CN113313943B - Road side perception-based intersection traffic real-time scheduling method and system - Google Patents

Abstract

The invention discloses a road side perception-based real-time crossing traffic scheduling method and system, which comprises the following specific steps of collecting corresponding scene information by a road side unit; identifying vehicles, pedestrians and non-motor vehicles in the scene information through information fusion to perform target level fusion to obtain fused target information in the current scene; extracting the target information of corresponding regional lane levels according to the road information of the high-precision map and the fused target information, and counting the traffic demands of different courses of the intersection; and inputting the traffic demands of the intersection with different courses into a traffic scheduling algorithm, calculating to obtain a lighting strategy of the intersection, and feeding back and adjusting. The traffic light control system can count the traffic demands of vehicles, pedestrians and non-motor vehicles with different courses in real time according to the result of the road side sensing module, and autonomously adjust the light-on strategy and light-on time of the traffic light at the intersection, so that the traffic efficiency at the intersection is improved, and the traffic pressure is relieved.

Description

A real-time scheduling method and system for intersection traffic based on roadside perception

technical field

The invention relates to the technical field of intersection traffic scheduling, in particular to a real-time intersection traffic scheduling method and system based on roadside perception.

Background technique

With the rapid development of economy and technology, people's daily life is becoming more and more intelligent. With the gradual improvement of living standards, people's requirements for the convenience of daily travel are also getting higher and higher. At present, almost all crossroads in China are equipped with traffic lights, which provide a guarantee for people to travel safely. However, in recent years, the number of private cars in my country has gradually increased, and the transportation systems of major cities are facing enormous pressure. Alleviating traffic pressure has gradually become a problem faced by the transportation systems of major cities.

There are many reasons for traffic jams, one of which cannot be ignored is that the signal lights are not intelligent enough, causing vehicles to form blockages at intersections. The existing traffic light system is still in the mode of setting a fixed lighting time, which is completely adjusted by human experience, and cannot intelligently change the lighting time according to the traffic flow of each road section, which is inefficient.

SUMMARY OF THE INVENTION

The purpose of the present invention is to overcome the shortcomings of the prior art. To achieve the above purpose, a real-time scheduling method for crossing traffic based on roadside perception is adopted to solve the problems raised in the above background technology.

The first technical solution adopted by the present invention: a real-time scheduling method for intersection traffic based on roadside perception, comprising:

Use the roadside unit to collect the corresponding scene information;

Identify vehicles, pedestrians and non-motor vehicles in the scene information through information fusion and perform target-level fusion to obtain the fused target information in the current scene;

According to the road information of the high-precision map and the fused target information, extract the target information of the corresponding area lane level, and make statistics on the traffic demand of the intersection in different directions;

Input the traffic demand of different directions at the intersection into the traffic scheduling algorithm, calculate the lighting strategy of the intersection, and feedback the adjustment.

As a further solution of the present invention: the specific steps of identifying vehicles, pedestrians and non-motor vehicles in the scene information through information fusion to perform target-level fusion to obtain the fused target information in the current scene include:

Collect the point cloud data of the target in the scene information for segmentation, clustering, feature extraction and target recognition, and extract the target center point and anchor frame information of the target clustering: (x, y, l, w, h, c), where (x ,y) is the abscissa and ordinate of the target in the radar coordinate system, (l,w,h) is the length, width and height of the target, and c is the category of the target;

At the same time, the image data of the target in the scene information is collected, and the deep learning model is used to identify vehicles, pedestrians and non-motor vehicles at the intersection, and the target center information and anchor frame information are extracted as: (x, y, l, w, c);

The target information in the radar coordinate system is mapped to the image, and the target-level fusion is performed according to the intersection of the two anchor boxes, and the position and category information of the target are output.

