A non-motor vehicle swarm wisdom flow state induction method and system based on microwave radar

Abstract

The application discloses a non-motor vehicle group wisdom flow state induction method and system based on a microwave radar, and belongs to the technical field of intelligent transportation. The method comprises the following steps: obtaining microwave radar echo data in a non-motor vehicle monitoring area of an intersection; performing target identification on the echo data to generate non-motor vehicle individual target parameters; aggregating the individual target parameters in a current period to generate non-motor vehicle group wisdom flow state parameters; inputting the non-motor vehicle group wisdom flow state parameters into a flow state evolution prediction model to generate predicted traffic demand parameters; generating and executing an adaptive induction strategy according to the predicted traffic demand parameters and traffic signal lamp control information to adjust non-motor vehicle signal lamp timing. Through microwave radar sensing, group wisdom flow state parameter construction and short-time traffic demand prediction, the application realizes forward-looking identification and active induction of non-motor vehicle group traffic demand, improves non-motor vehicle traffic efficiency at the intersection, and reduces the risk of stagnation, rush and traffic conflicts.

Description

Technical Field

[0001] This invention relates to the field of intelligent transportation technology, and in particular to a non-motorized vehicle swarm intelligent flow guidance method and system based on microwave radar. Background Technology

[0002] At urban road intersections, non-motorized traffic flows such as bicycles and electric bicycles are characterized by high randomness, significant group aggregation, large speed differences, and inconsistent starting behaviors. Traditional traffic signal control systems typically rely on motor vehicle detection data as their primary input, lacking the ability to accurately perceive the spatial aggregation of non-motorized groups, speed coordination, and traffic pressure approaching the stop line.

[0003] Existing non-motorized vehicle detection solutions mostly employ video recognition, inductive loop detectors, or manual timing based on experience. Video recognition is easily affected by factors such as low light at night, rain, snow, fog, haze, backlight, and obstructions; inductive loop detectors can only detect local occupancy and are difficult to obtain continuous position, speed, heading, and group evolution trends; fixed signal timing methods cannot be dynamically adjusted according to the real-time size and arrival trends of non-motorized vehicle groups. When the green light time for non-motorized vehicles is insufficient, congestion and rushing at the end of the green light can easily occur in the waiting area; when the green light time is too long, it can lead to phase idling and reduce the overall operational efficiency of the intersection.

[0004] Therefore, how to achieve all-weather, continuous, and reliable perception of non-motorized vehicle groups in complex environments, and how to convert discrete individual target information into group flow parameters that can be used for signal control, further predict short-term traffic demand, and generate forward-looking adaptive guidance strategies are technical problems that need to be solved in the field of intelligent traffic intersection control. Summary of the Invention

[0005] To address the aforementioned issues, this invention provides a non-motorized vehicle crowd intelligent flow guidance method and system based on microwave radar. Through microwave radar perception, crowd intelligent flow parameter construction, and short-term traffic demand prediction, it achieves forward-looking identification and proactive guidance of the traffic demand of non-motorized vehicle groups, improving the traffic efficiency of non-motorized vehicles at intersections and reducing the risks of congestion, overtaking, and traffic conflicts.

[0006] The above objectives can be achieved through the following approach: A non-motorized vehicle crowd intelligent flow guidance method based on microwave radar includes: acquiring signals collected by microwave radar deployed at an intersection within the non-motorized vehicle monitoring area, and generating microwave radar echo data; processing the microwave radar echo data to identify individual non-motorized vehicle targets and generate individual non-motorized vehicle target parameters for each target; aggregating all the individual non-motorized vehicle target parameters within the current period to quantify the overall movement characteristics of the group and generate non-motorized vehicle crowd intelligent flow parameters; inputting the non-motorized vehicle crowd intelligent flow parameters into a flow evolution prediction model to predict the future traffic trend of the group and generate predicted traffic demand parameters; generating an adaptive guidance strategy based on the predicted traffic demand parameters and combined with real-time traffic signal control information; and executing the adaptive guidance strategy to adjust the timing scheme of the non-motorized vehicle traffic signals.

[0007] Optionally, the step of processing the microwave radar echo data to identify individual non-motorized vehicle targets and generate individual non-motorized vehicle target parameters for each target includes: performing time-frequency analysis and filtering on the microwave radar echo data to extract effective target signals and generate radar point clusters; applying a spatial clustering algorithm to the radar point clusters to distinguish different physical entities and generate multiple target clusters; and performing state estimation on each target cluster to calculate its kinematic information and generate the individual non-motorized vehicle target parameters containing position, velocity, and heading information.

