Application of Recurrent Neural Networks for Drive Style Based Vehicular Accident Prediction

In our research, we endeavor to forecast vehicular accidents by leveraging internal measurement unit (IMU) data in conjunction with recurrent neural networks (RNNs). This methodology constitutes a pivotal component of a broader initiative aimed at utilizing IMU data for the analysis and prognostication of driver behavior. An overarching objective and obstacle within this project entails the utilization of a simplistic measurement device that is both facile to install and cost-effective for in-car deployment, thereby constraining the scope of attainable data collection. Our experimental endeavors encompass the acquisition of proprietary data on race tracks, followed by the input of this data into recurrent neural networks for analysis.