Strategies to Identify and Mitigate Secondary Crashes in Real-time

Authors

Angela KitaliFollow

Document Type

Dissertation

Degree

Doctor of Philosophy (PhD)

Major/Program

Civil Engineering

First Advisor's Name

Priyanka Alluri

First Advisor's Committee Title

Committee chair

Second Advisor's Name

Albert Gan

Second Advisor's Committee Title

Committee member

Third Advisor's Name

Mohammed Hadi

Third Advisor's Committee Title

Committee member

Fourth Advisor's Name

Xia Jin

Fourth Advisor's Committee Title

Committee member

Fifth Advisor's Name

Wensong Wu

Fifth Advisor's Committee Title

Committee member

Sixth Advisor's Name

Thobias Sando

Sixth Advisor's Committee Title

Committee member

Keywords

Secondary crashes, Incident spatiotemporal impact area, Penalized likelihood model, Imbalanced data, Dynamic binary classification model, Prevailing traffic data, Real-time weather data, Tertiary crashes

Date of Defense

11-13-2020

Abstract

Traffic incidents are the primary source of non-recurring congestion. In addition to affecting roadway operations, traffic congestion resulting from an incident exposes other vehicles to the risk of being involved in additional incidents, typically referred to as secondary crashes. Secondary crashes adversely affect traffic operations and impose risk on the safety of both road users and incident responders. Transportation agencies have been looking for ways to mitigate secondary crashes. However, secondary crash mitigation has several challenges. The length of the queue caused by an initial incident and the amount of time this queue lasts on the road varies, depending on the characteristics of the respective incident. Since identifying potential secondary crashes is difficult, investigating the factors that may influence these crashes becomes even more challenging. Moreover, the limited knowledge of what constitutes a secondary crash and its contributing factors largely impede mitigation strategies.

The goal of this research was to investigate approaches to mitigate secondary crashes on freeways. To achieve this goal, a readily implementable data-driven approach to identify secondary crashes in real-time was developed. This approach is more accurate in identifying secondary crashes since it better reflects the changes in traffic characteristics caused by the primary incident. Following the identification of secondary crashes, the next step involved developing a secondary crash likelihood model. This model established the relationship between the likelihood of secondary crashes and influential factors, i.e., incident characteristics, traffic flow attributes, temporal attributes, presence of work zone, and other geometric attributes. The model results indicate that the presence of work zones significantly influenced the occurrence of secondary crashes. Overall, as expected, roadway geometric, temporal, traffic flow, incident, and weather attributes were found to influence secondary crashes.

The probabilistic relationship between factors that influence the risk of cascading crashes was also explored. Crashes are termed as “cascading” when the subsequent secondary crashes occur within the impact area of the prior secondary crashes and the primary incident. Cascading crashes were found to be most likely to occur when traffic is in the transition state, i.e., when there is a platoon of vehicles traveling at high differential speeds.

Once an incident has occurred, traffic conditions upstream of the incident change with time, and so does the likelihood of secondary crashes. The likelihood model was implemented to dynamically predict the risk of a secondary crash in real-time. The proposed model accounts for the temporal variation of prevailing conditions that influence the likelihood of secondary crashes. This model could be used to develop an Advanced Traffic Management System (ATMS) to proactively prevent secondary crashes. Through this system, first responders will be more vigilant and better prepared in case secondary crashes occur. In addition, motorists upstream of the primary incident could be warned about the potential for secondary crashes.

Identifier

FIDC009240

ORCID

0000-0002-1962-162X

Previously Published In

Refereed Journal Papers

1. Kitali, A., Alluri, P., Sando, T., and Wu, W. (2019). “Identification of Secondary Crash Risk Factors using Penalized Logistic Regression Model,” Transportation Research Record: Journal of the Transportation Research Board, 2673(11), 901–914.
2. Kitali, A., Alluri, P., and Sando, T. (2019). “Impact of Primary Incident Spatiotemporal Influence Thresholds on the Detection of Secondary Crashes,” Transportation Research Record: Journal of the Transportation Research Board, 2673(10), 271–283.
3. Kitali, A., Alluri, P., Sando, T., Haule, H., Kidando, E., and Lentz, R. (2018). “Likelihood Estimation of Secondary Crashes Using Bayesian Complementary Log-Log Model,” Accident Analysis and Prevention, 119, 58–67.

Full-Paper Refereed Conference Proceedings

1. Kitali, A., Alluri, P., and Sando, T. “Influence of Incident Spatiotemporal Estimation Method in Secondary Crash Identification, Proceedings of the 100th Annual Meeting of the Transportation Research Board, Washington, DC.
2. Kitali, A., Alluri, P., Sando, T., and Wu, W. (2019). “Identification of Secondary Crash Risk Factors using Penalized Logistic Regression Model,” Proceedings of the 98th Annual Meeting of the Transportation Research Board, Washington, D.C.
3. Kitali, A., Alluri, P., and Sando, T. (2019). “Impact of Primary Incident Spatiotemporal Influence Thresholds on the Detection of Secondary Crashes,” Proceedings of the 98th Annual Meeting of the Transportation Research Board, Washington, D.C.

Recommended Citation

Kitali, Angela, "Strategies to Identify and Mitigate Secondary Crashes in Real-time" (2020). FIU Electronic Theses and Dissertations. 4547.
https://digitalcommons.fiu.edu/etd/4547