[0008] Optionally, the aggregation of all individual non-motorized vehicle target parameters within the current period to quantify the overall motion characteristics of the group and generate non-motorized vehicle swarm intelligence flow parameters includes: extracting the position coordinates of all targets from the individual non-motorized vehicle target parameters to construct the group spatial distribution and generate a group position dataset; calculating the average distance and distribution density between targets based on the group position dataset to assess the degree of group aggregation and generate a group density index; extracting the velocity vectors of all targets from the individual non-motorized vehicle target parameters, calculating the group average velocity and velocity variance to assess the group's velocity coordination and generate a group velocity dispersion index; and fusing the group density index and the group velocity dispersion index to form a comprehensive description of the group's current state and generate the non-motorized vehicle swarm intelligence flow parameters.

[0009] Optionally, the step of inputting the non-motorized vehicle crowd intelligent flow parameters into a flow evolution prediction model to estimate the future traffic trend of the group and generate predicted traffic demand parameters includes: inputting the non-motorized vehicle crowd intelligent flow parameters as an initial state into an evolution model based on time series analysis to predict its state at the next time step and generate future flow characteristics; calculating the time for the group's centroid to reach the stop line based on the future flow characteristics and the geometric position of the intersection stop line, and generating a predicted arrival time window; and combining the group size information in the future flow characteristics with the predicted arrival time window to quantify traffic pressure and generate the predicted traffic demand parameters.

[0010] Optionally, generating an adaptive guidance strategy based on the predicted traffic demand parameters and real-time traffic signal control information includes: requesting and receiving current traffic light phase and duration information from the traffic signal controller to generate the current traffic light phase state; comparing the predicted traffic demand parameters with the current traffic light phase state to determine whether the existing timing scheme meets the predicted demand, and generating a phase adjustment decision; and calculating the specific green light duration extension or phase start advance based on the phase adjustment decision to formulate an optimization scheme and generate the adaptive guidance strategy.

[0011] Optionally, before generating the adaptive guidance strategy, the method further includes: identifying non-motorized vehicle groups in other directions where there is a traffic conflict, and generating corresponding predicted traffic demand parameters for the groups in the conflicting directions to obtain the traffic demand parameters of the conflicting parties; establishing a right-of-way decision model, inputting the predicted traffic demand parameters of the current direction and the traffic demand parameters of the conflicting parties into the model simultaneously to perform priority judgment and generate a priority ranking result; adjusting the phase adjustment decision based on the priority ranking result to ensure that the traffic demand of high-priority directions is met first, and generating the final adaptive guidance strategy based on the adjusted decision.

[0012] Optionally, after generating the predicted traffic demand parameters, the method further includes: continuously acquiring and processing microwave radar data after executing the guidance strategy to obtain the actual situation of non-motorized vehicle groups crossing the stop line, generating actual traffic flow data; comparing the actual traffic flow data with the predicted traffic demand parameters generated before guidance to quantify the accuracy of the prediction model and generate a prediction model bias; using the prediction model bias, adjusting the internal parameters of the flow evolution prediction model through a feedback mechanism to improve the accuracy of subsequent predictions, generating a self-learning prediction model.

[0013] Optionally, acquiring signals collected by the microwave radar deployed at the intersection within the non-motorized vehicle monitoring area and generating microwave radar echo data includes: defining a dynamic key monitoring area covering the waiting area and upstream road segment based on the physical layout of the intersection and the direction of the non-motorized vehicle lanes; dynamically adjusting the radar's transmission power and beam scanning mode based on historical traffic flow data and real-time detected target density to focus on the dynamic key monitoring area, generating an optimized beam scanning strategy; and instructing the microwave radar to perform detection according to the optimized beam scanning strategy to enhance the signal-to-noise ratio of targets within the key area, thereby generating the microwave radar echo data.

[0014] Optionally, after generating the individual target parameters of the non-motorized vehicles, the method further includes: performing time-series trajectory tracking on the individual target parameters of the non-motorized vehicles for multiple consecutive periods to establish the motion history of each target and generate individual historical trajectory data; analyzing the individual historical trajectory data to identify whether there are behaviors that deviate from the normal driving mode, such as sudden changes in speed or direction, and generating abnormal motion trajectory identifiers; when the abnormal motion trajectory identifiers are detected, generating a high-priority emergency safety strategy and using it to replace the originally planned adaptive guidance strategy to deal with potential traffic safety incidents.

