Driving route decision-making method based on unmanned vehicle

Abstract

The invention relates to the technical field of driving route decision making, discloses a driving route decision making method based on an unmanned vehicle, and aims to solve the problem that the driving route of the unmanned vehicle is not intelligent and humanized enough during decision making at present. According to the method, the congestion type is quickly identified, the corresponding decision strategy is triggered, the route is dynamically adjusted, when the vehicle encounters the congestion condition, classified discussion can be carried out, the full congestion scene and the gap passable scene can be further distinguished so as to execute different decision strategies, the transverse gap width is smaller than the vehicle width + 0.6 m, the front vehicle is static for more than 30 seconds, and the full congestion state is achieved. The transverse gap is larger than or equal to the vehicle width + 0.6 m, the longitudinal gap distance is larger than the vehicle length + 2 m, the gap passable state is achieved, and decision-making intelligence is provided.

Description

Technical Field

[0001] The present invention relates to the technical field of driving route decision-making, and in particular to a driving route decision-making method based on an unmanned vehicle. Background Art

[0002] With the advancement of artificial intelligence and sensor technology, autonomous vehicles are gradually achieving autonomous driving in complex traffic scenarios through the coordinated integration of environmental perception, path planning, and control execution. Existing driving route decision-making methods typically include environmental perception, global path planning, local trajectory generation, control execution, and emergency response.

[0003] In view of the above-mentioned and existing related technologies, the inventors believe that the following defects often exist: In the existing technology, when an unmanned vehicle encounters a traffic jam during driving, it is impossible to judge the severity of the traffic jam. When a car stops in front, the unmanned vehicle will naturally stop. However, in some traffic jams, such as when the vehicle in front is performing a reverse parking operation, it will temporarily occupy the road, but it can be overtaken by using the road. However, the current unmanned vehicles will be judged as "unable to pass", resulting in the unmanned vehicle waiting in vain at the back, which is a waste of time and has low intelligence. Summary of the Invention

[0004] The technical problem to be solved by the present invention is that the existing technology has the disadvantage that the driving route decision-making of unmanned vehicles is not intelligent and humane enough. To this end, we propose a driving route decision-making method based on unmanned vehicles.

[0005] To achieve the above objectives, the present application adopts the following technical solution: a driving route decision method based on an unmanned vehicle, comprising the following steps: S1: Acquire the vehicle's surrounding environment information in real time and build an environmental model; S2: Based on the starting point and the end point, generate an initial feasible route that avoids static obstacles; S3: Dynamically adjust the global path based on the real-time environment to generate a safe and comfortable local driving trajectory; S4: Convert the planned trajectory into a control signal that can be executed by the vehicle; S5: Respond to emergencies, quickly identify congestion types, trigger corresponding decision-making strategies, and dynamically adjust routes.

[0006] Furthermore, the step S1 specifically includes the following steps: S11: A lidar is deployed on the roof to generate 3D point cloud data covering a 200m radius around the vehicle. A combination of forward-, side-, and surround-view cameras are deployed on the front windshield, left and right rearview mirrors, and the front logo to identify lane markings, traffic signals, pedestrians, and vehicles ahead for visual perception and path planning. It also monitors vehicles passing alongside, blind spots, and lane edges to assist with lane changes and parking. It also provides a 360° panoramic image for parking assistance, automatic parking, and surrounding environment modeling. A fisheye correction algorithm is used to eliminate distortion. A millimeter-wave radar is installed inside the front bumper of the vehicle, using radio waves to detect target distance, speed, and angle. It has strong penetration capabilities in rain, fog, and dust, making it suitable for all-weather perception. Twelve ultrasonic radars are deployed around the vehicle for low-speed obstacle detection. S12: Time synchronization and spatial calibration of the collected data are performed using the PTP protocol to align the camera and radar timestamps based on the lidar scanning period. The rotation matrix R and translation vector T are calculated using the Zhang calibration method and the hand-eye calibration method. S13: Filter out noise and invalid information in the original data to improve data quality and reduce the complexity of subsequent algorithm processing; S14: Integrate multi-source data to generate an environmental model that includes the locations of static obstacles and the real-time trajectories of dynamic targets, providing a unified environmental representation for subsequent path planning.

