A method and system for optimizing traffic state of a weaving area of a road
By collecting video data using drones, and combining it with the reinforcement learning environment and road channelization of the YOLO algorithm and SUMO traffic simulation software, the traffic conditions in the weaving areas of urban expressways and highways are optimized, solving the problem of low efficiency in existing technologies and achieving more efficient traffic flow management.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- TONGJI UNIV
- Filing Date
- 2023-04-06
- Publication Date
- 2026-06-16
AI Technical Summary
Existing traffic flow optimization methods in urban expressway weaving areas are inefficient and cannot meet the needs of practical applications.
This study employs a reinforcement learning algorithm and road channelization approach. Video image data is collected by drones, vehicle targets and trajectories are identified using the YOLO algorithm, and traffic conditions are optimized using SUMO traffic simulation software. A reinforcement learning environment and road channelization are set up to optimize traffic flow.
It improves the efficiency of traffic condition optimization in weaving areas, provides more valuable optimization results, is applicable to urban expressways and highways, and enhances the efficiency of traffic flow organization and management.
Smart Images
Figure CN116227230B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of road optimization technology, specifically to a method and system for optimizing traffic conditions in road weaving areas. Background Technology
[0002] With the accelerating pace of urbanization, urban traffic congestion has become increasingly prominent. As a crucial component of urban transportation, the primary function of urban expressways is to ensure the rapid and smooth flow of vehicles within the city. However, with the rapid increase in traffic volume, traffic congestion on urban expressways, especially in weaving areas, has become severe. Weaving areas refer to road sections where several different traffic flows intersect in the same general direction without the use of any traffic control devices. Due to urban land constraints, the distance between entrances and exits on urban expressways is typically short, leading to significant weaving phenomena and a marked reduction in efficiency.
[0003] The existing optimization of traffic conditions in urban expressway weaving areas mainly includes the following two methods: road channelization and ramp control.
[0004] Channelization on urban expressways typically uses traffic markings on the road surface to separate different lanes based on traffic volume or direction, allowing vehicles traveling at different routes and speeds to proceed in the prescribed direction without interfering with each other, much like water flowing in a channel. Channelization effectively and orderly organizes traffic flow through weaving areas, reduces traffic conflicts, and maximizes the use of road resources. Reasonable lane separation with continuity and closure can better alleviate traffic congestion and reduce the risk of collisions.
[0005] Ramp control refers to the use of signal control methods to limit the number of vehicles entering the main lane from urban expressway ramps. ALINEA (Automatic Highway Input Linear Control) is a well-known simple feedback loop controller used for ramp metering control; its theory has been proven to effectively reduce travel time and increase vehicle speed under appropriate traffic conditions. After investigating the impact of traffic demand levels, queue overflow handling strategies, and downstream bottleneck conditions on ALINEA performance, it was found that while ALINEA can increase traffic capacity, it also results in more travel and waiting times. With the rapid development of machine learning, many intelligent algorithms, especially reinforcement learning, have been applied to ramp control.
[0006] In practical applications, the aforementioned traffic flow optimization methods for urban expressway weaving areas cannot meet the practical needs of traffic flow in these areas, or are inefficient, leading to insufficient optimization or other problems. Summary of the Invention
[0007] To address the problem of low efficiency in existing methods for optimizing traffic flow in urban expressway weaving areas, this invention provides a method and system for optimizing traffic conditions in road weaving areas. Based on reinforcement learning algorithms and road channelization, the method continuously optimizes traffic conditions in urban expressway weaving areas, making the optimization results more practically valuable and solving the problems of insufficient optimization and low efficiency in existing technologies.
[0008] A method for optimizing traffic conditions in road weaving areas includes the following steps:
[0009] Collect video image data of intersecting road sections;
[0010] Video image data is input into a vehicle target and trajectory recognition and detection model based on the YOLO algorithm to obtain vehicle target and trajectory data;
[0011] Set up the reinforcement learning environment and road channelization in the SUMO traffic simulation scenario;
[0012] Vehicle target and trajectory data are input into SUMO traffic simulation software. Through vehicle target and trajectory data and car-following lane-changing model, the driving trajectory, basic driving parameters and overall traffic operation status of all vehicles in the weaving area are obtained.