[0015] Based on the same inventive concept, this invention also provides a non-motorized vehicle crowd intelligent flow guidance system based on microwave radar. The system includes: an acquisition module for acquiring signals collected by microwave radar deployed at an intersection within the non-motorized vehicle monitoring area, generating microwave radar echo data; a processing module for processing the microwave radar echo data to identify individual non-motorized vehicle targets and generate individual non-motorized vehicle target parameters for each target; an aggregation module for aggregating all the individual non-motorized vehicle target parameters within the current period to quantify the overall movement characteristics of the group and generate non-motorized vehicle crowd intelligent flow parameters; a prediction module for inputting the non-motorized vehicle crowd intelligent flow parameters into a flow evolution prediction model to predict the future traffic trend of the group and generate predicted traffic demand parameters; a strategy generation module for generating an adaptive guidance strategy based on the predicted traffic demand parameters and combined with real-time acquired traffic signal control information; and an execution module for executing the adaptive guidance strategy to adjust the timing scheme of the non-motorized vehicle traffic signals.

[0016] Compared with the prior art, the present invention has the following advantages: This invention uses microwave radar to continuously detect non-motorized vehicle monitoring areas at intersections. It can acquire distance, speed, and angle information of non-motorized vehicle targets in complex environments such as night, rain, snow, and fog, thereby improving the all-weather adaptability and data reliability of non-motorized vehicle traffic flow perception.

[0017] This invention generates non-motorized vehicle swarm intelligence flow parameters by aggregating and analyzing individual target parameters of non-motorized vehicles. It converts the position, speed and heading information of discrete targets into macroscopic features such as group density, average speed and speed dispersion, thereby realizing a quantitative description of the spatial aggregation degree and speed coordination of non-motorized vehicle groups.

[0018] This invention uses a flow evolution prediction model to make short-term predictions of the future traffic trends of non-motorized vehicle groups, and combines the geometric position of the stop line to generate predicted traffic demand parameters. This enables traffic signal control to no longer rely solely on current detection results, but to respond in advance to non-motorized vehicle groups that are about to reach the stop line.

[0019] This invention compares predicted traffic demand parameters with traffic signal phase status to generate green light extension or phase start advance, enabling non-motorized vehicle signal timing to be dynamically adjusted according to real-time traffic pressure, reducing waiting area congestion, rushing at the end of the green light, and phase idling.

[0020] This invention introduces conflict direction traffic demand parameters and right-of-way decision-making models to prioritize non-motorized vehicle groups from multiple directions when they approach an intersection simultaneously. This ensures that the traffic demand of high-priority directions is met first, which helps reduce the potential conflict risk between non-motorized vehicles, motorized vehicles, and pedestrians at intersections.

[0021] This invention compares actual traffic flow data with predicted traffic demand parameters and adjusts the internal parameters of the flow evolution prediction model using the prediction model deviation, enabling the system to have self-learning capabilities and adapt to changes in traffic demand at different intersections, at different times, and under different non-motorized vehicle traffic characteristics. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of the flow chart of the non-motorized vehicle swarm intelligent flow guidance method based on microwave radar according to an embodiment of the present invention.

[0023] Figure 2 This is a schematic diagram of the generation of key monitoring areas and target point clouds of microwave radar according to an embodiment of the present invention.

[0024] Figure 3 This is a schematic diagram illustrating how individual target parameters of non-motorized vehicles are aggregated into collective intelligent flow parameters according to an embodiment of the present invention.

[0025] Figure 4 This is a schematic diagram of flow state evolution prediction and adaptive induction strategy generation according to an embodiment of the present invention.

[0026] Figure 5 This is a framework diagram of a non-motorized vehicle crowd intelligent flow guidance system based on microwave radar, according to an embodiment of the present invention. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0028] Reference Figure 1 One embodiment of the present invention proposes a non-motorized vehicle crowd intelligent flow guidance method based on microwave radar. Through microwave radar perception, crowd intelligent flow parameter construction and short-term traffic demand prediction, it realizes the forward identification and active guidance of the traffic demand of non-motorized vehicle groups, improves the traffic efficiency of non-motorized vehicles at intersections, and reduces the risks of congestion, rushing and traffic conflicts.

[0029] The method described in this embodiment specifically includes: S1. Acquire the signals collected by the microwave radar deployed at the intersection in the non-motorized vehicle monitoring area and generate microwave radar echo data; In one embodiment of the present invention, step S1 includes the following steps: Based on the physical layout of the intersection and the direction of the non-motorized vehicle lanes, a dynamic key monitoring area covering the waiting area and the upstream road section is defined. Based on historical traffic flow data and real-time detected target density, the radar's transmit power and beam scanning mode are dynamically adjusted to focus on the dynamic key monitoring area, generating an optimized beam scanning strategy. The command microwave radar performs detection according to the optimized beam scanning strategy to enhance the signal-to-noise ratio of targets in key areas and generate microwave radar echo data.

