skip to main content

<< Return to Webinar Files

Webinar Presentation

HTML version of the presentation
Image descriptions are contained in brackets. [ ]

A Digital Twin for Traffic Monitoring and Proactive Incident Management
(May 11, 2021)

Developing a Digital Twin for Traffic Monitoring
Presenter: Xing Fu, PhD Candidate
Presenter’s Org: College of Engineering, University of Alabama

T3 and T3e webinars are brought to you by the Intelligent Transportation Systems (ITS) Professional Capacity Building (PCB) Program of the U.S. Department of Transportation (USDOT)’s ITS Joint Program Office (JPO). References in this webinar to any specific commercial products, processes, or services, or the use of any trade, firm, or corporation name is for the information and convenience of the public, and does not constitute endorsement, recommendation, or favoring by the USDOT.


[The slides in this presentation contain the University of Alabama’s logo.]

Slide 1: Student Presentation 1

Title: Developing a Digital Twin for Traffic Monitoring

Student: Xing Fu, PhD Candidate

Advisors: Jun Liu, PhD, Assistant Professor; Alex Hainen, PhD, Associate Professor

Motivations:

  • Sensors and cameras are not available at all sites
  • There are barriers to track all vehicles in the network

Digital Twin—A live representation of traffic flow in network

  • The goal of the research is to develop a real‐time traffic simulation with live traffic sensor and signal data
  • Applications include network‐wide traffic monitoring/management, incident management, and network‐wide performance evaluation (mobility, safety, and environment)

[This slide contains two images: (1) an aerial photo of an intersection and (2) a computer‐generated image of what appears to be the same intersection.]

Slide 2: Developing a Digital Twin for Traffic Monitoring

  • Research Background & Motivation

References:

[This slide contains three images: (1) a multilane road with a bike lane with a caption of “Limited Sensors.” There are rectangles on the road designating where the sensors can read. (2) a screenshot of a computer request for use your current location with a caption of “Privacy Concerns.” (3) a digital drawing of a book with the word “Law” on the spine and a judge’s gavel on top of the book with a caption of “Legal Issues.”]

Slide 3: Research Objective

  • Develop a traffic simulation model to represent the real‐world traffic patterns in a near real‐time manner

[This slide contains two images: (1) an aerial view of an intersection of two multilane roads and (2) a computer‐generated image of an intersection of two multilane roads.]

Slide 4: An open‐source microscopic traffic simulation platform

https://gfycat.com/discover/traffic-simulation-gifs

[This slide contains three images: (1) the logo of SUMO open source microscopic traffic simulation platform, (2) a computer‐generated image of a multilane road, and (3) a zoomed out computer‐generated aerial view of an intersection.]

Slide 5: SUMO Inputs and Outputs

Inputs

  • Network
  • Detector
  • Signal Events & Traffic Counts
  • Vehicle info

Outputs

  • Simulation
  • Vehicle Trajectories

[This slide contains six images: (1) an aerial photo of a developed area with a road highlighted with the word “network” below it, (2) an aerial view of a computer‐generated intersection with the word “detector” below it, (3) a zoomed in aerial view of a computer‐generated intersection with the word “simulation” beside it, (4) a table of signal events and traffic counts with the words “signal events & traffic counts” below it, (5) an overhead view of an region with several roads highlighted to show vehicle trajectories, and (6) a digital drawing showing six different types of cars and trucks with the words vehicle info below it.]

Slide 6: Challenges

  • Challenge 1‐ SUMO
    • Developing real‐time simulation in SUMO to capture live traffic dynamics
  • Challenge 2‐ Traffic flow
    • Reconstructing network‐wide and short‐term traffic flows based on short‐term traffic data (e.g. 5min, 10min)
  • Challenge 3‐ Traffic signal
    • Reflecting the traffic signals that control the traffic flows at intersections

https://en.wikipedia.org/wiki/Intersection_(road)

https://medium.com/yonohub/sumo-a-traffic-simulator-over-the-cloud-with-yonohub-f2bbf7f62990

[This slide contains five images: (1) an aerial view of a computer‐generated intersection, (2) a digital drawing showing a laptop with a line connecting it to a signal box, (3) a digital map of Tuscaloosa, Alabama, (4) a computer‐generated intersection with traffic pattern highlights, and (5) a table of 16 intersection diagrams with travel directions displayed in different colors.]