As a further scheme of the present invention: the specific steps of extracting the target information of the corresponding area lane level according to the road information of the high-precision map and the fused target information, and carrying out the statistics of the traffic demands of different headings at the intersection include:

Load the high-precision map of the intersection area and obtain the target information after target-level fusion;

According to the high-precision map, the fused target information is mapped to the current intersection area, and the target information of all lanes is extracted and counted as: (no, t, count _left , count _forward , count _rightt , count _cross );

Among them, no represents the current roadside unit number, t represents the GPS time at the time of radar information collection, (count _left , count _forward , count _rightt , count _cross ) represent the left-turn demand count, straight-through demand count, right Turn demand counts, as well as pedestrian and non-motor vehicle demand counts in zebra crossing areas.

As a further solution of the present invention: the specific steps of extracting and counting the target information of all lanes include:

Obtain the number of lanes in different directions, and make statistics on the traffic demand of vehicles in different lanes;

Left-turn vehicle traffic demand count:

Through-vehicle traffic demand count:

Right-turn vehicle traffic demand count:

Among them, a, b, and c are the number of independent straight, left, and right turn lanes at the intersection, respectively, d is the number of left-turn and straight collinear lanes, and e is right-turn and straight collinear lanes.

As a further scheme of the present invention: the specific steps of inputting the traffic demand of the intersection in different directions into the traffic scheduling algorithm, calculating the lighting strategy of the intersection, and feeding back the adjustment include:

Step 1: Obtain the statistical results of lane-level target information in different directions at the intersection;

Step 2: First, make sure that the two-way straight traffic lights are consistent with the zebra crossing traffic lights on both sides of the same direction;

Step 3: At the beginning of a traffic cycle, according to the statistical results of the target information, select the current traffic direction with the largest traffic demand, and delete this direction from the traffic demand list, and other traffic directions are prohibited. , and the time-consuming is less than the preset threshold t _thres , or when the current vehicle passing time exceeds the preset threshold t _thres , the direction is prohibited;

Step 4: In the current traffic cycle, select the direction with the greatest demand from the traffic list as the traffic direction, and prohibit traffic in other directions. At the same time, when the traffic in this direction ends and the time consumption is less than the preset threshold t _thres , or when this time When the passing time of the vehicle exceeds the preset threshold t _thres , the direction is prohibited;

Step 5: Iteratively execute until all directions in the current traffic cycle are completed, and the start of the new traffic cycle t _thres is the preset threshold for the release time.

The second technical solution adopted by the present invention: a system including a real-time scheduling method for intersection traffic based on roadside perception as described in any of the above, including:

The roadside perception module is used to collect scene data information in the intersection area and fuse the target information;

The traffic scheduling module is used to obtain the traffic requirements of different directions at the intersection and output the traffic scheduling strategy;

The traffic light module is used to receive the light-on strategy and light-on time output by the traffic scheduling module, and to indicate the traffic at the intersection;

The communication module is used for signal communication between the roadside sensing module, the traffic scheduling module and the traffic light module.

As a further solution of the present invention, the traffic scheduling module is respectively connected to the roadside sensing module and the traffic light module through the communication module.

As a further solution of the present invention: the roadside perception module includes a high-definition camera, a laser radar and an embedded computing unit.

Compared with the prior art, the present invention has the following technical effects:

By adopting the above technical solution, by setting roadside perception modules in different lane areas of the intersection, the information of vehicles, pedestrians and non-motor vehicles in different areas is collected, and the target information is fused through the embedded computing unit, and transmitted to the traffic scheduling. The module can set specific strategies, so as to output the time control amount to the traffic light control system, and then set a reasonable traffic time according to the vehicle traffic conditions at the intersection in real time, so as to adjust the traffic flow of each road section at different time periods, and intelligently change the lighting. Time, autonomously adjust the traffic light strategy and lighting time at the intersection, thereby improving the traffic efficiency of the intersection and alleviating the traffic pressure.

