Bayesian experimental design for safe transport

Research Project, 2026 – 2027

This project aims to deliver methodologies for scientific experimentation that can find its use in various transport domains. Using a Bayesian approach to Experimental Design, it will enable targeted scenario selection that accelerates convergence of models, reduces number of required evaluations, and quantifies uncertainties across diverse conditions. In trajectory planning for automated driving, this will allow real-time pruning of scenario trees, resulting in more efficient control decisions. For ADS safety assessment, it will generate optimally informative parameter conditions that enhance the robustness and interpretability of risk predictions, even under data scarcity. In injury assessment, the method will support the strategic use of multiple human body models to represent a wider range of human variability with fewer simulations, ultimately contributing to more inclusive and scientifically grounded vehicle safety evaluations.