基于LightGBM的高速公路隧道段驾驶人压力负荷评估

Driver Stress Load Assessment of Freeway Tunnel Sections Based on LightGBM

驾驶人在行驶中的压力负荷水平直接影响驾驶操作稳定性和适宜性,而隧道路段又是行驶过程中驾驶操作易出现风险的关键点。鉴于此,为了对驾驶人在高速公路隧道路段行驶时产生的压力负荷进行准确、有效的评估,对驾驶人驶经该路段时的压力负荷水平等级分类进行了研究。首先,通过设计实施自然驾驶试验构建数据集,获取了32名被试人员在广东省某高速公路上驶经30处隧道时产生的共117组生心理、环境、驾驶行为等多模态数据。然后,结合主观NASA_TLX评分与客观生心理指标,利用三维聚类方法得到了驾驶人在隧道路段的实时压力负荷等级分类结果。最后,综合考虑车辆位置、驾驶行为、驾驶环境等多模态指标,采用LightGBM方法建立了压力负荷等级分类模型,分类准确率达到95.7%,预测精度达到93.3%,与XGBoost方法相比,准确率高0.71%、精度高0.93%,较RF,GBDT,SVM等主流分类模型算法更优。根据研究结果,可对驾驶人通过高速公路隧道路段的压力负荷进行有效评估。

The driver stress load level during driving directly affects the stability and suitability of driving operation,and tunnels sections are the key points of driving operation risk during driving.In view of this,in order to accurately and effectively evaluate the stress load of drivers driving in the freeway tunnel section,the classification of driver stress load level when driving through the section was studied.First of all,a naturalistic driving test was designed and implemented to construct a data set by classifying the stress load level of drivers as the research objective,and a total of 117 sets of multimodal data on biopsychological,environmental,and driving behaviors were obtained from 32 subjects driving through 30 tunnels on a freeway in Guangdong Province.Then,combining the subjective NASA_TLX scores and objective biopsychological indicators,the real-time stress load classification results of drivers in the tunnel sections were obtained by using the 3D clustering method.Finally,the stress load classification model was established by LightGBM method considering the multimodal indicators such as vehicle location,driving behavior,and driving environment,the classification accuracy of which achieved 95.7%,and the prediction precision of which achieved 93.3%.Compared with XGBoost method,the accuracy is 0.71%higher and the precision is 0.93%higher,and it is better than other mainstream classification model algorithms such as RF,GBDT and SVM.According to the research results,it can effectively evaluate the pressure load of drivers passing through expressway tunnel sections.