To increase the speed and efficiency of a post traffic incident response, numerous highway-operating agencies have developed and implemented Traffic Incident Management (TIM) programs. When functioning properly, a TIM should inform an agency of an impact, calculate the optimal response, and ensure delay to the public is minimized. In collaboration with a research team from the University of Maryland’s A. James Clark School of Engineering, the MDOT SHA Office of Policy and Research a TIM system on Interstate 95.
This first Incident Duration Prediction Module (IDPM) was designed to assess a given incident and the number of vehicles involved, evaluating among other factors the likely clearance time. The success of the IDPM-I-95 led to work on a Transferability Assessment Method (TAM), allowing data and prediction tools designed for I-95 to design new IDPMs for other Interstates and major routes in Maryland.[i] The initial IDPM was tested in a corridor with major urban centers, but it can also benefit more rural corridors, particularly those with heavier use or higher risk of incidents.
The IDPMs fall under the aegis of a prototype TIM system called the Decision Support Tool (DST) software, intended to provide real-time response and incident management data. The original IDPM-I-95 was labeled DST-1, with the subsequent expansion of coverage to the four new IDPMs accordingly labeled as DST-2.[ii] The development of the TAM is of particular importance to the continued evolution of these devices, as previous methods of training proved far more time and resource intensive. Each of the four new IDPMs served a congested highway with distinct traffic and incident patterns, with the tools collectively encompassing two types of beltways, a typical commuting freeway, and a major expressway. Should TAM enable greater generalizability of TIM systems, responsible highway agencies will be able to account for a lack of data or quality data and improve performance.[iii]
When deployed, the IDPMs assessed incident clearance time and the resulting traffic impacts with an average prediction accuracy rate for I-495 of 82.14%, I-695 of 82.82%, I-70 of 81.67%, and US 29 of 77.27%. These figures can be compared to provided rates for the I-95 IDPM of 74.3%. In all cases, the testing datasets showed worse accuracy than with earlier training datasets, typically by low-to-mid single digits barring the case of US 29 where the difference was 16%.[iv] These predictions are likely to become increasingly accurate with advances in sensory equipment, computing, and design, and often are shaped by changing information in real-time just as would be a response without TIM assistance. MDOT’s system includes an estimated incident severity score that actively shifts as traffic management center operators collect new information. When compared to the null alternative of no technological assistance in incident response, accuracy in the range of 80% offers a clear and present advantage with tangible benefits to both safety and congestion.[v]
Human input remains a necessity to ensure systems are functional and correct for discovered inefficiencies. However, a stated long-term goal of future project research is to automate incident duration calculation, better increasing response time in a manner beneficial to both persons involving in traffic incidents and the commuting public.[vi] An additional goal is to account for the widespread lack of resources for highway sensor deployment, operations, and maintenance, a hindrance to the efficiency of TIM and other highway-adjacent systems that may warrant development of new technologies such as sensor-bearing drones.[vii]
The system is built for efficient performance and easy user interface, flexibly incorporating engineers’ knowledge as supplemental information, and modularly integrating with other existing systems. Its calculations are done with the maximum level of transparency, and it is designed to be capable of making predictions with minimal data and accommodating deficiencies in data’s quality and precision.[viii] In no circumstances is flawed or partial data preferable, but complete data are a luxury not always readily available. This problem is the central conceit of the TAM and the premise of future work to create a broadly generalizable IDPM for all Maryland highways, urban and rural, even where a lack of data or resources would have previously hindered deployment. As clearance operations conducted by the same incident response agency are expected to possess several common traits even when occurring on different highways, classification rules are expected to be transferable between IDPMs so long as they meet requisite confidence and support thresholds based on incident records.[ix]
Looking beyond MDOT’s Coordinated Highways Action Response Team (CHART) operations, similar transferability methods could apply to areas served by other response teams. Differences in methodologies presently require jurisdictions to develop their own tools, though improving transferability and data-sharing may change approaches in the long-term. The developed steps and methodologies of TAM and the lessons learned from IDPM may prove an asset to state- and local-level agencies pursuing their own TIM systems, with the MDOT SHA OPR and its partnered research team a potential resource in doing so.[x]
This research team effort has so far been implemented in a pilot, research context. As a result, resources required to deploy it widely are not yet known.
[i] Chang, G.-L., Huang, Y.-L., Lu, Y.-C., & Lin, Y.-T. (2021). “DEVELOPMENT OF A TRAFFIC MANAGEMENT DECISION SUPPORT TOOL (DST) FOR FREEWAY INCIDENT TRAFFIC MANAGEMENT PLAN DEVELOPMENT.” Maryland Department of Transportation State Highway Administration, https://www.roads.maryland.gov/OPR_Research/MD-21-SHAMD5-32_FITM-II_Report.pdf
[ii] Chang, G.-L., Huang, Y.-L., Lu, Y.-C., & Lin, Y.-T. (2021).
[iii] Chang, G.-L., Huang, Y.-L., Lu, Y.-C., & Lin, Y.-T. (2021).
[iv] Chang, G.-L., Huang, Y.-L., Lu, Y.-C., & Lin, Y.-T. (2021).
[v] Chang, G.-L., Huang, Y.-L., Lu, Y.-C., & Lin, Y.-T. (2021).
[vi] Personal communication with Dr. Gang-Len Chang, July 2022.
[vii] Chang, G.-L., Huang, Y.-L., Lu, Y.-C., & Lin, Y.-T. (2021).
[viii] Personal communication with Dr. Gang-Len Chang, July 2022.
[ix] Chang, G.-L., Huang, Y.-L., Lu, Y.-C., & Lin, Y.-T. (2021).
[x] Personal communication with Dr. Gang-Len Chang, July 2022.
This report was delivered to the U.S. Department of Transportation in 2023. It was primarily authored by NADO Associate Director Carrie Kissel and NADO Research Fellow Danny Tomares. Many transportation agency staff and others assisted with this project in a variety of ways. We offer deep and heartfelt thanks to all the individuals who have provided information and images, consented to be interviewed, and offered editorial guidance in support of this research. Any opinions, findings and conclusions, or recommendations expressed in this publication are those of the authors and do not necessarily reflect the views of U.S. DOT or the NADO Research Foundation.
To read more about the NADO Research Foundation’s ITS case studies, follow this link.