[0030] Specifically, the microwave radar is installed on poles above or to the side of the non-motorized vehicle lane at the intersection. The data source is the reflected echoes received by the radar antenna in the non-motorized vehicle waiting area and the upstream approach section. The waiting area is the area where non-motorized vehicles queue and accumulate in front of the stop line, and the upstream section is the approach area extending outwards along the direction of travel of the non-motorized vehicle lane. The radar control unit determines the dynamic key monitoring area based on the intersection channelization lines, the width of the non-motorized vehicle lane, the stop line position, and the upstream approach distance. The area of ​​the dynamic key monitoring area is written as: ; In the formula, Indicates the first The area of ​​the dynamic key monitoring area for each data collection cycle. Indicates the area of ​​the waiting area. Indicates the area of ​​the upstream road section. This indicates the data collection cycle number.

[0031] In the Within each acquisition cycle, the radar control unit reads the normal traffic flow level from historical traffic flow data and combines it with echo energy coarse detection to obtain a coarse target count, and calculates the real-time target density: ; In the formula, Indicates the first Real-time target density for each acquisition cycle Indicates the first The coarse count of targets is obtained by coarse detection of echo energy in each acquisition cycle. Indicates the first The area of ​​the dynamic key monitoring area for each collection cycle.

[0032] The radar transmit power and beam scanning period are adjusted based on real-time target density. Transmit power is written as: ; In the formula, Indicates the first Radar transmit power per acquisition cycle Indicates the radar's base transmit power. This represents the density adjustment coefficient. Indicates the first Real-time target density for each acquisition cycle.

[0033] Beam scanning period writing: ; In the formula, Indicates the first Beam scanning cycle of one acquisition cycle, Indicates the basic beam scanning period. This represents the density adjustment coefficient. Indicates the first Real-time target density for each acquisition cycle.

[0034] Optimized beam scanning strategy writing: ; In the formula, Indicates the first Optimized beam scanning strategy for each acquisition cycle Indicates the first The area of ​​the dynamic key monitoring area for each data collection cycle. Indicates the first Radar transmit power per acquisition cycle Indicates the first Beam scanning cycle of one acquisition cycle.

[0035] The radar continuously probes the dynamic key monitoring area according to an optimized beam scanning strategy and outputs microwave radar echo data: ; In the formula, Indicates the first Microwave radar echo data generated in each acquisition cycle Indicates the first Discrete sampling sequence of the first echo signal in each acquisition cycle, This represents the discrete sampling sequence of the second echo signal in the second acquisition cycle. Indicates the first The first in the collection cycle Discrete sampling sequence of the echo signal, Indicates the first Optimized beam scanning strategy for each acquisition cycle Indicates the first Timestamp of each collection period Indicates the first The number of echo channels acquired in each acquisition cycle.

[0036] For example, a 77GHz microwave radar is deployed on the non-motorized vehicle lane at the east entrance of a four-way intersection, with the waiting area taken as... The area of ​​the upstream road section is taken as ,get: ; Coarse detection of radar echo energy yields a coarse target count. Then the real-time target density is: ; Radar base transmit power The basic beam scanning period is taken The density adjustment coefficient is taken as ,but: ; ; This leads to an optimized beam scanning strategy: ; If four valid echo channels are acquired during the acquisition cycle, and the first three sampling points of each echo signal are taken as an encapsulation example, then: ; ; ; ; The number of echo channels is: ; Data collection timestamp The microwave radar echo data is then encapsulated as follows: The echo data includes echo amplitude sequence, dynamic key monitoring area, radar transmit power, beam scanning period, acquisition timestamp, and number of effective echo channels, which can be used as data input for subsequent time-frequency analysis, filtering, and point cloud generation.

[0037] S2. Process the microwave radar echo data to identify individual non-motorized vehicle targets and generate individual non-motorized vehicle target parameters for each target; In one embodiment of the present invention, step S2 includes the following steps: The microwave radar echo data is subjected to time-frequency analysis and filtering to extract effective target signals and generate radar point clouds; A spatial clustering algorithm is applied to the radar point cloud to distinguish different physical entities and generate multiple target clusters; State estimation is performed on each of the target clusters to calculate its kinematic information, generating individual non-motorized vehicle target parameters that include position, velocity, and heading information; The motion history of each non-motorized vehicle individual target is established by performing time-series trajectory tracking on the individual target parameters for multiple consecutive cycles, thereby generating individual historical trajectory data. The individual's historical trajectory data is analyzed to identify any deviations from the normal driving pattern, such as sudden changes in speed or direction, and an abnormal motion trajectory identifier is generated. When the abnormal motion trajectory identifier is detected, a high-priority emergency safety policy is generated, and the high-priority emergency safety policy is used as the priority input in the subsequent policy generation and execution process.