Description of drawings

Below in conjunction with the accompanying drawings, the specific embodiments of the present invention are described in detail:

FIG. 1 is a schematic diagram of steps of a real-time scheduling method for intersection traffic according to some embodiments disclosed in the present application;

2 is a schematic structural diagram of a system for a real-time scheduling method for intersection traffic according to some embodiments disclosed in the present application;

FIG. 3 is a flowchart of real-time scheduling of crossing traffic according to some embodiments disclosed in the present application.

Detailed ways

The technical solutions in the embodiments 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 a part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

Referring to FIG. 1 and FIG. 3 , in an embodiment of the present invention, a real-time scheduling method for intersection traffic based on roadside perception, the specific steps include:

S1. Use roadside units to collect corresponding scene information;

S2, identify vehicles, pedestrians and non-motor vehicles in the scene information through information fusion and perform target-level fusion to obtain the fused target information in the current scene. The specific steps include:

The point cloud data in the scene information is collected by lidar, the ground is filtered out by point cloud segmentation, the segmented point cloud is clustered, and the target center point and anchor frame information of the target cluster are extracted: (x, y, l, w, h, c), where (x, y) is the abscissa and ordinate of the target in the radar coordinate system, (l, w, h) is the length, width and height of the target, and c is the category of the target;

At the same time, the image data of the target in the scene information is collected, and the YoLo V4 deep learning model is used to identify vehicles, pedestrians and non-motor vehicles at the intersection, and the target center information and anchor frame information are extracted as: (x, y, l, w, c);

According to the external calibration matrix of lidar and high-definition camera, the target information in the radar coordinate system is mapped to the image, and the target-level fusion is performed according to the anchor frame intersection ratio of the actual target and the lidar target, and the position and category information of the target are output.

S3. According to the road information of the high-precision map and the fused target information, extract the target information of the corresponding area lane level, and perform statistics on the traffic demand of the intersection in different directions. The specific steps include:

Load the high-precision map of the intersection area and obtain the target information after target-level fusion;

According to the high-precision map information of the intersection, the fused target information is mapped to the current intersection area, and the target information of all lanes is extracted and counted as: (no, t, count _left , count _forward , count _rightt , count _cross ), through The ROS node publishes this information.

Among them, no represents the current roadside unit number, t represents the GPS time at the time of radar information collection, (count _left , count _forward , count _rightt , count _cross ) represent the left-turn demand count, straight-through demand count, right Turn demand counts, as well as pedestrian and non-motor vehicle demand counts in zebra crossing areas.

In some specific embodiments, the specific steps of extracting and counting target information of all lanes include:

Obtain the number of lanes in different directions, and make statistics on the traffic demand of vehicles in different lanes;

Left-turn vehicle traffic demand count:

Through-vehicle traffic demand count:

Right-turn vehicle traffic demand count:

Among them, a, b, and c are the number of independent straight, left and right turn lanes at the intersection, respectively, d is the number of left-turn and straight collinear lanes, and e is the right-turn and straight collinear lanes.

S4. Input the traffic demand of the intersection in different directions into the traffic scheduling algorithm, calculate the lighting strategy of the intersection, and feedback the adjustment. The specific steps include:

The traffic lights in all directions of the intersection are defined as one traffic cycle. In each traffic cycle, select the direction corresponding to the largest demand from the traffic demand list as the traffic direction. The corresponding lane and zebra crossing are green lights, and the rest are red lights. , to judge whether the passage is over or whether the lighting time exceeds the set threshold. If so, select the one with the largest demand from the remaining headings from the passing demand list as the current lighting heading, and cycle until the end of a passing period. At the beginning of the next traffic cycle, the lighting sequence is determined according to the real-time sensing results of the roadside sensing module.

