Travel time and incident risk assessment

Satu Innamaa, Eetu Pilli-Sihvola, Ilkka Norros

Research output: Book/ReportReportProfessional

Abstract

The goal of the project was to find methods for creating an accurate overall understanding of the current status of the transport system and to predict changes in traffic conditions. The most important task in achieving this was travel time prediction. Another goal was to find methods to assess the risk of incidents. In addition, an overall assessment of the monitoring system was performed considering specifically the needs of short-term prediction and incident risk assessment. Best practices were sought among other road operators, in the literature and from small data pilots. Existing applications in use by the Finnish Transport Agency were evaluated based on theory and practice found in literature and in data pilot studies. A slightly modified version of the dynRP travel time prediction model of the Danish Road Directorate was piloted on Ring I of the Helsinki Metropolitan Area. The main results showed that a 15-minute prediction model gave better travel time estimates than just using the latest measurement, especially in congested conditions. The model did not fulfil the threshold of keeping maximum errors between 10–25% prevalent in the literature. Nevertheless, if decisions must be made proactively, the use of this forecast, although not perfect, would lead to better decisions more often than when using just the latest measurement. Therefore the use of this model can be recommended. Furthermore, shorter-than-15 min prediction models provided more accurate estimates than the 15 min model. However, with the former, the latest measurement served better as an estimate, and the difference from the prediction model was small if any, or even negative. Therefore, the use of these shorter-term models cannot be recommended. The Dutch Rijkswaterstaat Traffic Management Centre has several procedures that can be considered best practice in incident risk assessment and management. It is recommended that a procedure be set up to systematically collect and use information on events that affect traffic. These annual forecasts should be studied in weekly meetings to detect abnormalities of traffic in the coming week and find solutions for (proactively) operating the traffic. The success of the previous week’s operations should also be evaluated in order to perform better next time. The annual traffic forecast can also be used in the planning of timing of road works. Incident data analysis for Ring I included road weather conditions in addition to traffic flow status information. The results indicate that some circumstances have higher incident risk than others, like evening rush hour, reduced visibility and moderate or abundant snowfall. However, the statistical significance of the results could not be studied here. This should be examined further with a larger dataset. Travel time is a reactive measure, as it can be measured only with delay. Therefore it is recommended that in areas with regular congestion, the traffic flow be monitored using sufficiently densely-spaced cross-section specific detectors capable of monitoring reliably at least the traffic volume and speed. In areas where regular congestion does not take place, traffic monitoring serves incident management and traffic information (e.g. media). In such areas, travel time monitoring would be sufficient to indicate the consequences of incidents and the level of congestions. The system could be supplemented by road user notifications.
Original languageEnglish
Place of PublicationHelsinki
PublisherFinnish Transport Agency
Number of pages74
ISBN (Electronic)978-952-255-327-0
Publication statusPublished - 2013
MoE publication typeD4 Published development or research report or study

Publication series

NameResearch reports of the Finnish Transport Agency
PublisherFinnish Transport Agency
No.31/2013
ISSN (Print)1798-6656
ISSN (Electronic)1798-6664

Fingerprint

Travel time
Risk assessment
Monitoring
Information use
Snow
Risk management
Visibility
Detectors
Planning