[0038] Specifically, the object of processing is the microwave radar echo data output in step S1. First, the echo signal is subjected to range-direction Fast Fourier Transform, Doppler frequency shift analysis, and angle estimation to filter out stationary background, roadside guardrail reflections, and low-energy noise points, resulting in a radar point cloud. Each point cloud point consists of range, azimuth, radial velocity, and echo intensity, and is converted into longitudinal and lateral coordinates in the road coordinate system. ; In the formula, Indicates the first In the first collection cycle The coordinates of a point cloud point in the longitudinal direction of the road. Indicates the first In the first collection cycle The radial distance of each point cloud point relative to the radar. Indicates the first In the first collection cycle The azimuth angle of each point cloud point relative to the radar normal direction. Indicates the point cloud point number. This represents the cosine function.

[0039] ; In the formula, Indicates the first In the first collection cycle The coordinates of a point cloud point in the horizontal direction of the road. Indicates the first In the first collection cycle The radial distance of each point cloud point relative to the radar. Indicates the first In the first collection cycle The azimuth angle of each point cloud point relative to the radar normal direction. Indicates the point cloud point number. This represents the sine function.

[0040] Dot Cloud Writing: ; In the formula, Indicates the first In the first collection cycle Point cloud points, Represents the longitudinal coordinates of the road. Indicates the lateral coordinates of the road. Indicates radial velocity, This indicates the echo intensity.

[0041] A spatial clustering algorithm is used to cluster radar points, with the clustering radius denoted as . The minimum number of points is denoted as After clustering, multiple target clusters are obtained. Each target cluster is denoted as . ,in Indicates the first The target cluster sequence number within each acquisition period. State estimation is performed on the target clusters to obtain the individual non-motorized vehicle target parameters: ; In the formula, Indicates the first In the first collection cycle Individual target parameters for each non-motorized vehicle target. Indicates the first The longitudinal position of each non-motorized vehicle target. Indicates the first The lateral position of each non-motorized vehicle target. Indicates the first The speed of movement of the non-motorized vehicle target Indicates the first The heading angle of the non-motorized vehicle target This indicates the target serial number of the non-motorized vehicle.

[0042] The position and velocity of the target cluster are determined by the mean of the point cloud points within the cluster: ; In the formula, Indicates the first The longitudinal position of each non-motorized vehicle target. Indicates belonging to the target cluster Point cloud points, Indicates the first In the first collection cycle One target cluster, Represents point cloud points The vertical coordinate, Indicates the target cluster The number of points in the point cloud.

[0043] ; In the formula, Indicates the first The lateral position of each non-motorized vehicle target. Indicates belonging to the target cluster Point cloud points, Indicates the first In the first collection cycle One target cluster, Represents point cloud points The horizontal coordinate, Indicates the target cluster The number of points in the point cloud.

[0044] ; In the formula, Indicates the first The speed of movement of the non-motorized vehicle target Indicates belonging to the target cluster Point cloud points, Indicates the first In the first collection cycle One target cluster, Represents point cloud points radial velocity, Indicates the target cluster The number of points in the point cloud.

[0045] After generating individual target parameters for non-motorized vehicles, trajectory association is performed on individual target parameters across multiple consecutive periods. The association rule is that the distance between target positions in adjacent periods is less than the trajectory association threshold. Calculate the rate of change of velocity and the change of heading for the same target after association: ; In the formula, Indicates the first The non-motorized vehicle target in the first Rate of change of speed per acquisition cycle Indicates the first The non-motorized vehicle target in the first The speed of movement per acquisition cycle Indicates the first The non-motorized vehicle target in the first The speed of movement per acquisition cycle This indicates the time interval between adjacent acquisition cycles.

[0046] ; In the formula, Indicates the first The non-motorized vehicle target in the first The change in heading over each data collection cycle Indicates the first The non-motorized vehicle target in the first The heading angle of motion in each acquisition cycle, Indicates the first The non-motorized vehicle target in the first The heading angle of motion in each acquisition cycle, This represents absolute value operations.

[0047] Abnormal motion trajectory identifiers are generated according to the following rules: ; ; In the formula, Indicates the first The non-motorized vehicle target in the first Abnormal motion trajectory identifiers for each acquisition cycle Represents the rate of change of velocity. Indicates the safety threshold for the rate of change of velocity. Indicates the change in heading. Indicates the safety threshold for changes in heading.