[0007] Furthermore, the step S2 specifically includes the following steps: S21: Acquires lane-level road information and real-time vehicle location to provide geographic references and constraints for global path planning, enabling map loading and positioning fusion. Map loading includes loading lane centerline coordinates, traffic rules, and real-time congestion data. Positioning fusion outputs vehicle posture through RTK-GNSS and vision / laser SLAM. S22: Based on the starting point, the end point and real-time traffic data, a coarse-grained route is generated to avoid static obstacles and the overall driving direction is determined; S23: Decompose the coarse-grained route into a specific lane change sequence, clarify the target lane and driving speed for each section, provide detailed guidance for local trajectory planning, and output a reference path with lane information.

[0008] Furthermore, the step S3 specifically includes the following steps: S31: Predict the movement trajectories of pedestrians and vehicles in the next 3-5 seconds, assess potential collision risks, and provide evidence for behavioral decision-making; S32: Quantify risk through collision time and lateral safety distance indicators, triggering corresponding obstacle avoidance or yield strategies; S33: Based on the risk assessment results, the optimal driving strategy for the current scenario is selected and the local trajectory planning goal is clarified. The state priority is: emergency vehicle > pedestrian crossing > slow-moving vehicle in front > normal driving. If a pedestrian is detected entering the vehicle's lane and the collision time is less than 3 seconds, the vehicle is forced to enter the "stop state" and other low-priority instructions are ignored. S34: Generates safe and comfortable driving trajectories that comply with vehicle dynamics constraints, ensuring smooth passage through complex scenarios.

[0009] Furthermore, the step S4 specifically includes the following steps: S41: Calculate steering commands based on the planned trajectory to ensure that the vehicle follows the target path and compensate for understeering or oversteering characteristics of the vehicle; S42: Generate throttle / brake commands based on the target speed to control vehicle acceleration and deceleration to ensure that the speed is stable within a safe range; S43: Ensures control commands are within the vehicle hardware capabilities, avoids oversteering or sudden acceleration / braking, and switches to safe mode if a sensor fails.

[0010] Furthermore, the step S5 specifically includes the following steps: S51: Identify road congestion through vehicle density and speed indicators, distinguish between normal driving, slow traffic and severe congestion, and provide a basis for subsequent classification decisions; S52: Further distinguish between “completely blocked” and “passable” scenarios to implement different decision strategies; S53: Execute corresponding strategies for different congestion types to ensure a balance between safety and efficiency. Wait or reroute when completely blocked. Calculate feasibility and attempt to pass when there is a gap. In completely congested state: Activate automatic parking, maintain a distance of 2 meters between vehicles, request traffic conditions every 20 seconds, and trigger rerouting after more than 5 minutes; Gap passable state: Calculation of longitudinal safety time window ,satisfy When the speed is 10km / h, the gap width changes are monitored in real time. is the longitudinal available distance of the gap, is the length of the vehicle, is the planned passing speed of the vehicle, For the car in front The displacement change within For the following car Internal displacement changes.

[0011] S54: When local adjustments cannot resolve congestion, re-search the global path to avoid congestion, and optimize the decision model through historical data to improve the ability to respond to future scenarios.

[0012] Furthermore, in step S13, voxel filtering and statistical filtering are used to process the data collected by the lidar, the point cloud density is optimized to 10 points / cube, median filtering and histogram equalization are used to process the data collected by the camera to improve the low-light scene recognition accuracy, and extended Kalman filtering is used to process the data collected by the millimeter-wave radar.

[0013] Furthermore, in step S14, the multi-source data integration includes early fusion and late fusion. In the early fusion, the PointPainting technology is used to map the camera semantic labels to the lidar point cloud, and the PointPillars network is input to detect the target. In the late fusion, the Hungarian algorithm is used to associate the multi-sensor detection results to generate a dynamic target list with ID, where the list includes position, speed, and predicted trajectory.

[0014] Furthermore, in S23, based on the lane connection relationship in the map, a lane sequence is generated to ensure lane change advance and avoidance of restricted driving periods. The advance amount in high-speed scenarios is 200m, and the advance amount on urban roads is 50m. For example, bus lanes are prohibited from driving from 7:00 to 9:00.

