Decision support traffic model for assessing transportation safety performance

本研究将开发一种行为微观模型，以识别不同道路几何条件下交通流中潜在的不安全车辆相互作用。该模型基于许多基本交通协议，包括车辆跟随、车道变化、间隙接受和超车。随着时间的推移，安全性能表示为成对车辆速度和间距差异，车辆减速轮廓和制动能力的函数。所提出的研究的一个主要方面涉及基于可观察到的真实世界车辆跟踪数据的交通模型输入参数的校准。该过程包括：1)识别单独或组合对安全性能有统计显著影响的参数，2)使用多准则拟合函数建立这些参数的“最佳估计值”。模型参数的全面标定为模拟安全性能提供了良好的客观依据。这项拟议的研究将评估模拟安全性能措施实时捕捉观察到的碰撞发生的能力（即在每次观察到的碰撞发生之前的一段时间间隔）。将采用一些统计测试，以确定安全性能措施在识别交通流中的高风险驾驶员-车辆行为方面的能力。在不同的道路几何形状和交通情况下，这种高风险行为预计会导致碰撞。迄今为止进行的研究表明，虽然行为微观交通模型可能无法预测碰撞，但它们可以复制或识别交通流中的高风险驾驶情况。微观模拟模型的有用性研究了两个基本的安全问题：哪些位置本质上是不安全的？以及哪些对策最有可能提高这些地点的安全性。

This research will develop a behavioural microscopic model to identify potentially unsafe vehicle interactions in the traffic stream for varying road geometric conditions. The model is based on a number of fundamental traffic protocols, including car-following, lane change, gap acceptance and over-taking. Safety performance is expressed over time as a function of pair-wise vehicle speed and spacing differentials, vehicle deceleration profiles and braking capabilities. A major aspect of the proposed research concerns the calibration of traffic model input parameters, based on observable real-world vehicle tracking data. The process involves: 1) identifying parameters that individually or in combination have a statistically significant effect on safety performance, and 2) using a multi-criteria fitting function to establish "best estimate" values for these parameters. A comprehensive calibration of model parameters provides a sound objective basis for simulating safety performance. This proposed research will assess the ability of the simulated safety performance measures to capture observational crash occurrence in real-time (i.e. for a time interval preceding each observed crash). A number of statistical tests will be applied to establish the ability of safety performance measures in identifying high risk driver-vehicle behaviour in the traffic stream. This high risk behaviour is expected to result in crashes for different road geometric and traffic scenarios. Research carried out thus far suggests that while behavoural microscopic traffic models may be unable to predict crashes they can replicate or identify high risk driving situations in the traffic stream. The usefulness of the microscopic simulation model is investigated with respect to two fundamental questions of safety: Which locations are inherently unsafe? and Which countermeasures have the highest potential for improving safety at these locations.