Abstract
Traffic flow modeling and simulation is central to transportation system analysis. Existing research has been primarily focusing on cars, while trucks are overlooked or modeled separately from cars. Unfortunately, characteristics of freight demand, such as when and how trucks travel, and how truck flow interacts with car flow, are unclear. This becomes the main hurdle for improving truck mobility. This research aims at developing a holistic framework for mesoscopic traffic simulation that mixes both cars and trucks, by considering their interrelations simultaneously. The result includes the prediction of travel time, travel delay, vehicle-mile-traveled and emissions for both cars and trucks, at each road segment and intersection by time of day. Thus, potential traffic management strategies for both passenger cars and freight transportation can be evaluated and deployed.
This project is a continuation of the research from the Mobility21 project ‘Data-driven Network Models for Analyzing Multi-modal Transportation Systems’ in FY 2018 led by PI Qian. It further extends the data-driven multi-modal modeling on passenger transportation (cars and buses) to the one that integrates both passenger and freight transportation. While the former model is still being improved, the latter model is the focus of FY 2019 that will bring more potential deployment partners from governmental agencies and private trucking companies.
The expected outcome of this research is a framework of car-truck modeling in the regional transportation network, followed by a prototype web application that implements it using data of cars and trucks collected over many years in the state. The application also provides user interfaces to manage various scenarios of road closures/extensions and visualize the resultant system metrics for both cars and trucks. The simulation models and web application will be integrated into an open-source dynamic network analysis toolkit to test its effectiveness in the Philadelphia Metro network.
Description
Traffic flow modeling and simulation play a central role in dynamic transportation system analysis and management. Existing research has been primarily focusing on cars, while trucks are overlooked or modeled separately from cars. Unfortunately, characteristics of freight demand, such as when and how trucks travel, and how truck flow interacts with car flow, are unclear. This becomes the main hurdle for improving truck mobility. This research aims at developing a holistic framework for mesoscopic bi-modal traffic simulation, which considers the interaction of both private cars and trucks simultaneously. This research aims at developing a holistic framework for mesoscopic traffic simulation that mixes both cars and trucks, by considering their interrelations simultaneously. The result includes the prediction of travel time, travel delay, vehicle-mile-traveled and emissions for both cars and trucks, at each road segment and intersection by time of day. Thus, potential traffic management strategies for both passenger cars and freight transportation can be evaluated and deployed.
Traditional mesoscopic traffic simulation models perform simulation of a single vehicle class, namely standard passenger cars. It models the flow propagation using kinetic wave traffic flow model, which can achieve reasonable accuracy for large-scale regional networks. In traditional models, homogeneous passenger cars are assumed and first-in-first-out (FIFO) can be enforced within links. Arguably, such assumptions may not be realistic in the real-world settings. A more realistic setting is a network with multiple classes of vehicles, where traffic flow is a mixture of passenger cars and trucks on links and nodes at any time. For example a truck, which takes more road space and flow capacity, usually progresses in a lower free-flow speed than a standard passenger car. A passenger car usually takes over a truck when the traffic has low level of congestion. In our new mesoscopic traffic simulation framework, we specifically consider vehicle interrelations in this bi-modal network (note: we define bi-modal to be the mixture of passenger and freight transportation). The interactions among the two classes of flow is incorporated in a multi-class kinetic wave model. The multi-class flow model allows cars to overtake trucks when the total flow is in or close to free-flow, while the cross-class congestion shockwave can still be properly captured. The bi-modal modeling and simulation is critical to encapsulate the mobility of cars and trucks in a more realistic and accurate way. In addition, it also enables the utilization of large-scale data of cars and trucks collected over many years.
This project is a continuation of the research from the Mobility21 project ‘Data-driven Network Models for Analyzing Multi-modal Transportation Systems’ in FY 2018 led by PI Qian. It further extends the data-driven multi-modal modeling on passenger transportation (cars and buses) to the one that integrates both passenger and freight transportation. While the former model is still being improved, the latter model is the focus of FY 2019 that will bring more potential deployment partners from governmental agencies and private trucking companies.