[0048] When any objective satisfies At that time, generate high-priority emergency security policies: ; In the formula, Indicates the first High-priority emergency security strategies for each data collection cycle Indicates the first Abnormal movement trajectory markers for non-motorized vehicles. Indicates the safety protection phase. Indicates the duration of safety protection.

[0049] For example, in a low-light scenario on a rainy night, the microwave radar in the first... Each acquisition cycle outputs echo data. After time-frequency analysis and filtering, the target cluster was obtained. The target cluster contains 3 point cloud points, and the point cloud point numbers are respectively taken as follows: , and The distance and azimuth of the three points are respectively , , The corresponding radial velocities are respectively , , The corresponding normalized echo intensities are respectively , , Perform coordinate transformation on the 11th point cloud point to obtain: ; ; Perform coordinate transformation on the 12th point cloud point to obtain: ; ; Perform coordinate transformation on the 13th point cloud point to obtain: ; ; Substituting the coordinates, velocity, and normalized echo intensity into the point cloud point expression, we obtain: ; ; ; Therefore, the target cluster is represented as: ; target cluster The position and velocity are estimated as follows: ; ; ; The first individual target parameter for non-motorized vehicles was obtained: ; If the speed of the same target in the previous cycle was The heading angle is The data collection period interval is ,but: ; ; Safety threshold for rate of change of velocity Heading change safety threshold ,therefore: ; The target did not trigger an abnormal motion trajectory indicator, and the individual target parameters proceed to the next step of the group aggregation process. If another target in an adjacent period... Sudden increase to The data collection period interval is ,but: ; because ,get: ; At this point, a high-priority emergency safety strategy is generated. This strategy is then incorporated into the subsequent strategy generation and execution process, and is given priority for handling potential traffic safety incidents.

[0050] S3. Aggregate all the individual target parameters of the non-motorized vehicles in the current period to quantify the overall motion characteristics of the group and generate the intelligent flow parameters of the non-motorized vehicle group; In one embodiment of the present invention, step S3 includes the following steps: The location coordinates of all targets are extracted from the individual non-motorized vehicle target parameters to construct the group spatial distribution and generate a group location dataset; The average distance and distribution density between targets are calculated based on the group location dataset to assess the degree of group aggregation and generate a group density index. Extract the speed vectors of all targets from the individual non-motorized vehicle target parameters, calculate the group average speed and speed variance to evaluate the group's speed coordination, and generate a group speed dispersion index. By integrating the population density index and the population speed dispersion index, a comprehensive description of the current state of the population is formed, and the non-motorized vehicle swarm intelligent flow parameters are generated.

[0051] Specifically, the data source for this step is the target parameters of all non-motorized vehicles in the current period output by step S2. Let the first... The number of non-motorized vehicle targets identified within each collection cycle is: The target set is First, calculate the group centroid based on the position coordinates of all targets: ; In the formula, Indicates the first The longitudinal centroid position of the population in each collection cycle. Indicates the first The longitudinal position of each non-motorized vehicle target. Indicates the first The number of non-motorized vehicle targets identified within each collection cycle. This indicates the target serial number of the non-motorized vehicle.

[0052] ; In the formula, Indicates the first The position of the lateral centroid of the population in each collection cycle Indicates the first The lateral position of each non-motorized vehicle target. Indicates the first The number of non-motorized vehicle targets identified within each collection cycle. This indicates the target serial number of the non-motorized vehicle.

[0053] The average distance between targets is used to describe the degree of group clustering: In the formula, Indicates the first Average distance between targets over one acquisition cycle This indicates the first target number in the target distance calculation. This indicates the second target number in the target distance calculation. Indicates the first The longitudinal position of each non-motorized vehicle target. Indicates the first The longitudinal position of each non-motorized vehicle target. Indicates the first The lateral position of each non-motorized vehicle target. Indicates the first The lateral position of each non-motorized vehicle target. Indicates the first The number of non-motorized vehicle targets identified within a collection cycle.

[0054] Distribution density is calculated based on the area of ​​the dynamic key monitoring zone: ; In the formula, Indicates the first Population distribution density for each collection period Indicates the first The number of non-motorized vehicle targets identified within each collection cycle. Indicates the first The area of ​​the dynamic key monitoring area for each collection cycle.

[0055] Writing the Normalized Population Density Index: ; In the formula, Indicates the first Normalized population density index for each collection period, Indicates the first Population distribution density for each collection period This indicates the preset maximum population density reference value.