Step 1: Obtain the statistical results of lane-level target information of different headings at the intersection, and comprehensively count the vehicle information of different heading requirements in the current intersection area, as well as the pedestrian and non-motor vehicle traffic demand information on the zebra crossing. Assuming that the current intersection road is east-west direction and north-south direction, the traffic demand variables after statistics are as follows:

c _{l_sw} is the count of vehicles required to turn left in the direction of south to west (north to east);

c _{l_wn} is the count of vehicles required to turn left in the direction of west-to-north (east-to-south) direction;

c _{f_ew} is the count of the vehicle going straight demand in the east-west (west-east) direction;

c _{f_sn} is the count of vehicles going straight in the north-south (north-south) direction;

c _{r_ew} counts the right-turn demand of vehicles in the east-west direction;

c _{r_sn} is the right-turn demand count of vehicles in the north-south direction;

p _{f_e} is the count of pedestrian and non-motor vehicle traffic demand on the east lane zebra crossing;

p _{f_w} is the count of pedestrian and non-motor vehicle traffic demand on the west zebra crossing;

p _{f_s} is the count of pedestrian and non-motor vehicle traffic demand on the south zebra crossing;

p _{f_n} is the count of pedestrian and non-motor vehicle traffic demand on the north zebra crossing.

Step 2: Input the above parameters into the traffic scheduling algorithm, and then determine that the two-way straight traffic lights and the zebra crossing traffic lights on both sides of the same direction keep the same signal, specifically:

The traffic lights going straight in the east-west (west-east) direction are consistent with the traffic lights on the south and north zebra crossings;

The traffic lights going straight in the north-south (north-south) direction are consistent with the traffic lights on the east and west zebra crossings.

Step 3: According to max{c _{l_sw} ,c _{l_wn} ,c _{f_ew} +p _{f_s} +p _{f_n} ,c _{f_sn} +p _{f_e} +p _{f_w} }, at the beginning of a traffic cycle, according to the statistical results of the target information, select the one with the largest traffic demand as The current direction of travel will be deleted from the traffic demand list, and other directions will be prohibited. At the same time, when the traffic in this direction ends and the time-consuming is less than the preset threshold t _thres , or when the current traffic time exceeds the preset threshold t _thres , the direction is prohibited;

The implementation manner of the specific embodiment is:

Assuming that the traffic demand in the north-south direction is the largest, then the north-south (north-south) direction will pass through (the green light will be on, and the red lights will be on in other directions), and the east-west zebra crossing will pass. If the accumulated traffic in the north-south direction ends, and the time is less than t _thres , then it is prohibited to go straight in that direction (the red light for straight travel is on); if the cumulative time for this traffic exceeds t _thres , then the straight travel is prohibited in that direction (the red light for straight travel is on) ;

Step 4: In the current traffic cycle, select the direction with the greatest demand from the traffic list as the traffic direction, and prohibit traffic in other directions. At the same time, when the traffic in this direction ends and the time consumption is less than the preset threshold t _thres , or when this time When the passing time of the vehicle exceeds the preset threshold t _thres , the direction is prohibited; the specific steps are:

In step 3, when two-way straight driving in this direction is prohibited, according to max{c _{l_sw} ,c _{l_wn} ,c _{f_ew} +p _{f_s} +p _{f_n} }, the left turning direction with the largest traffic demand in this direction is obtained, then the vehicle in this left turning direction can pass, Other directions are prohibited, and when the vehicle in this direction ends and the time consumption is less than the preset threshold t _thres , or when the current vehicle passing time exceeds the preset threshold t _thres , the direction is prohibited;

The implementation manner of the specific embodiment is:

If the traffic demand in the left-turn direction from south to west (north to east) is the largest, the green light for left turn from south to west and north to east will be on, and the red lights in other directions, including the zebra crossing, will be on; if the cumulative traffic in this direction ends, And the time-consuming is less than t _thres , the direction is prohibited (red light is on), and if the accumulated time of this pass exceeds t _thres , the direction is prohibited (red light is on).

