METHOD OF PREDICTING TRAFFIC CONGESTION AND CONTROLLING TRAFFIC SIGNALS BASED ON DEEP LEARNING AND SERVER FOR PERFORMING THE SAME

• US 20200135018A1
• Filed: 11/12/2018
• Published: 04/30/2020
• Est. Priority Date: 10/24/2018
• Status: Active Grant

• Current Assignee: Bluesignal Corporation
• Original Assignee: Bluesignal Corporation
• Inventors: BAIK, Seng Tae
• Granted Patent: US 10,902,719 B2
• Time in Patent Office: Days
• CPC Class Codes: G06F 18/217 Validation; Performance evaluation; Active pattern learning techniques
• CPC Class Codes: G06F 18/251 of input or preprocessed data
• CPC Class Codes: G06N 3/08 Learning methods
• CPC Class Codes: G06N 5/02 Knowledge representation; Symbolic representation
• CPC Class Codes: G06V 10/764 using classification, e.g. of video objects
• CPC Class Codes: G06V 10/803 of input or preprocessed data
• CPC Class Codes: G06V 10/82 using neural networks
• CPC Class Codes: G06V 20/54 of traffic, e.g. cars on the road, trains or boats
• CPC Class Codes: G08G 1/0116 from roadside infrastructure, e.g. beacons
• CPC Class Codes: G08G 1/0129 for creating historical data or processing based on historical data
• CPC Class Codes: G08G 1/0133 for classifying traffic situation
• CPC Class Codes: G08G 1/0145 for active traffic flow control
• CPC Class Codes: G08G 1/04 using optical or ultrasonic detectors
• CPC Class Codes: G08G 1/08 according to detected number or speed of vehicles
• CPC Class Codes: G08G 1/081 Plural intersections under common control

Abstract

A method of predicting traffic congestion and controlling traffic signals based on deep learning according to an embodiment of the present invention includes: analyzing regional network outflow behavior based on a per-intersection traffic demand pattern or traffic signal control, and determining a specific intersection to be a control target intersection based on the results of the analysis; generating 2D space-time images by analyzing the data of predetermined customized composite data corresponding to the control target intersection and a plurality of pieces of image data for the control target intersection in terms of time and space; generating a real-time traffic congestion index by using the 2D space-time image of the control target intersection and the 2D space-time image of the data of the customized composite data corresponding to the control target intersection; and controlling the traffic signals of the control target intersection based on the real-time traffic congestion index.

Claims

1. A method of predicting traffic congestion and controlling traffic signals based on deep learning, the method being performed in a deep learning-based traffic congestion prediction and traffic signal control server, the method comprising:

analyzing regional network outflow behavior based on a per-intersection traffic demand pattern or traffic signal control, and determining a specific intersection to be a control target intersection based on results of the analysis;

generating second-dimensional (2D) space-time images by analyzing data of predetermined customized composite data corresponding to the control target intersection and a plurality of pieces of image data for the control target intersection in terms of time and space;

generating a real-time traffic congestion index by using the 2D space-time image of the control target intersection and the 2D space-time image of the data of the customized composite data corresponding to the control target intersection; and

controlling traffic signals of the control target intersection based on the real-time traffic congestion index.

2. The method of claim 1, wherein analyzing the per-intersection traffic demand pattern or regional network outflow behavior based on traffic signal control, and determining the specific intersection to be the control target intersection based on the results of the analysis comprises:

determining a change in regional traffic volume by analyzing a traffic demand (outbound demand) directed to areas outside a downtown road network area and a traffic demand (local demand) directed to an inside of the area, such as a residential area road and a parking space; and

determining a specific intersection to be the control target intersection based on the change in regional traffic volume.

3. The method of claim 1, wherein generating the 2D space-time images by analyzing the data of the predetermined customized composite data corresponding to the control target intersection and the plurality of pieces of image data for the control target intersection in terms of time and space comprises:

generating a grid for a 2D space-time domain by dividing a space into a specific number of sections in the 2D space-time image of the control target intersection, and then calculating average travel speeds for the respective sections; and

determining a corresponding section to be any one of a plurality of channels based on its average travel speed, and inputting a numerical value corresponding to the determined channel into the grid.

4. The method of claim 1, wherein generating the 2D space-time images by analyzing the data of the predetermined customized composite data corresponding to the control target intersection and the plurality of pieces of image data for the control target intersection in terms of time and space comprises:

generating the 2D space-time images by combining weather information and unexpected situation information of the customized composite data for each piece of time environment information and corresponding space information with road environment information; and

labeling the 2D space-time images based on the road environment information.

5. A deep learning-based traffic congestion prediction and traffic signal control server, comprising:

a control target intersection determination unit configured to analyze regional network outflow behavior based on a per-intersection traffic demand pattern or traffic signal control, and to determine a specific intersection to be a control target intersection based on results of the analysis;

a 2D space-time image generation unit configured to generate 2D space-time images by analyzing data of predetermined customized composite data corresponding to the control target intersection and a plurality of pieces of image data for the control target intersection in terms of time and space;

a real-time traffic congestion index generation unit configured to generate a real-time traffic congestion index by using the 2D space-time image of the control target intersection and the 2D space-time image of the data of the customized composite data corresponding to the control target intersection; and

a traffic signal control unit configured to control traffic signals of the control target intersection based on the real-time traffic congestion index.

6. The deep learning-based traffic congestion prediction and traffic signal control server of claim 5, wherein the control target intersection determination unit is further configured to:

determine a change in regional traffic volume by analyzing a traffic demand (outbound demand) directed to areas outside a downtown road network area and a traffic demand (local demand) directed to an inside of the area, such as a residential area road and a parking space; and

determine a specific intersection to be the control target intersection based on the change in regional traffic volume.

7. The deep learning-based traffic congestion prediction and traffic signal control server of claim 5, wherein the 2D space-time image generation unit is further configured to:

generate a grid for a 2D space-time domain by dividing a space into a specific number of sections in the 2D space-time image of the control target intersection, and then calculate average travel speeds for the respective sections; and

determine a corresponding section to be any one of a plurality of channels based on its average travel speed, and input a numerical value corresponding to the determined channel into the grid.

8. The deep learning-based traffic congestion prediction and traffic signal control server of claim 5, wherein the 2D space-time image generation unit is further configured to:

generate the 2D space-time images by combining weather information and unexpected situation information of the customized composite data for each piece of time environment information and corresponding space information with road environment information; and

label the 2D space-time images based on the road environment information.