[0015] Furthermore, in S51, density detection is performed: on a highway, if the number of vehicles within 50m ahead is greater than 15, and on a city road, if the number of vehicles within 50m ahead is greater than 8, it is determined to be slow passage; speed detection is performed: on a city road, if the average vehicle speed for 30 consecutive seconds is less than 5km / h, and on a highway, if the average vehicle speed for 30 consecutive seconds is less than 20km / h, it is determined to be slow passage.

[0016] The technical effects and advantages of the present invention are as follows: In the present invention, S1: real-time acquisition of vehicle surrounding environment information and construction of an environmental model; S2: based on the starting point and the end point, an initial feasible route avoiding static obstacles is generated; S3: the global path is dynamically adjusted according to the real-time environment to generate a safe and comfortable local driving trajectory; S4: the planned trajectory is converted into a control signal executable by the vehicle; S5: to deal with emergency scenarios, quickly identify the type of congestion and trigger the corresponding decision-making strategy, and dynamically adjust the route. When the vehicle encounters a congestion, it can be classified and discussed to further distinguish between "complete congestion" and "gap passable" scenarios in order to execute different decision-making strategies. If the lateral gap width is less than the vehicle width + 0.6m and the front vehicle is stationary for more than 30 seconds, it is a complete congestion state. If the lateral gap is greater than or equal to the vehicle width + 0.6m and the longitudinal gap distance is greater than the vehicle length + 2m, it is a gap passable state, thereby improving the intelligence of decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The disclosure of the present invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of the present invention. In the drawings, the same reference numerals are used to refer to the same components: Figure 1 It is the overall flow chart of the present invention. DETAILED DESCRIPTION

[0018] It is easy to understand that according to the technical solution of the present invention, without changing the essential spirit of the present invention, a person skilled in the art can propose a variety of interchangeable structural modes and implementation modes. Therefore, the following specific embodiments and drawings are only exemplary descriptions of the technical solution of the present invention and should not be regarded as the entire invention or as a limitation or restriction of the technical solution of the present invention.

[0019] Reference Figure 1 As shown, the present invention provides a technical solution: a driving route decision method based on an unmanned vehicle, comprising the following steps: S1: Acquire the vehicle's surrounding environment information in real time and build an environmental model; S2: Based on the starting point and the end point, generate an initial feasible route that avoids static obstacles; S3: Dynamically adjust the global path based on the real-time environment to generate a safe and comfortable local driving trajectory; S4: Convert the planned trajectory into a control signal that can be executed by the vehicle; S5: Respond to emergencies, quickly identify congestion types, trigger corresponding decision-making strategies, and dynamically adjust routes.

[0020] Reference Figure 1 As shown, in this embodiment: in step S1, the following steps are specifically included: S11: A lidar is deployed on the roof to generate 3D point cloud data covering a 200m radius around the vehicle. A combination of forward-, side-, and surround-view cameras are deployed on the front windshield, left and right rearview mirrors, and the front logo to identify lane markings, traffic signals, pedestrians, and vehicles ahead for visual perception and path planning. It also monitors vehicles passing alongside, blind spots, and lane edges to assist with lane changes and parking. It also provides a 360° panoramic image for parking assistance, automatic parking, and surrounding environment modeling. A fisheye correction algorithm is used to eliminate distortion. A millimeter-wave radar is installed inside the front bumper of the vehicle, using radio waves to detect target distance, speed, and angle. It has strong penetration capabilities in rain, fog, and dust, making it suitable for all-weather perception. Twelve ultrasonic radars are deployed around the vehicle for low-speed obstacle detection. S12: Time synchronization and spatial calibration of the collected data are performed using the PTP protocol to align the camera and radar timestamps based on the lidar scanning period. The rotation matrix R and translation vector T are calculated using the Zhang calibration method and the hand-eye calibration method. S13: Filter out noise and invalid information in the original data to improve data quality and reduce the complexity of subsequent algorithm processing; S14: Integrate multi-source data to generate an environmental model that includes the locations of static obstacles and the real-time trajectories of dynamic targets, providing a unified environmental representation for subsequent path planning.

[0021] In step S2, the following steps are specifically included: S21: Acquires lane-level road information and real-time vehicle location to provide geographic references and constraints for global path planning, enabling map loading and positioning fusion. Map loading includes loading lane centerline coordinates, traffic rules, and real-time congestion data. Positioning fusion outputs vehicle posture through RTK-GNSS and vision / laser SLAM. S22: Based on the starting point, the end point and real-time traffic data, a coarse-grained route is generated to avoid static obstacles and the overall driving direction is determined; S23: Decompose the coarse-grained route into a specific lane change sequence, clarify the target lane and driving speed for each section, provide detailed guidance for local trajectory planning, and output a reference path with lane information.

