Machine Learning for Predictive Traffic Safety

Overview

Improving road safety requires traffic infrastructure to anticipate potential hazards rather than merely detect them. This project explored the use of Digital Twin concepts and machine learning techniques to predict the future behavior of traffic participants based on infrastructure sensor data.

Using Eclipse SUMO and custom Python-based workflows, traffic scenarios were simulated and used to train and evaluate trajectory prediction models. The resulting concept demonstrated how edge-based processing and predictive analytics can support the early identification of potentially hazardous traffic situations.

Key Contributions:

• Development of a Digital Twin representation of traffic participants
• Investigation of machine learning methods for trajectory prediction
• Traffic simulation and model evaluation using Eclipse SUMO
• Custom Python tooling for data processing and analysis
• Exploration of edge-based predictive safety concepts for intelligent transport systems (ITS)