In the FY 2018 project, we have built a sophisticated transportation network model for the Pittsburgh Metro Area and Philadelphia Metro Area that describes individual travel activities on transportation systems, all in standard passenger cars. Operational strategies and policies can therefore be examined in the network model in terms of system delay, crash risk, reliability, vehicle-miles traveled (VMT), fuel consumption and emissions. It was based upon a research project with PennDOT (Philadelphia traffic impact study) and one with the City of Pittsburgh (traffic impact study of Greenfield Bridge closure). Southwestern Pennsylvania Commissions is another potential deployment partner. A prototype web application that simulates traffic in the regional network and provides system performance metrics in high resolutions is developed, and initially tested by PennDOT District 6.
To build the framework of mesoscopic car-truck (bi-modal) traffic flow modeling and simulation, the following two fundamental questions will be tackled in this research,
1) How to estimate the dynamic traffic demands of private cars and trucks with traffic counts and speed data of both classes?
2) How can consider the interactions of private cars and trucks into our traffic flow model? What is their interrelation in terms of their respective travel time and choices?
This research primarily addresses car-truck traffic simulation and traffic states prediction, in the context of large-scale regional networks. The research approach is as follows.
First, we estimate the dynamic origin-destination (OD) demand of private cars and trucks using traffic counts and speed data. The traffic count data are obtained from PennDOT and contain both flow counts of private cars and trucks on selected road segments; the speed data are obtained from HERE/INRIX and contain the 5-min speed measurement for private cars and trucks, respectively. The start-of-the-art dynamic OD estimation method will be adopted to obtain the dynamic OD for every 15 minutes for both classes. The estimated demand will be used as input of the dynamic traffic flow models.
Second, we consider the heterogeneity of traffic flow where private cars and trucks are mixed, and develop a bi-modal traffic flow model. Some preliminary work has been done and published in a recent publication in Transportation Research Part B led by PI Qian. For modeling the heterogeneous traffic flow in general large networks, the model would pragmatically generalize the Cell Transmission Model to multi-class heterogeneous traffic flow. Each class of vehicles possesses identical vehicle attributes, which are encapsulated by a unique well-defined (least requirements usually include continuity and concavity) fundamental diagram. Within this general framework, we develop a generic, yet simple, class-specific capacity allocation and flux scheme to capture cross-class flow interrelations.
Third, we will develop a new dynamic regional network model (simulation framework) incorporating the bi-modal traffic flow. The DNL model uses the per-vehicle-class demand data from the first step and the bi-modal traffic flow model developed in the second step to compute the flow on each link and intersection, and produce the spatio-temporal distributions of each class of vehicles. Through the regional network model, we simulate individual cars and trucks in the reginal network, and model their route choices, travel time and mixed traffic flow conditions. The results include the travel time, travel delay, vehicle-mile-traveled and emissions, and can be presented and estimated at different scales, from each road segment and intersection by time of day, to the entire network of the peak hours. The new simulation framework will be implemented in a prototype web application using an open-source dynamic network analysis toolkit, Mobility Data Analytics Center - Prediction, Optimization, and Simulation (MAC-POSTS) developed by the Mobility Data Analytics Center. In addition, the web application will be further advanced to incorporate visualization of car and truck flow in the network, as well as the animation of traffic evolution over time of day. It will also provide interactive user interfaces to visualize model outputs.
Fourth, we examine our new simulation framework in two test cases, a real network of the Borough of McKees Rocks in the Pittsburgh Metro Area, and the Philadelphia Metro Area. We perform the regional network model under two scenarios, by adding additional truck flow to the McKees Rocks, and by closing several road segments along the I-95 near North of Downtown Philadelphia, as proof-of-concept experiments. The traffic impact can be measured by time-of-day performance metrics at both the street level and the regional level, such as total traffic delay, average travel time, emissions, energy use, vehicle-miles traveled, etc., all for both cars and trucks. Those results will again be added to the web application for visualization and animation.
The expected outcome of this research is a framework of car-truck modeling in the regional transportation network, followed by a prototype web application that implements it using data of cars and trucks collected over many years in the state. The application also provides user interfaces to manage various scenarios of road closures/extensions and visualize the resultant system metrics for both cars and trucks. The simulation models and web application will be integrated into an open-source dynamic network analysis toolkit to test its effectiveness in the Pittsburgh and Philadelphia Metro network.
