Emergency-braking Distance Prediction using Deep Learning
This addresses safety features for vehicles, but it is incremental as it applies existing deep learning methods to a new dataset.
The study tackled predicting emergency-braking distance for vehicle collision avoidance by proposing a deep-learning model that uses 0.25 seconds of 3D acceleration data as input, achieving accuracy within 3 feet across two road surfaces.
Predicting emergency-braking distance is important for the collision avoidance related features, which are the most essential and popular safety features for vehicles. In this study, we first gathered a large data set including a three-dimensional acceleration data and the corresponding emergency-braking distance. Using this data set, we propose a deep-learning model to predict emergency-braking distance, which only requires 0.25 seconds three-dimensional vehicle acceleration data before the break as input. We consider two road surfaces, our deep learning approach is robust to both road surfaces and have accuracy within 3 feet.