Dynamic Risk Assessment for Automated Driving System using Artificial Neural Network

This paper introduces an approach for Dynamic Risk Assessment (DRA) for Automated Driving System (ADS) using an Artificial Neural Network (ANN) model. With the increasing complexity and intelligence of ADS, traditional risk assessment methods, which are adequate for human-driven vehicles, fall short in addressing the dynamic and evolving operational environment of ADS. Our proposed approach leverages the ANN model, a deep learning technique, to analyze riskspecific context information. This analysis enables the identification and classification of severity and controllability ratings, vital for continuous risk assessment for ADS. ANN models have demonstrated significant success in various real-world ADS applications, particularly due to their ability to efficiently process and recognize patterns within large, complex datasets. By applying ANNs to risk-specific context information, our approach provides an accurate and objective runtime assessment of risk levels, thereby enhancing the situational awareness of ADS. This enhancement enables ADS to more effectively predict and assess risks that could lead to accidents. This capability marks a step forward, overcoming the limitations of traditional risk assessment methods.