Prediction of Occupant Safety Using Machine Learning

This white paper discusses the application of Machine Learning and CARLA autonomous driving simulation software to enhance vehicle safety by predicting occupant injuries in crash scenarios. It outlines how traditional crash tests are limited to controlled environments and regulated scenarios, while this study utilizes 'real-case' data from CARLA to simulate various traffic accident scenarios. The paper details the process of using Finite Element analysis to assess occupant safety and injury prediction based on simulated data. It describes the creation of datasets for training Machine Learning models that optimize safety parameters such as airbag deployment and seatbelt triggering. The study focuses on rear-end collisions, the most common type of traffic accident, and explains the methodology for collecting and analyzing data to improve vehicle design safety. Additionally, it presents the development of a Predictor model that forecasts head injury criteria, enhancing the efficiency of safety system optimizations across different crash scenarios.