[0056] The velocity vector is obtained by decomposing the velocity and heading angle: ; In the formula, Indicates the first The non-motorized vehicle target in the first The longitudinal velocity component of each acquisition cycle, Indicates the first The speed of movement of the non-motorized vehicle target Indicates the first The heading angle of the non-motorized vehicle target This represents the cosine function.

[0057] ; In the formula, Indicates the first The non-motorized vehicle target in the first The lateral velocity component of each acquisition cycle Indicates the first The speed of movement of the non-motorized vehicle target Indicates the first The heading angle of the non-motorized vehicle target This represents the sine function.

[0058] Group average velocity vector writing: ; In the formula, Indicates the first The average longitudinal velocity of the population over a collection period. Indicates the first The longitudinal velocity component of a non-motorized vehicle target. Indicates the first The number of non-motorized vehicle targets identified within each collection cycle. This indicates the target serial number of the non-motorized vehicle.

[0059] ; In the formula, Indicates the first The average lateral velocity of the population over a collection period Indicates the first The lateral velocity component of a non-motorized vehicle target. Indicates the first The number of non-motorized vehicle targets identified within each collection cycle. This indicates the target serial number of the non-motorized vehicle.

[0060] Velocity variance is used to characterize the velocity coordination of a group: ; In the formula, Indicates the first The speed variance of each acquisition cycle Indicates the first The longitudinal velocity component of a non-motorized vehicle target. Indicates the first The average longitudinal velocity of the population over a collection period. Indicates the first The lateral velocity component of a non-motorized vehicle target. Indicates the first The average lateral velocity of the population over a collection period Indicates the first The number of non-motorized vehicle targets identified within a collection cycle.

[0061] Writing the normalized population velocity dispersion index: ; In the formula, Indicates the first Normalized population velocity dispersion index for each acquisition cycle Indicates the first The speed variance of each acquisition cycle This indicates the preset maximum speed variance reference value.

[0062] Finally, the following non-motorized vehicle swarm intelligence flow parameters are generated: ; In the formula, Indicates the first Non-motorized vehicle crowd flow parameters for each collection cycle Indicates the target number of non-motorized vehicles. Indicates the position of the longitudinal centroid of the group. Indicates the position of the group's lateral centroid. Indicates the average longitudinal velocity of the group. Indicates the average lateral velocity of the group. This represents the average distance between targets. This represents the normalized population density index. This represents the dispersion index of normalized population velocity.

[0063] For example, following step S2, four non-motorized vehicle targets are identified within a certain collection period, and the individual target parameters are as follows: ; ; ; ; The location of the group's centroid is: ; ; The distances between the 6 groups of targets are as follows: , , , , , The average distance between targets is: ; Step S1, taking the area of ​​the dynamic key monitoring area. The population distribution density is: ; The maximum population density reference value is taken as The normalized population density index is: ; Based on velocity and heading angle decomposition, the longitudinal velocity components of the four targets are as follows: , , , The transverse velocity components are respectively , , , The average velocity of the group is: ; ; The velocity variance is calculated as follows: ; Maximum speed variance reference value is taken The normalized population velocity dispersion index is: ; The final non-motorized vehicle crowd flow parameters are obtained as follows: ; The results indicate that the non-motorized vehicle group consists of 4 vehicles, with the group's centroid close to the stop line, a high density index, and a low speed dispersion. This suggests that the group is approaching the stop line at a relatively consistent speed, and the next step of traffic demand prediction can proceed.

[0064] S4. Input the non-motorized vehicle group intelligent flow parameters into the flow evolution prediction model to estimate the future traffic trend of the group and generate predicted traffic demand parameters; In one embodiment of the present invention, step S4 includes the following steps: The non-motorized vehicle crowd flow parameters are input as the initial state into an evolutionary model based on time series analysis to predict its state at the next time step and generate flow characteristics for future time steps. Based on the flow characteristics at the future moment and the geometric position of the intersection stop line, the time for the group's centroid to reach the stop line is calculated, and a predicted arrival time window is generated. By combining the population size information in the future flow characteristics with the predicted arrival time window, the traffic pressure is quantified, and the predicted traffic demand parameters are generated. After implementing the guidance strategy, microwave radar data is continuously acquired and processed to obtain the actual situation of non-motorized vehicle groups crossing the stop line and generate actual traffic flow data. The actual traffic flow data is compared with the predicted traffic demand parameters generated before induction to quantify the accuracy of the prediction model and generate the prediction model bias. By utilizing the bias in the prediction model, the internal parameters of the flow state evolution prediction model are adjusted through a feedback mechanism to improve the accuracy of subsequent predictions and generate a self-learning prediction model.

