Scholarlink

<p>The national highway system of India still records an unequal number of serious accidents, with horizontal curves accounting for 54,593 reported accidents and 20,573 accidents being fatal annually. The research problem of this study is to investigate how car velocities change at the approach, middle, and exit points of circular horizontal curves, and the changes that determine the side friction necessitated to travel safely. Predictive models were created to help manage proactive speed through multiple linear regression, artificial neural networks, and a customized gradient-boosting model, based on XG-Boost. The gradient-boosting model with individual settings was the most reliable, with mean absolute percentage errors ranging from 0.35 to 10.58%, indicating its ability to represent nonlinear interactions between speed behavior, curve geometry, and driver responses. To identify potentially unsafe working conditions, a moving safety index (<i>D</i><sub><i>i</i></sub>) was developed as a comparison of the current friction demand with the maximum attainable friction (<i>f</i><sub><i>max</i></sub>). Vehicle-actuated and adaptive speed-display systems were used to create advisory speeds that could be used to generate a real-time warning. Field observations of cars, four-wheelers, buses, and trucks revealed significant variations in speed at the midpoint of the curves on national highways, with standard deviations ranging from 8.2 to 12.5&#xa0;km/h. The most substantial variation was observed in cars, where mean speeds decreased from 75.0&#xa0;km/h at the approach to 59.0&#xa0;km/h in the middle, and then increased to 73.1&#xa0;km/h at the exit. The results show that a combination of predictive modeling and the proposed dynamic safe advisory-speed warning system provides a strong basis for controlling speeds along the highway curves.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Integrated machine-learning prediction of speed profiles and friction demand for enhanced curve safety advisory systems

  • 期刊论文

  • 发表日期：

    2026年6月20日

查看全文
  • R. Srinivasa Kumar,
  • Hamid Noori

摘要

The national highway system of India still records an unequal number of serious accidents, with horizontal curves accounting for 54,593 reported accidents and 20,573 accidents being fatal annually. The research problem of this study is to investigate how car velocities change at the approach, middle, and exit points of circular horizontal curves, and the changes that determine the side friction necessitated to travel safely. Predictive models were created to help manage proactive speed through multiple linear regression, artificial neural networks, and a customized gradient-boosting model, based on XG-Boost. The gradient-boosting model with individual settings was the most reliable, with mean absolute percentage errors ranging from 0.35 to 10.58%, indicating its ability to represent nonlinear interactions between speed behavior, curve geometry, and driver responses. To identify potentially unsafe working conditions, a moving safety index (Di) was developed as a comparison of the current friction demand with the maximum attainable friction (fmax). Vehicle-actuated and adaptive speed-display systems were used to create advisory speeds that could be used to generate a real-time warning. Field observations of cars, four-wheelers, buses, and trucks revealed significant variations in speed at the midpoint of the curves on national highways, with standard deviations ranging from 8.2 to 12.5 km/h. The most substantial variation was observed in cars, where mean speeds decreased from 75.0 km/h at the approach to 59.0 km/h in the middle, and then increased to 73.1 km/h at the exit. The results show that a combination of predictive modeling and the proposed dynamic safe advisory-speed warning system provides a strong basis for controlling speeds along the highway curves.