Collaborative Approach to Sensor Perception Modelling (SENSI)

Research Project, 2025 – 2026

Automated driving systems (ADS) hold potential for safer roads, but their widespread adoption relies on robust and reliable perception systems. This project addresses a lack of open-source, transparent perception models that accurately reflect the uncertainties of real-world driving scenarios. Currently, virtual safety assessments of ADS often rely on idealized perception models, neglecting the real-world stochasticity of the environment. Additionally, existing complex, physicsbased perception models are typically not open-source, difficult to set-up, and computationally expensive for real-time control. This project aims to bridge this gap by developing open-source, transparent, and computationally efficient stochastic perception models with different levels of modelling complexity. These abstracted models will be applied in two key use cases: 1) The models will be implemented into virtual simulations, enabling more realistic and transparent safety assessments of ADS. 2) The models will be integrated into the prediction module for optimal,
model-based, real-time control in ADS. This will lead to enhanced real-time decision-making and control capabilities, further contributing to safer and more efficient automated driving.