[0022] In step S3, the following steps are specifically included: S31: Predict the movement trajectories of pedestrians and vehicles in the next 3-5 seconds, assess potential collision risks, and provide evidence for behavioral decision-making; S32: Quantify risk through collision time and lateral safety distance indicators, triggering corresponding obstacle avoidance or yield strategies; Specifically, for longitudinal risk, the collision time = longitudinal distance / relative speed. The safety threshold collision time is greater than 2 seconds. If it is lower than the threshold, deceleration or parking will be triggered. For lateral risk, if the lateral distance is less than 0.8m, an early warning will be triggered. If it is less than 0.5m, an emergency brake will be triggered. S33: Based on the risk assessment results, the optimal driving strategy for the current scenario is selected and the local trajectory planning goal is clarified. The state priority is: emergency vehicle > pedestrian crossing > slow-moving vehicle in front > normal driving. If a pedestrian is detected entering the vehicle's lane and the collision time is less than 3 seconds, the vehicle is forced to enter the "stop state" and other low-priority instructions are ignored. S34: Generates safe and comfortable driving trajectories that comply with vehicle dynamics constraints, ensuring smooth passage through complex scenarios.

[0023] In step S4, the following steps are specifically included: S41: Calculate steering commands based on the planned trajectory to ensure that the vehicle follows the target path and compensate for understeering or oversteering characteristics of the vehicle; S42: Generate throttle / brake commands based on the target speed to control vehicle acceleration and deceleration to ensure that the speed is stable within a safe range; S43: Ensures control commands are within the vehicle hardware capabilities, avoids oversteering or sudden acceleration / braking, and switches to safe mode if a sensor fails.

[0024] In step S5, the following steps are specifically included: S51: Identify road congestion through vehicle density and speed indicators, distinguish between normal driving, slow traffic and severe congestion, and provide a basis for subsequent classification decisions; S52: Further distinguish between “completely blocked” and “passable” scenarios to implement different decision strategies; If the lateral gap is less than the vehicle width + 0.6m and the vehicle ahead is stationary for more than 30 seconds, it is considered a complete blockage. The lateral clearance is greater than or equal to the vehicle width + 0.6m, and the longitudinal clearance is greater than the vehicle length + 2m, which means the vehicle is passable. S53: Execute corresponding strategies for different congestion types to ensure a balance between safety and efficiency. Wait or reroute when completely blocked. Calculate feasibility and attempt to pass when there is a gap. In completely congested state: Activate automatic parking, maintain a distance of 2 meters between vehicles, request traffic conditions every 20 seconds, and trigger rerouting after more than 5 minutes; Gap passable state: Calculation of longitudinal safety time window ,satisfy When the speed is 10km / h, the gap width changes are monitored in real time. is the longitudinal available distance of the gap, is the length of the vehicle, is the planned passing speed of the vehicle, For the car in front The displacement change within For the following car Internal displacement changes.

[0025] S54: When local adjustments cannot resolve congestion, re-search the global path to avoid congestion, and optimize the decision model through historical data to improve the ability to respond to future scenarios.

[0026] In step S13, voxel filtering and statistical filtering are used to process the data collected by the lidar, and the point cloud density is optimized to 10 points / cube. Median filtering and histogram equalization are used to process the data collected by the camera to improve the low-light scene recognition accuracy. Extended Kalman filtering is used to process the data collected by the millimeter-wave radar.

[0027] In step S14, multi-source data integration includes early fusion and late fusion. In early fusion, PointPainting technology is used to map camera semantic labels to lidar point clouds and input them into the PointPillars network to detect targets. In late fusion, the Hungarian algorithm is used to associate multi-sensor detection results to generate a dynamic target list with ID, which includes position, speed, and predicted trajectory.

[0028] In S23, a lane sequence is generated based on the lane connection relationship on the map to ensure lane change advance and avoidance of restricted driving periods. The advance amount in high-speed scenarios is 200m, and the advance amount in urban roads is 50m. For example, bus lanes are prohibited from driving from 7:00 to 9:00.