According to step 3, cyclic judgment is performed on vehicles, pedestrians and non-motor vehicles in the other direction of the intersection, where t _thres is a preset threshold for the release time. The implementation manner of the specific embodiment is:

When the direction of travel is prohibited in step 3, according to the corresponding traffic demand in max{c _{l_wn} ,c _{f_ew} +p _{f_s} +p _{f_n} }, if the east-west direction (west-east) requires a large amount of straight travel, then the direct travel in the direction (the green light for straight travel is on) , the red lights are on in other directions), and the zebra crossing on the north and south sides will pass. If the accumulated traffic in the east-west direction ends, and the time is less than t _thres , then the direct _driving in this direction is prohibited (the red light is on). ;

When the north-south (north-south), east-west (west-east), and south-to-west and north-to-east-central directions are prohibited, the green lights for left turns from west to north and east to south are on, and the red lights are on for other directions, including zebra crossings ; If the accumulated traffic in this direction ends and the time is less than t _thres , the direction is prohibited (the red light is on), and if the accumulated time of this traffic exceeds t _thres , the direction is prohibited (the red light is on).

Step 5: Iteratively execute the above steps until all directions in the current traffic cycle are completed, and a new traffic cycle begins.

The second technical solution adopted by the present invention: please refer to FIG. 2, a system including a real-time scheduling method for intersection traffic based on roadside perception as described in any of the above, including:

The roadside perception module is used to collect scene data information in the intersection area and fuse the target information; in a specific embodiment, a roadside perception module is installed at the intersections of solid-line roads in four directions in the south, south, and northwest. The installation position of the roadside sensing module is at the midpoint between the beginning of the pre-intersection and the center of the intersection, and is located 1m away from the road on the right side of the road, and the installation height is 3 to 4m above the ground. Its field of view covers all solid lanes of the current road pre-junction.

The traffic scheduling module is used to obtain the traffic requirements of different directions at the intersection and output the traffic scheduling strategy;

The traffic light module is used to receive the light-on strategy and light-on time output by the traffic scheduling module, and to indicate the traffic at the intersection;

The communication module is used for signal communication between the roadside sensing module, the traffic scheduling module and the traffic light module.

In some specific embodiments, the traffic scheduling module is respectively connected to the roadside sensing module and the traffic light module through a communication module. Specifically, the roadside perception module and the traffic scheduling module construct a local area network through the communication module, and the information interaction between the roadside perception module and the traffic scheduling module is realized by ROS. Select one of the embedded computing units as ROSmaster.

In some specific embodiments, the roadside perception module includes a high-definition camera for identifying objects such as vehicles, pedestrians, and non-motor vehicles on the road using an image object recognition algorithm, and a The data is extracted from the lidar of the target clustering within the current field of view and the embedded computing unit for loading the high-definition map of the intersection area.

Specifically, the roadside perception module includes a Sagitar Juchuang 128 first lidar RS-Ruby, three Amerson AR023Z high-definition cameras and an NVIDIA Xavier embedded computing unit. The lidar and the Xavier embedded computing unit are connected through a network port , the high-definition camera and the Xavier embedded computing unit are connected through the USB3.0 interface. The lidar is installed at a downward tilt, with a tilt angle of 25.647° to 29.036°. Three cameras are installed directly above the radar, covering the left, middle, and right parts of the current road respectively.

Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, and substitutions can be made in these embodiments without departing from the principle and spirit of the invention and modifications, the scope of the present invention is defined by the appended claims and their equivalents, and should be included within the protection scope of the present invention.

Claims (6)

1. A road side perception-based intersection traffic real-time scheduling method is characterized by comprising the following steps:

collecting corresponding scene information by using a road side unit;

identifying vehicles, pedestrians and non-motor vehicles in scene information through information fusion to perform target level fusion to obtain fused target information in the current scene, and the method specifically comprises the following steps:

collecting point cloud data of a target in scene information, segmenting, clustering, extracting characteristics and identifying the target, and extracting a target clustering target central point and anchor frame information: (x, y, l, w, h, c), wherein (x, y) is the abscissa and the ordinate of the target in the radar coordinate system, (l, w, h) is the length, width and height of the target, and c is the category of the target;

