<< Return to Webinar Files

Webinar Presentation

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

Raising Awareness of Artificial Intelligence for Transportation System Management and Operations
(February 13, 2020)

Raising Awareness of AI Applications in Transportation System Management and Operations
Presenter: Peter Huang, Ph.D., P.E.
Presenter’s Org: Federal Highway Administration (FHWA)

T3 webinars are brought to you by the Intelligent Transportation Systems (ITS) Professional Capacity Building (PCB) Program of the 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.


Slide 1: Raising Awareness of AI Applications in Transportation System Management and Operations

Peter Huang Ph.D., P.E.
Office of Operations R&D, FHWA, TFHRC,
Peter.Huang@dot.gov

[This slide contains a photo of a large grassy lawn with a three-story building in the background.]

Slide 2: Why AI Now?

AI applications should be built on the foundation of existing ITS infrastructure;
And it is the time to do it!

  • AI
    AI applications for higher level intelligent operations, predictive data analysis, knowledge base, learning from successful experience, get “smarter over time”
  • ITS facility & operations
    Decision support programs, TMC operations, ITS planning and operations
  • Data collection/processing
    Central computers, data base, data storage, etc.
  • Field devices and infrastructure
    Advanced sensors and field devices, communications, and TMC buildings, etc.

[This slide contains an image of a four-level pyramid with each level labeled and with supplemental information to the right of each level. This information in the pyramid and the supplemental information is reproduced above.]

Slide 3: Why AI Now?

AI technology is mature and advanced

In 1997 IBM DeepBlue Computer won the Chess game over the world champion!

In 2016, Google’s AlphaGo defeated World Champion (Go Chess) Lee Sedol!

[This slide contains two images: (1) a photo of the 1997 chess game where the IBM DeepBlue computer beat the world champion and (2) a photo from the 2016 chess game where Google’s AlphaGo computer defeated the world champion.]

Slide 4: Why AI Now?

Another Example: The Jeopardy! Game;
It was a very tough fight between Human and IBM Watson, and AI won the game over human!

[This slide contains a still image from the airing of the Jeopardy television show of three contestants playing against, and losing to, the IBM Watson computer.]

Slide 5: And How to use AI?

  • Example 1: AI Applications in TMC Operations
    • Similar to a chess game: Mr. NETWORK vs Mr. TMC Operator
    • Rules: Whenever NETWORK has a move for generating negative impacts, the Operator will have a move to minimize such impacts.
  • What DeepBlue and AlphaGo would inspire to us?
    • To gain high confidence level in AI applications
      And why: The difficult level (DL) of AlphaGo is much higher than the DL of TMC Operations!

[This slide contains two images: (1) a wall within a TMC that is displaying twelve (4x3) live feeds from traffic cameras and (2) a photo of a TMC technician on the phone looking at the monitor screen with a thought balloon above him displaying a chessboard.]

Slide 6: And How to use AI?

But the fact is that one operator is playing multiple games simultaneously—no human beings could handle such heavy work load.

Operator’s tasks: monitor thousands of sensors, understand the network situations by processing each sensor’s output, and conducting assessment using the output.

[This slide contains the same two images from the previous slide, except there are nine chessboards in the thought balloon in the second image.]

Slide 7: So, use AI this way!

Finding the similar character of Chess Gaming and TMC Operations

Both chess playing and TMC operations have some special characters:

  1. They all could learn knowledge from previous successful experiences and lessons learned.
  2. They all need to consider the impacts of each move—need thinking of short term impacts as well as long term impacts.
  3. They both can accumulate knowledge, so for a long run, both should be getting “Smarter over Time.”

Slide 8: Examples of TMC AI

The TMC Working Process

  • Big Data ↓
  • Useful Data ↓
  • Knowledge ↓
  • Analysis ↓
  • Decision and Implementation ↓
  • Knowledge and Experience Accumulation

The AI Application Process

  • Data Management ↓
  • Logic & Reasoning ↓
  • Prioritizing/Tradeoff Analysis ↓
  • Decision Support ↓
  • Assessment/Knowledge Building ↓
  • Knowledge Application

[This slide contains the TMC Working Process flowchart alongside the AI Application Process flowchart. Information from the flowcharts is reproduced above]

Slide 9: Examples of TMC AI

The Architecture of the AI [Expert System] Tool

Examples of TMC Applications

Data Interface
Expert System with a Knowledge Base
  • Collecting & arranging data
  • Performing data analysis
  • Performing simulation
  • Building knowledge base
  • Providing analysis outcome

Simulation
SW Interface
↓ ↑
User Interface

Key Features/Advantages:

  • Network-wide data analysis
  • On-line decision support or offline training
  • Fast and accurate response
  • Emulate human reasoning and problem-solving capability; learn and become smarter over time

[This slide contains a flowchart of the overall system architecture. The flowchart information is reproduced in the table above.]