[0029] In S51, density detection: on a highway, if the number of vehicles within 50m ahead is greater than 15, on a city road, if the number of vehicles within 50m ahead is greater than 8, it is determined to be slow traffic. Speed ​​detection: on a city road, if the average speed for 30 consecutive seconds is less than 5km / h, on a highway, if the average speed for 30 consecutive seconds is less than 20km / h, it is determined to be slow traffic. Working principle: When an unmanned vehicle is driving, it collects data through multi-source sensors, obtains real-time information about the vehicle's surrounding environment, builds an environmental model, obtains lane-level road information and the vehicle's real-time position, provides geographical references and constraints for global path planning, realizes map loading and positioning fusion, and generates an initial feasible route that avoids static obstacles based on the starting point and end point. At the same time, it dynamically adjusts the global path according to the real-time environment to generate a safe and comfortable local driving trajectory, and then converts the planned trajectory into a control signal that can be executed by the vehicle. It identifies road congestion through vehicle density and speed indicators, and distinguishes between normal driving, slow traffic and severe congestion. In addition, it further distinguishes between "complete congestion" and "gap passable" scenarios in order to implement different decision-making strategies. Vehicles in a "complete congestion" state cannot move and can only park and brake. Vehicles in a "gap passable" state can try to pass, saving time and improving the level of intelligence.

[0030] The technical scope of the present invention is not limited to the contents of the above description. Those skilled in the art can make various deformations and modifications to the above embodiments without departing from the technical idea of ​​the present invention, and these deformations and modifications should all fall within the protection scope of the present invention.

Claims

1. A driving route decision method based on an unmanned vehicle, characterized in that: The following steps are involved: S1: Acquire the vehicle's surrounding environment information in real time and build an environmental model; S2: Based on the starting point and the end point, generate an initial feasible route that avoids static obstacles; S3: Dynamically adjust the global path based on the real-time environment to generate a safe and comfortable local driving trajectory; S4: Convert the planned trajectory into a control signal that can be executed by the vehicle; S5: Respond to emergencies, quickly identify congestion types, trigger corresponding decision-making strategies, and dynamically adjust routes.

2. The driving route decision method based on an unmanned vehicle according to claim 1, characterized in that: The step S1 specifically includes the following steps: S11: A lidar is deployed on the roof to generate 3D point cloud data covering a 200m radius around the vehicle. A combination of forward-, side-, and surround-view cameras are deployed on the front windshield, left and right rearview mirrors, and the front logo to identify lane markings, traffic signals, pedestrians, and vehicles ahead for visual perception and path planning. It also monitors vehicles passing alongside, blind spots, and lane edges to assist with lane changes and parking. It also provides a 360° panoramic image for parking assistance, automatic parking, and surrounding environment modeling. A fisheye correction algorithm is used to eliminate distortion. A millimeter-wave radar is installed inside the front bumper of the vehicle, using radio waves to detect target distance, speed, and angle. It has strong penetration capabilities in rain, fog, and dust, making it suitable for all-weather perception. Twelve ultrasonic radars are deployed around the vehicle for low-speed obstacle detection. S12: Time synchronization and spatial calibration of the collected data are performed using the PTP protocol to align the camera and radar timestamps based on the lidar scanning period. The rotation matrix R and translation vector T are calculated using the Zhang calibration method and the hand-eye calibration method. S13: Filter out noise and invalid information in the original data to improve data quality and reduce the complexity of subsequent algorithm processing; S14: Integrate multi-source data to generate an environmental model that includes the locations of static obstacles and the real-time trajectories of dynamic targets, providing a unified environmental representation for subsequent path planning.

3. The driving route decision method based on an unmanned vehicle according to claim 1, characterized in that: The step S2 specifically includes the following steps: S21: Acquires lane-level road information and real-time vehicle location to provide geographic references and constraints for global path planning, enabling map loading and positioning fusion. Map loading includes loading lane centerline coordinates, traffic rules, and real-time congestion data. Positioning fusion outputs vehicle posture through RTK-GNSS and vision / laser SLAM. S22: Based on the starting point, the end point and real-time traffic data, a coarse-grained route is generated to avoid static obstacles and the overall driving direction is determined; S23: Decompose the coarse-grained route into a specific lane change sequence, clarify the target lane and driving speed for each section, provide detailed guidance for local trajectory planning, and output a reference path with lane information.