Upon the completion of this project, we plan to actively seek both industrial and federal funding based on this initial development. Our framework is applicable to any large traffic networks with private cars and trucks. This generality will attract the attentions from various agencies and private trucking companies. Potential funding agencies/collaborators include the Department of Transportation, Federal Highway Administration, National Science Foundation, and National Institute of Standards and Technology, MPOs and local trucking companies.
Timeline
1 Develop multiclass vehicle demand estimation methods: 2 months
2 Bi-model traffic flow model: 3 months
3 Dynamic network loading model implementation: 3 months
4 Test on real networks and policy implications: 4 months
Strategic Description / RD&T
Deployment Plan
We will use the web application developed in this project to simulate traffic evolution in the roadway network of the Borough of McKees Rocks in the Great Pittsburgh Area, as well as for the network of Philadelphia Metro Area. The output will provide the spatio-temporal distributions of both private cars and trucks, and hence we can obtain the travel time, travel delay, vehicle-mile-traveled and emissions for each road segment and intersection by time of day for each vehicle class. We will work closely with the Borough of McKees Rocks, DVRPC and PennDOT to seek their feedback, improve the web application and help their decision making.
Upon the completion of this project, we plan to actively seek both industrial and federal funding based on this initial development. Our framework is applicable to any large traffic networks with private cars and trucks. This generality will attract the attentions from various agencies and private trucking companies. Potential funding agencies/collaborators include the Department of Transportation, Federal Highway Administration, National Science Foundation, and National Institute of Standards and Technology, MPOs and local trucking companies.
Expected Outcomes/Impacts
The expected outcome of this research is a framework of car-truck modeling in the regional transportation network, followed by a prototype web application that implements it using data of cars and trucks collected over many years in the state. The application also provides user interfaces to manage various scenarios of road closures/extensions and visualize the resultant system metrics for both cars and trucks. The simulation models and web application will be integrated into an open-source dynamic network analysis toolkit to test its effectiveness in the Pittsburgh and Philadelphia Metro network.
Expected Outputs
TRID
Individuals Involved
Email |
Name |
Affiliation |
Role |
Position |
matthew.battifarano@gmail.com |
Battifarano, Matt |
Carnegie Mellon University |
Other |
Other |
weima@cmu.edu |
Ma, Wei |
Carnegie Mellon University |
Other |
Student - PhD |
seanqian@cmu.edu |
Qian, Sean |
Carnegie Mellon University |
PI |
Faculty - Untenured, Tenure Track |
Budget
Amount of UTC Funds Awarded
$90000.00
Total Project Budget (from all funding sources)
$100197.00
Documents
Type |
Name |
Uploaded |
Data Management Plan |
dmp_FiY4Ynh.docx |
Jan. 12, 2018, 1:16 p.m. |
Publication |
N/A |
Sept. 23, 2018, 7:35 p.m. |
Presentation |
Car-truck simulation for Pittsburgh region |
Sept. 23, 2018, 7:35 p.m. |
Presentation |
Car-truck simulation for Pittsburgh region |
Sept. 23, 2018, 7:35 p.m. |
Presentation |
Rethink O-D estimation |
Sept. 23, 2018, 7:35 p.m. |
Progress Report |
176_Progress_Report_2018-09-30 |
Sept. 23, 2018, 7:35 p.m. |
Presentation |
Car-truck simulation for Pittsburgh region |
March 24, 2019, 9:36 p.m. |
Progress Report |
176_Progress_Report_2019-03-30 |
March 24, 2019, 9:36 p.m. |
Final Report |
176_-_Final_Report.pdf |
June 3, 2019, 8:45 a.m. |
Publication |
Large-Scale Mesoscopic Network Modeling with Cars and Trucks: A Case Study in Pittsburgh |
Dec. 2, 2020, 9:23 a.m. |
Match Sources
No match sources!
Partners
Name |
Type |
PennDOT |
Deployment Partner Deployment Partner |
DVRPC |
Deployment Partner Deployment Partner |