Slide 7: Estimation of Network Flows/Vehicle Routes

Inputs:

  • Traffic detector data (counts per min or hour)

Outputs:

  • Origins and destinations (of vehicle trips)
  • Vehicle routes (chain of road links/segments)

Methods:

  • Origin‐destination (OD) Demand Estimation
    • Gravity Model
    • Maximum likelihood method
    • Statistical methods
  • Vehicle Route Estimation
    • Kalman filter
    • Optimization methods
    • Maximum entropy model

[This slide contains two images: (1) a map of an area with lines connecting to many locations on the map and (2) a digital map interface with multiple colored circles each with a directional arrow in the center showing a driving route from one start point to an endpoint.]

Slide 8: Current Practices in SUMO

What do we need in this project?

  1. A reasonable OD demand generation method
  2. Keep the balance between the accuracy and efficiency
  3. Reduce the time‐lag between the traffic and simulation

https://realtimeapi.io/category/realtime-data/

https://en.wikipedia.org/wiki/Dijkstra%27s_algorithm#/media/File:DijkstraDemo.gif

[This slide contains three images: (1) an aerial photo of two parallel multilane roads overlaid with computer‐generated images of two multilane roads with the caption “Trip OD,” (2) an aerial photo of an intersection with many lines connecting different points in the photo with the caption “Shortest Path in Network,” and (3) an aerial photo of an intersection overlaid with a computer‐generated image of an intersection and an image of a laptop with a line connecting it to a signal box with the caption “Data Feed.”]

Slide 9: Our SUMO Simulation Method

Learning the network regarding all possible OD pairs and all shortest paths between the ODs
Possible Routes (no flow)
Assigning the traffic flow to possible routes and try to minimize the error between the assigned flow and the detector counts
Vehicle Route Flows
Real‐time simulation in SUMO with short‐term data feeds

[This slide contains three images: (1) a clipart image of two right angles, (2) a clipart image of a funnel, and (3) a clipart image of a square within a square with many dots around the edges of the outer square.]

Slide 10: Mathematicallyā€¦

Learning the network regarding all possible OD pairs and all shortest paths between the ODs.

  1. Search the shortest route of each OD pair (Dijkstra’s algorithm)
  2. Build the relationship matrix of traffic routes and roads

The value of š¯‘´(š¯‘¯,š¯’¸) is 1 if road š¯‘› is on the route š¯‘¯.

[This slide contains three images: (1) a digital drawing of two separate lines with starting points that converge in the middle, (2) an image of many interconnected lines, and (3) an image of a mathematical formula relating a road to a route.]

Slide 11: Mathematicallyā€¦

Assigning the traffic flow to possible routes and try to minimize the error between the assigned flow and the detector counts.

Control the Vehicle Inputs

minΣι∈Ι(errι(t))2 minimize the total estimation error
The detector data in the network input roads are used to ensure that conservation of total number of vehicles Σρ∈ΡΜρ,ιΧρ(τ) = ƒlowι(τ) ∀ι∈ΙΑ
The detectors in the network non‐input roads are used to optimize the distribution of route flows Σρ∈ΡΜρ,ιΧρ(τ) + errι(τ) = ƒlowι(τ)
∀ι∈ΙΒ
χρ(τ)≥0 ∀ρ∈Ρ
errι(τ)∈ℜ ∀ι∈Ι*
ΙΑ∪ΙΒ=Ι*

Minimized the Error within the Network

[This slide contains two images: (1) a clipart image of a funnel and (2) an image of a grid of roads with arrows showing direction and images of cars.]