[0013] Furthermore, the video image data of the road weaving section is collected in real time by drones, and the collected video image data is saved to hardware devices.
[0014] Furthermore, the construction process of the vehicle target and trajectory recognition and detection model includes the following steps:
[0015] Acquire video image data of road weaving sections collected by drones;
[0016] Manually calibrate vehicle targets and trajectories in video image data to establish a vehicle target and trajectory recognition dataset;
[0017] A YOLO vehicle target and trajectory recognition and detection model was built based on a vehicle target and trajectory dataset.
[0018] Furthermore, the basic driving parameters include the vehicle's speed, acceleration, heading angle, braking status, and vehicle status lights.
[0019] Furthermore, it also includes calibrating the car-following lane-changing model using the least squares method or genetic optimization algorithm.
[0020] Furthermore, the reinforcement learning methods include traditional greedy strategy-based reinforcement learning methods and deep reinforcement learning methods.
[0021] Furthermore, the setup of the reinforcement learning method includes the following steps:
[0022] Construct a reinforcement learning environment for the SUMO simulation scenario;
[0023] Install traffic lights that can change in real time according to traffic conditions;
[0024] Configure the executable actions of the agent in reinforcement learning;
[0025] The road and traffic information provided by SUMO software serves as the environment for reinforcement learning methods.
[0026] Furthermore, the road channelization setting is achieved by using the editing function of SUMO simulation software to change the dashed and solid lines of road markings.
[0027] Furthermore, the vehicle operation data and overall traffic conditions of the weaving area output by the SUMO traffic simulation are compared and analyzed with the detection results obtained by inputting video image data into the vehicle target and trajectory recognition and detection model based on the YOLO algorithm to determine whether the optimization results have met expectations.
[0028] Furthermore, an optimization system for traffic conditions in road weaving areas includes:
[0029] The acquisition module is used to acquire video image data of intersecting road sections;
[0030] The recognition and detection module is used to input video image data into the vehicle target and trajectory recognition and detection model based on the YOLO algorithm to obtain vehicle target and trajectory data;
[0031] The reinforcement learning environment setting module is used to set up the reinforcement learning method environment and road channelization in the SUMO traffic simulation scenario;
[0032] The output module is used to input vehicle target and trajectory data into the SUMO traffic simulation software. Through the vehicle target and trajectory data and the car-following lane-changing model, the driving trajectory, basic driving parameters, and overall traffic operation status of the weaving area of all vehicles are obtained.
[0033] This invention provides a method and system for optimizing traffic conditions in road weaving areas, which has the following beneficial effects:
[0034] (1) This invention compares and analyzes the driving trajectories, basic driving parameters and overall traffic operation status of all vehicles obtained after setting up a reinforcement learning environment and channelizing roads in the SUMO traffic simulation scenario with the results detected by the YOLO vehicle target and trajectory recognition detection model. Based on reinforcement learning and road channelization, the optimization method for traffic status optimization in urban expressway weaving areas has more efficient optimization results and greater application value.
[0035] (2) This invention is not only applicable to urban expressways, but also to highways. By retraining the traffic state optimization method for weaving areas based on reinforcement learning and road channelization, the traffic state optimization of weaving areas can be completed relatively easily.
[0036] (3) Compared with traditional single optimization methods, the present invention combines reinforcement learning and road channelization to better optimize and improve the traffic conditions in weaving areas. Attached Figure Description
[0037] Figure 1 This is a flowchart illustrating the traffic state method optimization in an embodiment of the present invention;
[0038] Figure 2 This is a schematic diagram illustrating step 2 of the present invention, which uses YOLO series algorithms to perform vehicle target and trajectory recognition on video data of urban expressway weaving sections collected by UAVs.
[0039] Figure 3 This is an example diagram illustrating the reinforcement learning method environment and road channelization settings in step 4 of this embodiment of the invention;
[0040] Figure 4 This is an example diagram illustrating step 5 of the present invention, representing a SUMO traffic simulation. Detailed Implementation
[0041] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.
[0042] This invention provides a method for optimizing traffic conditions in road weaving areas; the steps are as follows:
[0043] Step 1: Based on video data collected by drones from urban expressway weaving sections, real-time and video recording data are generated and saved to hardware devices; this invention uses drones for data extraction, which can maximize the authenticity of traffic conditions in urban expressway weaving areas and improve the rationality of the invention.
