编织
弹道
运输工程
工程类
计算机科学
汽车工程
物理
天文
机械工程
作者
Hongcheng Ge,Zhongyin Guo,Difei Jing,Cancan Song,Renjie Zhao,Qiong Sun,Jinming Zhou
标识
DOI:10.1061/jtepbs.teeng-9141
摘要
Highway-weaving areas face significant traffic efficiency challenges due to frequent merging and diverging movements. This study quantifies the precise thresholds at which risky driving behaviors impact traffic efficiency, using high-resolution drone data from two weaving sections in Kunming, China, collected during peak hours. The data cover more than 2 h of detailed vehicle motion, including speed, acceleration, and lane-change activities. Four risky driving behaviors—rapid acceleration, rapid deceleration, lane changes, and speeding—were analyzed to assess their impact on traffic flow. A traffic efficiency model was developed using traffic volume and the duration of risky behaviors within a 30-s window, considering both upstream and weaving area conditions. Rapid deceleration was most frequent near merging points, creating bottlenecks, while rapid acceleration and speeding were common after diverging points. High converging ratios were linked to longer travel times and reduced traffic efficiency. Random forest models achieved the highest predictive accuracy (R2=0.612–0.723) among the machine learning approaches tested. SHAP analysis identified specific critical thresholds (e.g., mainline traffic volume > 14.22 pcu, merging traffic > 8.12 pcu) beyond which traffic efficiency significantly deteriorates. Notably, a nonlinear relationship was discovered where moderate levels of rapid deceleration before the weaving area (below 17.74 s) actually improved traffic flow. The study provides actionable parameters for traffic management systems to implement targeted interventions such as extended merging lanes and adaptive speed controls, moving beyond descriptive observations to prescriptive guidance for improving traffic flow in weaving areas.
科研通智能强力驱动
Strongly Powered by AbleSci AI