An Adaptive Multi-Staged Forward Collision Warning System Using a Light Gradient Boosting Machine

The existing forward collision warning (FCW) systems that adopt kinematic or perceptual parameters have some drawbacks in the performance because of poor adaptability to users ineffectiveness warnings. To solve problems adaptability, several FCW models been proposed based on algorithms (machine learning, deep learning). However, there is a lack consideration for multi-staged avoid an abrupt may startle distract driver. In this study, light gradient boosting machine (LGBM) was adopted develop FCW. model trained and evaluated platform driving simulator by twenty drivers. Through Shapley Additive Explanations (SHAPs), output explained. Specifically, front vehicle acceleration, time-to-collision (TTC), relative speed were found strongly affect stages from model. evaluate utility acceptability developed model, it compared with three terms subjective objective indicators. As result, trade-off between user acceptance. Additionally, comparison study also indicated outperformed other previous due not only high accuracy but suitable trigger timing each participant.