Naren Bao, "Subjective Risk Feeling Estimation Using Quantile Random Forest Regression"

2018/04/25 2:19 に Yuan Sheng が投稿

In this study, I first propose learn from demonstration framework for individual driving styles. and I investigated which features influence individual risk perception in lane change situations. A random forest method was applied to assess and rank the importance of each of these various risk features individually for each participant. i then estimated overall perceived risk levels for each participant for lane change scenes. Experimental results showed MSE and R2 value of estimated perceived risk for ten experimental participants using RF, so this estimation can not be considered as valid. So I assume the individual risk feeling is not easily estimated by RF, and we conduct QRF to catch the individual risk feeling intervals, and the result showed we still need more investigation.