传统运行速度预测模型以二维平、纵线元参数为自变量，其预测的速度变化无法与实际道路线形空间几何特性、驾驶行为保持同步。相应的一致性评价方法无法识别线形中存在的安全隐患，是目前线形设计安全和优化面临的主要困境之一。运行速度预测应立足于线形空间几何特性，其变化应体现人-车-路（环境）的耦合作用。因此，本项目从三维线形本质出发，以山区高速为研究对象，依据微分几何曲线论，采用空间曲线几何不变量对不同组合线形空间几何特性进行统一描述和分析。结合自然驾驶实验与虚拟仿真等技术，分析线形空间几何特性变化对驾驶行为和速度选择的内在影响机理。综合考虑横断面和路侧环境影响，采用人工神经网络算法构建三维运行速度预测模型。在此基础上，提出基于速度连续均衡变化的线形安全性评价体系，形成完善的安全性评价原型系统。力图从源头上避免因山区高速线形设计不足而引发的交通事故，为提高线形设计安全性提供新的视角。

Traditional operating speed prediction model were carried out with a function of the parameters of two-dimensional (2D) horizontal and vertical elements. The changes of the predicted speed could not keep pace with the spatial geometric characteristics of highway alignments and driving behavior. This is the main dilemma in the safety evaluation and optimization of highway alignment design, as the hidden safety risks cannot be identified by the speed consistency evaluation method. Operating speed prediction should be established with spatial alignment geometric characteristics, and its variations should reflect the coupling effects of diver-vehicle-road (environment). Therefore, this project takes the mountain expressways as the research subject. New indices, spatial geometric intrinsic invariants are adopted to describe and analyze the spatial geometric properties of different combined alignments with view of highway three-dimensional (3D) nature and the differential geometric principles. And the influence mechanisms of the spatial geometric characteristics changes on driving behavior and speed selection are analyzed, with the natural driving experiment and driving simulation. The artificial network model is used to develop the 3D operating speed model, considering the influences of cross section and roadside environment. On the base of the above research, a new alignment safety evaluation frame based on the continuity and homogeneity of speed variation is proposed, and an improved safety evaluation system is formed. This project tries to avoid traffic accidents of mountain expressways caused by the alignment design defect from the origin, and bring a new perspective to improve the safety level of alignment design.