Slide 12: Real‐time simulation in SUMO with Short‐term data feeds

[This slide contains a flowchart showing the process of getting from Map/Network Topology and Traffic Cameras, Radars, Detectors to Python through the use of network processing, real‐time data input, and the proposed model for route flow.]

Slide 13: Traffic Signal Configuration

  • Load traffic signal events
  • Generate phase strings with timestamps
  • Replay traffic signal in SUMO according to phase strings

[This slide contains a flowchart showing the process of getting from Traffic Light Events and Permissive Control File to Update Signal States.]

Slide 14: Example Simulation Inputs

  • Input Files
    • Network: AL‐69 S in Tuscaloosa, AL
    • Detectors
    • Detector Traffic Counts: 2020‐02‐05 06:00:00–10:00:00
    • Signal: Signal Events
  • Vehicle Route Flow Estimation
    • SUMO Methods
      • DFROUTER
      • FLOWROUTER
      • ROUTESAMPLER
    • Our Method

[This slide contains two images: (1) a computer‐generated image of an intersection with the crossroads highlighted and (2) an aerial photograph of a main road with roads perpendicularly intersecting it.]

Slide 15: Method Comparisons (non‐real‐time)

  • Root Mean Square Error (RMSE)
Method DFROUTER FLOWROUTER ROUTESAMPLER Proposed Method
RMSE 98.42 64.56 50.08 27.72
  • Absolute Percentage Error Distribution

[This slide contains four graphs: (1) a bar graphs showing the Absolute Percentage Error Distribution of the DFROUTER, (2) a bar graphs showing the Absolute Percentage Error Distribution of the FLOWROUTER, (3) a bar graph showing the Absolute Percentage Error Distribution of the ROUTESAMPLER, and (4) University of Alabama’s method.]

Slide 16: Method Comparison

Features SUMO Methods Proposed Method
DFROUTER FLOWROUTER JTCROUTER ROUTESAMPLER
Input Detector information;
Observed edge flow;
Sumo network
Detector information;
Observed edge flow;
Sumo network
Turn‐count data;
Sumo network
Route file;
Turn‐count data;
Sumo network
Detector information;
Observed edge flow;
Sumo network
Output Route;
Vehicles;
The route choice probabilities of vehicles
Routes;
Route flow
Routes;
Vehicles
Routes;
Vehicles
Routes;
Route flow
Optimization Function
Flow Conservation
Self‐Searching Routes
The Topology Extraction of Complex Networks
Computation Cost 30% 60% 30% 30% 30%

Slide 17: Simulation Overview

[This slide contains an aerial photograph of a main road with roads perpendicularly intersecting it.]

Slide 18: More Simulation Details

[This slide contains three images: (1) an aerial photo of an intersection overlaid with a computer‐generated image of an intersection with a title “Real‐time Simulation,” (2) a screenshot of many lines of Python code, and (3) a computer‐generated image of a road with two representative vehicles with the title “Detector Simulation Error.”]

Slide 19: Next Step

  • OD Constraints
    • Use regional OD information (Streetlight) to improve the demand estimation
  • Event Response
    • Dynamically adjusting vehicle routes in response to traffic events
  • Implementation on Large Network
    • Expand the road network and the implement the simulation

https://www.urban-control.com/our-traffic-control-sensors/

[This slide contains three images: (1) an image of a map with two roads highlighted, (2) a photo of a multilane road with a car crash, and (3) an image of a map with a driving route in a city highlighted.]

↑ Return to top

SUPPORT

Technical Assistance is available to Federal, State and local transportation agencies through:

ITS Peer Program - The ITS Peer-to-Peer Program puts you in touch with technical experts or experienced peers.

ITS Help Line - The ITS Help Line provides technical support by email or telephone at 866-367-7487.

STAY CONNECTED

go to twitter    go to Facebook    go to Instagram    go to Linkedin    go to email