Andrzej P. Tarko

HETEROGENEITY CONSIDERATIONS IN ACCIDENT MODELING

Panel data sets are becoming readily available and increasingly popular in safety research. Despite its advantages, panel data raises new specification issues, the most important of which is heterogeneity, which have not been addressed in previous studies in the safety area. Based on a county accident data set, the present analysis extends prior research in a significant direction. There is an explicit effort to control for cross-section heterogeneity that may otherwise seriously bias the resulting estimates and invalidate statistical tests. Because common modeling techniques such as the fixed and random effects models, developed to account for heterogeneity, are impractical for count data, this study uses cluster analysis to overcome this. First, observations are disaggregated into homogeneous clusters. Then, separate negative binomial models including a time trend factor are developed for each group. The results clearly indicate that there are significant differences between the models developed, and that separate models describe data more efficiently than the joint model.

Toward a class of link travel time functions for dynamic assignment models on signalized networks

This paper investigates time-dependent travel time functions for dynamic assignment on signalized arterial network links. Dynamic link travel times are first classified according to various applications. Subsequently, stochastic and deterministic travel time functions for longer and shorter time horizons are discussed separately, and two sets of functions are recommended for dynamic transportation network problems. The implications of those functional forms are analyzed and some modifications for dynamic network models are suggested. In addition, based on dynamic link travel time functions, we discuss how many independent variables are necessary to describe the temporal traffic flow and properly estimate the time-dependent travel time and flow propagation over an arterial link. As a result, six link flow variables and corresponding link state and flow propagation equations are proposed as the basis to formulate dynamic transportation network models.

METHODOLOGY FOR IDENTIFYING HIGHWAY SAFETY PROBLEM AREAS

Safety programs can provide benefits only if they apply effective countermeasures to the locations and areas that really need safety treatment. A methodology of areawide safety analyses is presented to detect these areas (states, counties, townships, etc.) that should be considered for safety treatment. The method is implemented for Indiana at the county level. The method uses regression models to estimate the normal number of crashes in individual counties. The above-norm numbers of crashes calculated for several past years are used to predict the above-norm value for the future year when the safety treatment is to be applied. The counties are priority ranked using the combined criterion including both the above-norm number of crashes and the confidence level. Above-norm crashes represent the magnitude of the problem—specifically, how many crashes could be avoided if the safety level was normal. A confidence level represents the chance that the observed excessive number of crashes is not just a random ef...

Accident prediction models for signalized crosswalks

In Poland, there is an obvious need for effective countermeasures to elevate traffic safety, especially pedestrian safety. The paper presents results of study of pedestrian safety on signalized crosswalks. A regression technique was used in the investigation of factors-accidents relationships and, then, the accident predictive models have been developed. The models have been used to highlight and quantify the effects of traffic and signals impacts on the pedestrian safety. The study has shown that short-term measurements carried on specially for a purpose of safety analysis or these conducted by traffic authorities on a regular basis provide sufficient data base for developing accident regression models. The method of accident prediction applicable in non-stationary conditions faced in Poland and other East European countries, has been proposed in the paper.

Using Imaging Technology to Evaluate Highway Safety

Crash-based safety analysis is set back by several shortcomings such as randomness and rarity of crash occurrences, lack of timeliness, and inconsistency in crash reporting. Non-crash-based safety analysis has been around for more than three decades but its potential was limited due to a difficulty in the data collection and the evaluation. Recent advancement in digital videos and image detection technology renewed interest in facilitating the data collection and improving the evaluation method. Two image detection systems for the measurement of traffic characteristics were evaluated: (a) a commercial video detection system and (b) proprietary image processing software. The measurement evaluation revealed that both systems were still not sufficiently accurate for the safety evaluation purpose and thus a manual measurement from digitized video clips was preferred for a collection of evaluation data. The authors proposed a novel application of extreme value theory to safety evaluation based on observable traffic characteristics. The proposed method was evaluated by applying to right-angle collisions at signalized intersections. A traffic characteristic so-called post-encroachment time (PET) was collected at selected intersections as a surrogate safety measure. Based on PET characteristics, risk and frequency of right-angle crashes at the studied intersection or individual conflict zone can be estimated using only the data collected at the location. For comparison, a traditional approach to safety analysis using Poisson and negative binomial regression analyses was also examined. Both evaluation methods – extreme value approach and regression – indicate a significant relationship between PETs and historical crash data. Simulation experiments were carried out to examine the effect of observation period on a variance of estimates obtained the proposed method. Advantages and problems with the proposed method are described in this study. A simple method for an evaluation of the risk of right-angle collisions at signalized intersections is also provided in the appendix.

Development of a Portable Video Detection System for Counting Turning Vehicles at Intersections

This research was conducted to test the feasibility of using existing video detection techniques for counting turning volumes with a portable installation. This was accomplished by integrating a forty-five foot mechanical tower mounted on a van with two video detection systems, Autoscope and VideoTrak. The research project has produced results in three categories: prototype methods of counting turning volumes, evaluation results, and general specifications of a portable video-based system for counting vehicles at intersections. The method based on spot detection uses redundancy of data to improve the results quality. The method VideoTrak one dimensional tracking classifies maneuvers based on the location where vehicles enter and exit a tracking strip. Both the evaluated methods in their current versions do not meet the accuracy expectations expressed by the INDOT representatives. The general system specifications were developed to help build a prototype unit. The biggest challenge is the structure of the system that has to be portable, stable during data collection, and protected against tempering with. The report advises postponing building a prototype system by the time needed to develop satisfactory image processing and interpretation software for identifying vehicle maneuvers at intersections.

Congestion Detection in Large-Scale Advanced Traveler Information Systems

Abstract A new method of data integration to detect congestion in signalized urban networks is presented. The method uses imprecise knowledge within the framework of fuzzy operator logic and the modified Dempster-Shafer rule of combination. The results of the method evaluation indicate that the amount of congestion information produced by the algorithm increases several times after the intelligent processing is applied to sparse data. The developed algorithm can be also used to reduce amount of source data required.