualitative test script analysis (YAML rule design, Python enhancements). - Supported HiL test benches, including modifications, troubleshooting, and service tasks. - Analyzed findings, documented issues, and communicated results within the test and quality teams.