Cite this

Innamaa, S., Pilli-Sihvola, E., & Norros, I. (2013). Travel time and incident risk assessment. Helsinki: Finnish Transport Agency. Research Reports of the Finnish Transport Agency, No. 31/2013
Innamaa, Satu ; Pilli-Sihvola, Eetu ; Norros, Ilkka. / Travel time and incident risk assessment. Helsinki : Finnish Transport Agency, 2013. 74 p. (Research Reports of the Finnish Transport Agency; No. 31/2013).
@book{fa573a9700f7425b8f4ac8c672293c85,
title = "Travel time and incident risk assessment",
abstract = "The goal of the project was to find methods for creating an accurate overall understanding of the current status of the transport system and to predict changes in traffic conditions. The most important task in achieving this was travel time prediction. Another goal was to find methods to assess the risk of incidents. In addition, an overall assessment of the monitoring system was performed considering specifically the needs of short-term prediction and incident risk assessment. Best practices were sought among other road operators, in the literature and from small data pilots. Existing applications in use by the Finnish Transport Agency were evaluated based on theory and practice found in literature and in data pilot studies. A slightly modified version of the dynRP travel time prediction model of the Danish Road Directorate was piloted on Ring I of the Helsinki Metropolitan Area. The main results showed that a 15-minute prediction model gave better travel time estimates than just using the latest measurement, especially in congested conditions. The model did not fulfil the threshold of keeping maximum errors between 10–25{\%} prevalent in the literature. Nevertheless, if decisions must be made proactively, the use of this forecast, although not perfect, would lead to better decisions more often than when using just the latest measurement. Therefore the use of this model can be recommended. Furthermore, shorter-than-15 min prediction models provided more accurate estimates than the 15 min model. However, with the former, the latest measurement served better as an estimate, and the difference from the prediction model was small if any, or even negative. Therefore, the use of these shorter-term models cannot be recommended. The Dutch Rijkswaterstaat Traffic Management Centre has several procedures that can be considered best practice in incident risk assessment and management. It is recommended that a procedure be set up to systematically collect and use information on events that affect traffic. These annual forecasts should be studied in weekly meetings to detect abnormalities of traffic in the coming week and find solutions for (proactively) operating the traffic. The success of the previous week’s operations should also be evaluated in order to perform better next time. The annual traffic forecast can also be used in the planning of timing of road works. Incident data analysis for Ring I included road weather conditions in addition to traffic flow status information. The results indicate that some circumstances have higher incident risk than others, like evening rush hour, reduced visibility and moderate or abundant snowfall. However, the statistical significance of the results could not be studied here. This should be examined further with a larger dataset. Travel time is a reactive measure, as it can be measured only with delay. Therefore it is recommended that in areas with regular congestion, the traffic flow be monitored using sufficiently densely-spaced cross-section specific detectors capable of monitoring reliably at least the traffic volume and speed. In areas where regular congestion does not take place, traffic monitoring serves incident management and traffic information (e.g. media). In such areas, travel time monitoring would be sufficient to indicate the consequences of incidents and the level of congestions. The system could be supplemented by road user notifications.",
author = "Satu Innamaa and Eetu Pilli-Sihvola and Ilkka Norros",
year = "2013",
language = "English",
series = "Research reports of the Finnish Transport Agency",
publisher = "Finnish Transport Agency",
number = "31/2013",
address = "Finland",

}

Innamaa, S, Pilli-Sihvola, E & Norros, I 2013, Travel time and incident risk assessment. Research Reports of the Finnish Transport Agency, no. 31/2013, Finnish Transport Agency, Helsinki.

Travel time and incident risk assessment. / Innamaa, Satu; Pilli-Sihvola, Eetu; Norros, Ilkka.

Helsinki : Finnish Transport Agency, 2013. 74 p. (Research Reports of the Finnish Transport Agency; No. 31/2013).