Slide 10: Examples of TMC AI

Another Application: Using Artificial Neural Network (ANN) for TMC’s Traffic Demand Analysis and Prediction

Dynamic O-Ds, and demand prediction for 5, 10, and 15 minutes, NRC postdoc researcher Yi Zhao, 2018

[This slide contains Traffic Demand Analysis and Prediction charts.]

Slide 11: Examples of TMC AI

The Project Covered Area (in Northern Delaware)

  • It is the corridor of I-95 from DC to NYC, with parallel arterial US-40
  • Wilmington and Newark, the state’s largest cities, are clustered here
  • This corridor has high incident rate: on average 10~15 crashes/accidents each quarter
  • The test bed area includes 8 miles of I-95 freeway, 43 miles of state highway with 98 signalized intersections
  • Key roadway segments include I-95, US 13, US 40, DE 4, DE 7, DE 72, DE 273, DE 896 and Old Baltimore Pike

[This slide contains a map of Delaware and a map of the project area in Delaware.]

Slide 12: Examples of TMC AI

The Detection System
Blue-digital radar, Green-other traffic sensors

[This slide contains a map showing the traffic signal expert system in the DelDOT Newark area. The map is marked with blue icons (Wavetronix sensors for freeway volume, speed, and occupancy data collection) and green icons (signal system detectors for arterial volume and occupancy data collection). Some roadways are colored to show near real-time traffic conditions (volume, speed).]

Slide 13: Another Example of Using AI

Another AI Application: Vehicle Tracking using loop signature

[This slide contains two images: (1) a graphic of a car with an arrow pointing to the Loop and Loop Amplifier and (2) a line chart with Magnet Strength plotted against Time.]

Slide 14: Another Example of Using AI

The Vehicle Tracking; Concept of Operations
Obtaining space mean speed and routing data

[This slide contains two images: (1) a graphic of a car with an arrow pointing to Loop #1, which has an arrow pointing to Loop #2, which has an arrow pointing to a computer; and (2) two line charts: “Getting Next Signature” and “Getting Next Signature and reidentifying.”]

Slide 15: Another Example of Using AI

Signal matching or reidentification has been a daily activity for many business and agencies.

In recent years, AI has become a mature technology to carry out this mission, as well as similar missions, such as face identification, and voice identification, etc.

[This slide contains two identical photos of a person signing the bottom of a document.]

Slide 16: Examples of AI for signature matching

Concept of Quick Incident Detection using Loop Tracking

Example: if suddenly it was much longer than 30 seconds for a vehicles or a platoon to arrive at 2nd station in a lane.
Meanwhile other lanes did not see significant delay during same period - There must have been something happened.

Therefore, after this 30 seconds period, one special situation has been alerted.
If the situation lasts longer than a threshold, say, 90 seconds, an incident could be identified.

Nice concept, but how to make it happen? The answer should be: using AI to match loop signatures for vehicle tracking.

[This slide contains a graphic image of three vehicles traveling on the highway showing the concept of quick incident detection using loop tracking.]

Slide 17: The General Concept

We could formulated a TE (traffic engineering) problem as a signature matching problem, which AI has mature methods and products to address it.

Slide 18: Deep Learning makes system smarter

“Smarter Over Time” by Deep Learning:

The TMC Working Process

  • Big Data ↓
  • Useful Data ↓
  • Knowledge ↓
  • Analysis ↓
  • Decision and Implementation ↓
  • Knowledge and Experience Accumulation

The AI Application Process

  • Data Management ↓
  • Logic & Reasoning ↓
  • Prioritizing/Tradeoff Analysis ↓
  • Decision Support ↓
  • Assessment/Knowledge Building ↓
  • Knowledge Application

[This slide contains the two flowcharts from Slide 8 with two of the boxes highlighted. The information from these two flowcharts is reproduced above. The two highlighted boxes are in bold above.]

Slide 19: That is All I want to present

Summary

  • AI technology and tools are available now.
  • Traffic engineering problems can be formulated to the problems which AI has the strength to deal with.
  • The Advantage of using AI - It could make the system “Smarter Over Time” by its learning capability.

↑ Return to top

SUPPORT

Get technical 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 twittergo to email