4. The driving route decision method based on an unmanned vehicle according to claim 1, characterized in that: The step S3 specifically includes the following steps: S31: Predict the movement trajectories of pedestrians and vehicles in the next 3-5 seconds, assess potential collision risks, and provide evidence for behavioral decision-making; S32: Quantify risk through collision time and lateral safety distance indicators, triggering corresponding obstacle avoidance or yield strategies; S33: Based on the risk assessment results, the optimal driving strategy for the current scenario is selected and the local trajectory planning goal is clarified. The state priority is: emergency vehicle > pedestrian crossing > slow-moving vehicle in front > normal driving. If a pedestrian is detected entering the vehicle's lane and the collision time is less than 3 seconds, the vehicle is forced to enter the "stop state" and other lower-priority instructions are ignored. S34: Generates safe and comfortable driving trajectories that comply with vehicle dynamics constraints, ensuring smooth passage through complex scenarios.

5. The driving route decision method based on an unmanned vehicle according to claim 1, characterized in that: The step S4 specifically includes the following steps: S41: Calculate steering commands based on the planned trajectory to ensure that the vehicle follows the target path and compensate for understeering or oversteering characteristics of the vehicle; S42: Generate throttle / brake commands based on the target speed to control vehicle acceleration and deceleration to ensure that the speed is stable within a safe range; S43: Ensures control commands are within the vehicle hardware capabilities, avoids oversteering or sudden acceleration / braking, and switches to safe mode if a sensor fails.

6. The driving route decision method based on an unmanned vehicle according to claim 1, characterized in that: The step S5 specifically includes the following steps: S51: Identify road congestion through vehicle density and speed indicators, distinguish between normal driving, slow traffic and severe congestion, and provide a basis for subsequent classification decisions; S52: Further distinguish between "completely blocked" and "passable" scenarios to implement different decision strategies; S53: Execute corresponding strategies for different congestion types to ensure a balance between safety and efficiency. Wait or reroute when completely blocked. Calculate feasibility and attempt to pass when there is a gap. In completely congested state: Activate automatic parking, maintain a distance of 2 meters between vehicles, request traffic conditions every 20 seconds, and trigger rerouting after more than 5 minutes; Gap passable state: Calculation of longitudinal safety time window ,satisfy When the speed is 10km / h, the gap width changes are monitored in real time. is the longitudinal available distance of the gap, is the length of the vehicle, is the planned passing speed of the vehicle, For the car in front The displacement change within For the following car Displacement changes within S54: When local adjustments cannot resolve congestion, re-search the global path to avoid congestion, and optimize the decision model through historical data to improve the ability to respond to future scenarios.

7. The driving route decision method based on an unmanned vehicle according to claim 2, characterized in that: In step S13, voxel filtering and statistical filtering are used to process the data collected by the lidar, and the point cloud density is optimized to 10 points / cube. Median filtering and histogram equalization are used to process the data collected by the camera to improve the low-light scene recognition accuracy. Extended Kalman filtering is used to process the data collected by the millimeter-wave radar.

8. The driving route decision method based on an unmanned vehicle according to claim 2, characterized in that: In step S14, the multi-source data integration includes early fusion and late fusion. In the early fusion, the PointPainting technology is used to map the camera semantic labels to the lidar point cloud, and the PointPillars network is input to detect the target. In the late fusion, the Hungarian algorithm is used to associate the multi-sensor detection results to generate a dynamic target list with ID, where the list includes position, speed, and predicted trajectory.

9. The driving route decision method based on an unmanned vehicle according to claim 3, characterized in that: In S23, based on the lane connection relationship in the map, a lane sequence is generated to ensure lane change lead time and avoidance of restricted traffic periods. The lead time in high-speed scenarios is 200m, and the lead time on urban roads is 50m.

10. The driving route decision method based on an unmanned vehicle according to claim 6, characterized in that: In S51, density detection: on the highway, the number of vehicles within 50m ahead is greater than 15, and on the city road, the number of vehicles within 50m ahead is greater than 8, which is judged as slow passage. Speed ​​detection: on the city road, the average vehicle speed for 30 consecutive seconds is less than 5km / h, and on the highway, the average vehicle speed for 30 consecutive seconds is less than 20km / h, which is judged as slow passage.

Images