[0044] Step 2: Based on the YOLO series algorithms, vehicle target and trajectory recognition is performed on video data of urban expressway weaving sections collected by the UAV. Basic driving parameters include vehicle speed, acceleration, heading angle, braking status, and vehicle status lights. YOLO algorithms include, but are not limited to, the already released YOLO, YOLOv2, YOLOv3, YOLOv5, YOLOX, and other target recognition algorithms based on YOLO extensions. The recognition process is as follows: First, acquire video image data of urban expressway weaving sections collected by the UAV; second, manually calibrate the vehicle targets and trajectories in the traffic vehicle image data of the urban expressway weaving sections to establish a vehicle target and trajectory recognition dataset; then, construct a YOLO vehicle target and trajectory recognition detection model based on the vehicle target and trajectory dataset; finally, input the video image data to be detected into the YOLO vehicle target and trajectory recognition detection model to obtain the detection results.
[0045] Step 3: Transmit the identified vehicle targets and trajectory data to the SUMO traffic simulation software and calibrate some car-following and lane-changing models. The calibration method involves using the least squares method or genetic optimization algorithm to calibrate the car-following and lane-changing models based on the identified vehicle target and trajectory data. The calibration targets are the parameters of the W99 or IDM car-following model and the SL2015 lane-changing model, calibrating parameters for different types of vehicle car-following and lane-changing models. Setting up the SUMO simulation scene requires these parameters from the car-following and lane-changing models, providing the actual traffic parameters for the subsequent SUMO traffic simulation.
[0046] Step 4: Set up the reinforcement learning environment and road channelization settings. Reinforcement learning methods include traditional greedy strategy-based reinforcement learning and deep reinforcement learning. First, construct the reinforcement learning environment for the SUMO simulation scenario, setting up traffic lights that change in real-time according to traffic conditions. Simultaneously, set the executable actions of the agent in reinforcement learning (here, the traffic lights can change to red, yellow, and green). Use the road and traffic information provided by the SUMO software (such as road length, width, number of lanes, vehicle speed, traffic flow, road vehicle density, and queue length) as the environment for the reinforcement learning method. Road channelization settings are implemented using the editing function of the SUMO simulation software to change the solid and dashed lines of road markings.
[0047] Step 5: Execute the SUMO traffic flow simulation software, preset the speed and trajectory of the driving vehicles and the vehicles in front of them in the video sampled by the drone, and record the driving trajectory and basic driving parameters of all vehicles in the entire simulation environment, as well as the overall traffic operation status of the weaving area.
[0048] Step 6: Optimize traffic flow in the target weaving area using a traffic efficiency optimization method for urban expressway weaving areas based on reinforcement learning and road channelization. For example... Figure 3 The diagram shown illustrates step 4 of this embodiment of the reinforcement learning method, including the environment and road channelization settings. ABCDEF represent the area numbers of the weaving zone, A represents the area before merging, BC represents the area where weaving mainly occurs (also the area where road channelization is set up), DE represents the area where weaving is not severe, and F represents the area where vehicles separate lanes. A traffic light exists between areas AB, representing the traffic light used for reinforcement learning control. The arrows in areas BC indicate where vehicles can change lanes and the directions in which they can change lanes (the directions in which lanes can change change after road channelization).
[0049] Step 7: The optimization effect of this method is judged by analyzing the vehicle operation data and overall traffic conditions of the weaving area output by SUMO traffic simulation. The vehicle operation data and overall traffic conditions output by SUMO traffic simulation are compared with the vehicle speed, traffic flow, traffic density, and queue length calculated from actual video footage and vehicle target and trajectory data identified using YOLO. If the simulation output of these indicators is better than the indicators calculated from actual video footage, the optimization effect is considered good. Simultaneously, the traffic service level of the area can be calculated, and the optimization effect of this method is evaluated based on the level of traffic service.
[0050] The specific basic driving parameters are shown in Table 1 below:
[0051] Table 1 Basic Driving Parameters
[0052]
[0053]
[0054] An optimization system for a method to optimize traffic conditions in road weaving areas includes:
[0055] The acquisition module is used to acquire video image data of rapidly intersecting urban areas;
[0056] The calibration module is used to manually calibrate vehicle targets and trajectories in video image data, thereby establishing a dataset for vehicle target and trajectory recognition;
[0057] The recognition and detection module is used to build a YOLO vehicle target and trajectory recognition and detection model based on the dataset. The video image data to be detected is input into the YOLO vehicle target and trajectory recognition and detection model to obtain the detection results.