simultaneously, acquiring image data of a target in scene information, identifying vehicles, pedestrians and non-motor vehicles at the intersection by using a deep learning model, and extracting target center information and anchor frame information as follows: (x, y, l, w, c);

mapping target information in a radar coordinate system to an image, performing target-level fusion according to the intersection and comparison of anchor frames of the target information and the image, and outputting position and category information of a target;

according to the road information of the high-precision map and the fused target information, extracting the target information of corresponding regional lane levels, and counting the traffic demands of different courses of the intersection, the method specifically comprises the following steps:

loading a high-precision map of the intersection area, and acquiring target information after target-level fusion;

according to the high-precision map, mapping the fused target information to the current intersection area, and extracting the target information of all lanesAnd the statistics are as follows: (no, t, count)_left，count_forward，count_rightt，count_cross)；

Wherein no represents the current road side unit number, t represents the GPS time of the radar information acquisition time, (count)_left，count_forward，count_rightt，count_cross) Respectively representing the left turn demand count, the straight travel demand count and the right turn demand count of the vehicles in the current intersection area, and the demand counts of pedestrians and non-motor vehicles in the zebra crossing area;

and inputting the traffic demands of the intersection with different courses into a traffic scheduling algorithm, calculating to obtain a lighting strategy of the intersection, and feeding back and adjusting.

2. The method for real-time traffic crossing scheduling based on roadside perception according to claim 1, wherein the specific steps of extracting and counting the target information of all lanes comprise:

acquiring the number of lanes in different directions, and counting the passing demands of vehicles in different lanes;

counting the passing demands of left-turn vehicles:

counting the passing demands of straight-going vehicles:

right-turn vehicle traffic demand counting:

wherein, a, b and c are the independent number of lanes for straight running, left turning and right turning at the intersection, d is the number of lanes for collinear left turning and straight running, and e is the number of lanes for collinear right turning and straight running.

3. The method for real-time traffic dispatching at the intersection based on roadside awareness as claimed in claim 2, wherein the specific steps of inputting traffic demands of the intersection with different headings into a traffic dispatching algorithm, calculating a lighting strategy of the intersection, and performing feedback regulation comprise:

the method comprises the following steps: acquiring the statistical results of the target information of lane levels of different courses of the intersection;

step two: firstly, determining that signals of bidirectional straight traffic lights are consistent with signals of zebra crossing traffic lights on two sides in the same direction;

step three: when a passing period begins, selecting the direction with the largest passing demand as the current passing direction according to the target information statistical result, deleting the direction from the passing demand list, forbidding the passing in other passing directions, and simultaneously when the passing of the vehicles in the direction is finished and the consumed time is less than a preset threshold value t_thresOr when the vehicle passing time exceeds the preset threshold t_thresIf so, the direction is disabled;

step four: in the current passing period, selecting the direction with the largest demand from the passing list as the passing direction, forbidding in other directions, and simultaneously when the passing of the vehicles in the direction is finished and the consumed time is less than a preset threshold value t_thresOr when the vehicle passing time exceeds the preset threshold t_thresIf so, the direction is disabled;

step five: the iteration is carried out until all the directions in the current passing period are passed, and a new passing period starts t_thresA threshold is preset for the release time.

4. A system comprising a road side perception-based intersection traffic real-time scheduling method according to any one of claims 1 to 3, comprising:

the roadside sensing module is used for acquiring scene data information of the intersection area and fusing target information;

the traffic scheduling module is used for acquiring traffic requirements of different courses of the intersection and outputting a traffic scheduling strategy;

the traffic light module is used for receiving the light-up strategy and the light-up time output by the traffic scheduling module and performing intersection traffic indication;

and the communication module is used for signal communication among the road side sensing module, the traffic scheduling module and the traffic light module.

5. The system of claim 4, wherein the traffic scheduling module is respectively connected to the roadside sensing module and the traffic light module through communication modules.

6. The system of claim 5, wherein the roadside awareness module comprises a high-definition camera, a laser radar and an embedded computing unit.

Images