Research output: Book/ReportReportProfessional

TY - BOOK

T1 - Travel time and incident risk assessment

AU - Innamaa, Satu

AU - Pilli-Sihvola, Eetu

AU - Norros, Ilkka

PY - 2013

Y1 - 2013

N2 - The goal of the project was to find methods for creating an accurate overall understanding of the current status of the transport system and to predict changes in traffic conditions. The most important task in achieving this was travel time prediction. Another goal was to find methods to assess the risk of incidents. In addition, an overall assessment of the monitoring system was performed considering specifically the needs of short-term prediction and incident risk assessment. Best practices were sought among other road operators, in the literature and from small data pilots. Existing applications in use by the Finnish Transport Agency were evaluated based on theory and practice found in literature and in data pilot studies. A slightly modified version of the dynRP travel time prediction model of the Danish Road Directorate was piloted on Ring I of the Helsinki Metropolitan Area. The main results showed that a 15-minute prediction model gave better travel time estimates than just using the latest measurement, especially in congested conditions. The model did not fulfil the threshold of keeping maximum errors between 10–25% prevalent in the literature. Nevertheless, if decisions must be made proactively, the use of this forecast, although not perfect, would lead to better decisions more often than when using just the latest measurement. Therefore the use of this model can be recommended. Furthermore, shorter-than-15 min prediction models provided more accurate estimates than the 15 min model. However, with the former, the latest measurement served better as an estimate, and the difference from the prediction model was small if any, or even negative. Therefore, the use of these shorter-term models cannot be recommended. The Dutch Rijkswaterstaat Traffic Management Centre has several procedures that can be considered best practice in incident risk assessment and management. It is recommended that a procedure be set up to systematically collect and use information on events that affect traffic. These annual forecasts should be studied in weekly meetings to detect abnormalities of traffic in the coming week and find solutions for (proactively) operating the traffic. The success of the previous week’s operations should also be evaluated in order to perform better next time. The annual traffic forecast can also be used in the planning of timing of road works. Incident data analysis for Ring I included road weather conditions in addition to traffic flow status information. The results indicate that some circumstances have higher incident risk than others, like evening rush hour, reduced visibility and moderate or abundant snowfall. However, the statistical significance of the results could not be studied here. This should be examined further with a larger dataset. Travel time is a reactive measure, as it can be measured only with delay. Therefore it is recommended that in areas with regular congestion, the traffic flow be monitored using sufficiently densely-spaced cross-section specific detectors capable of monitoring reliably at least the traffic volume and speed. In areas where regular congestion does not take place, traffic monitoring serves incident management and traffic information (e.g. media). In such areas, travel time monitoring would be sufficient to indicate the consequences of incidents and the level of congestions. The system could be supplemented by road user notifications.

AB - The goal of the project was to find methods for creating an accurate overall understanding of the current status of the transport system and to predict changes in traffic conditions. The most important task in achieving this was travel time prediction. Another goal was to find methods to assess the risk of incidents. In addition, an overall assessment of the monitoring system was performed considering specifically the needs of short-term prediction and incident risk assessment. Best practices were sought among other road operators, in the literature and from small data pilots. Existing applications in use by the Finnish Transport Agency were evaluated based on theory and practice found in literature and in data pilot studies. A slightly modified version of the dynRP travel time prediction model of the Danish Road Directorate was piloted on Ring I of the Helsinki Metropolitan Area. The main results showed that a 15-minute prediction model gave better travel time estimates than just using the latest measurement, especially in congested conditions. The model did not fulfil the threshold of keeping maximum errors between 10–25% prevalent in the literature. Nevertheless, if decisions must be made proactively, the use of this forecast, although not perfect, would lead to better decisions more often than when using just the latest measurement. Therefore the use of this model can be recommended. Furthermore, shorter-than-15 min prediction models provided more accurate estimates than the 15 min model. However, with the former, the latest measurement served better as an estimate, and the difference from the prediction model was small if any, or even negative. Therefore, the use of these shorter-term models cannot be recommended. The Dutch Rijkswaterstaat Traffic Management Centre has several procedures that can be considered best practice in incident risk assessment and management. It is recommended that a procedure be set up to systematically collect and use information on events that affect traffic. These annual forecasts should be studied in weekly meetings to detect abnormalities of traffic in the coming week and find solutions for (proactively) operating the traffic. The success of the previous week’s operations should also be evaluated in order to perform better next time. The annual traffic forecast can also be used in the planning of timing of road works. Incident data analysis for Ring I included road weather conditions in addition to traffic flow status information. The results indicate that some circumstances have higher incident risk than others, like evening rush hour, reduced visibility and moderate or abundant snowfall. However, the statistical significance of the results could not be studied here. This should be examined further with a larger dataset. Travel time is a reactive measure, as it can be measured only with delay. Therefore it is recommended that in areas with regular congestion, the traffic flow be monitored using sufficiently densely-spaced cross-section specific detectors capable of monitoring reliably at least the traffic volume and speed. In areas where regular congestion does not take place, traffic monitoring serves incident management and traffic information (e.g. media). In such areas, travel time monitoring would be sufficient to indicate the consequences of incidents and the level of congestions. The system could be supplemented by road user notifications.

M3 - Report

T3 - Research reports of the Finnish Transport Agency

BT - Travel time and incident risk assessment

PB - Finnish Transport Agency

CY - Helsinki

ER -

Innamaa S, Pilli-Sihvola E, Norros I. Travel time and incident risk assessment. Helsinki: Finnish Transport Agency, 2013. 74 p. (Research Reports of the Finnish Transport Agency; No. 31/2013).