[0058] The car-following lane-changing model calibration module is used to input the test results into the SUMO traffic simulation software for car-following lane-changing model calibration;
[0059] The optimization module obtains the driving trajectories, basic driving parameters, and overall traffic operation status of all vehicles in the weaving area by setting up a reinforcement learning environment and road channelization in the SUMO traffic simulation scenario.
[0060] The comparative analysis module compares and analyzes traffic operation status with detection results to determine whether the optimization results have met expectations.
[0061] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.
Claims
1. A method for optimizing traffic conditions in road weaving areas, characterized in that, Includes the following steps: Collect video image data of intersecting road sections; Video image data is input into a vehicle target and trajectory recognition and detection model based on the YOLO algorithm to obtain vehicle target and trajectory data; Set up the reinforcement learning environment and road channelization in the SUMO traffic simulation scenario; Vehicle target and trajectory data are input into SUMO traffic simulation software. Through vehicle target and trajectory data and car-following lane-changing model, the driving trajectory, basic driving parameters and overall traffic operation status of all vehicles in the weaving area are obtained.
2. The method for optimizing traffic conditions in a road weaving area according to claim 1, characterized in that, The video image data of the road weaving section is collected in real time by drones, and the collected video image data is saved to hardware devices.
3. The method for optimizing traffic conditions in a road weaving area according to claim 1, characterized in that, The construction process of the vehicle target and trajectory recognition and detection model includes the following steps: Acquire video image data of road weaving sections collected by drones; Manually calibrate vehicle targets and trajectories in video image data to establish a vehicle target and trajectory recognition dataset; A YOLO vehicle target and trajectory recognition and detection model was built based on a vehicle target and trajectory dataset.
4. A method for optimizing traffic conditions in a road weaving area as described in claim 1, characterized in that, The basic driving parameters include vehicle speed, acceleration, heading angle, braking status, and vehicle status lights.
5. The method for optimizing traffic conditions in a road weaving area according to claim 1, characterized in that, It also includes the calibration of the car-following lane-changing model using the least squares method or genetic optimization algorithm.
6. The method for optimizing traffic conditions in a road weaving area according to claim 1, characterized in that, The reinforcement learning methods include traditional greedy strategy-based reinforcement learning methods and deep reinforcement learning methods.
7. The method for optimizing traffic conditions in a road weaving area according to claim 1, characterized in that, The setup of the reinforcement learning method includes the following steps: Construct a reinforcement learning environment for the SUMO simulation scenario; Install traffic lights that can change in real time according to traffic conditions; Configure the executable actions of the agent in reinforcement learning; The road and traffic information provided by SUMO software serves as the environment for reinforcement learning methods.
8. The method for optimizing traffic conditions in a road weaving area according to claim 1, characterized in that, The road channelization setting is achieved by using the editing function of SUMO simulation software to change the dashed and solid lines of road markings.
9. The method for optimizing traffic conditions in a road weaving area according to claim 1, characterized in that, By comparing and analyzing the vehicle operation data and overall traffic conditions of the weaving area output by the SUMO traffic simulation with the detection results obtained by inputting video image data into the vehicle target and trajectory recognition and detection model based on the YOLO algorithm, it is possible to determine whether the optimization results have met expectations.
10. An optimization system based on the road weaving zone traffic state optimization method according to claim 1, characterized in that, include: The acquisition module is used to acquire video image data of intersecting road sections; The recognition and detection module is used to input video image data into the vehicle target and trajectory recognition and detection model based on the YOLO algorithm to obtain vehicle target and trajectory data; The reinforcement learning environment setting module is used to set up the reinforcement learning method environment and road channelization in the SUMO traffic simulation scenario; The output module is used to input vehicle target and trajectory data into the SUMO traffic simulation software. Through the vehicle target and trajectory data and the car-following lane-changing model, the driving trajectory, basic driving parameters, and overall traffic operation status of the weaving area of all vehicles are obtained.