[0065] Specifically, the data source for this step is the non-motorized vehicle swarm intelligence flow parameters output in step S3. Population size, population centroid, average population velocity, population density index, and velocity dispersion index are used as input states for the flow regime evolution prediction model. ; In the formula, Indicates the first Each acquisition cycle inputs the state vector of the flow regime evolution prediction model. Indicates the target number of non-motorized vehicles. Indicates the position of the longitudinal centroid of the group. Indicates the position of the group's lateral centroid. Indicates the average longitudinal velocity of the group. Indicates the average lateral velocity of the group. This represents the normalized population density index. This represents the dispersion index of normalized population velocity.

[0066] Short-term evolution prediction uses a time-step-based state extrapolation model. The predicted location of the population centroid at the next time step is written as follows: ; In the formula, This indicates the predicted longitudinal centroid position of the population at the next time step. Indicates the current vertical centroid position of the group. This represents the current average longitudinal velocity of the group. This indicates the prediction time step.

[0067] ; In the formula, This indicates the predicted lateral centroid position of the group at the next time step. Indicates the current position of the horizontal centroid of the group. This represents the average lateral velocity of the current group. This indicates the prediction time step.

[0068] The predicted group size will remain at the current target number within a very short time frame: ; In the formula, This indicates the predicted group size for the next time step. This indicates the current target number of non-motorized vehicles.

[0069] The location of the stop line at the intersection is represented by the longitudinal coordinates of the road, denoted as . The predicted arrival time window is calculated based on the remaining distance from the group's center of mass to the stop line and the longitudinal velocity: ; In the formula, Indicates the first The predicted arrival time window is generated for each collection cycle. Indicates the longitudinal coordinate of the stop line at the intersection. This indicates the predicted longitudinal centroid position of the population at the next time step. This represents the current average longitudinal velocity of the group. This represents the minimum speed reference value to prevent division by zero at low speeds. This represents the maximum value operation. Indicates the first Arrival time prediction bias for each acquisition cycle.

[0070] The parameters for predicting traffic demand integrate population size, population density, speed dispersion, and arrival time window: ; In the formula, Indicates the first Predicted traffic demand parameters for each data collection period This indicates the predicted group size for the next time step. Indicates the population density weight. This represents the normalized population density index. Indicates the weight of velocity dispersion. This represents the dispersion index of normalized population velocity. Indicates the predicted arrival time window. This represents the time stability constant.

[0071] After the guidance strategy is implemented, the actual non-motorized vehicle flow data crossing the stop line is continuously acquired, and the difference between the actual arrival time and the predicted arrival time is used as the prediction model bias: ; In the formula, Indicates the first Prediction model time deviation for each data collection cycle This represents the actual arrival time calculated from actual traffic flow data. This represents the predicted arrival time window generated before induction.

[0072] A feedback mechanism is used to correct the arrival time prediction bias: ; In the formula, Indicates the first Prediction bias correction for each acquisition cycle This represents the model's feedback learning rate. This indicates the time bias of the prediction model.

[0073] ; In the formula, This indicates the arrival time prediction bias for the next acquisition cycle. This indicates the arrival time prediction bias for the current acquisition period. This represents the prediction bias correction amount for the current data collection period. Feedback corrections are used to form a self-learning prediction model, improving the accuracy of subsequent predictions of traffic demand parameters.

[0074] For example, the non-motorized vehicle swarm flow parameters obtained in step S3 are as follows: ; Prediction time step Then the predicted group centroid position for the next time step is: ; ; The predicted population size is: ; The vertical coordinate of the stop line is taken The minimum speed reference value is taken as Arrival time prediction bias The predicted arrival time window is: ; Population density weighting The speed dispersion weight is taken as follows The time stability constant is taken as The predicted traffic demand parameters are: ; This value indicates that the group reached the stop line in a very short time, with high density and relatively uniform speed, indicating strong traffic pressure in the current direction. After the induction strategy is implemented, the radar continuously monitors the actual crossing of the stop line. If the actual arrival time is... The time bias of the prediction model is then: ; Model feedback learning rate The prediction bias correction is: ; The arrival time prediction bias for the next acquisition cycle is updated as follows: ; The updated model appropriately extends the arrival time estimate in subsequent predictions to reduce invalid phase adjustments caused by excessively early release.

[0075] S5. Based on the predicted traffic demand parameters and combined with the real-time traffic signal control information, generate an adaptive guidance strategy. In one embodiment of the present invention, step S5 includes the following steps: Request and receive current traffic light phase and duration information from the traffic signal controller, and generate the current traffic light phase status; The predicted traffic demand parameters are compared with the current traffic light phase status to determine whether the existing timing scheme meets the predicted demand and generate a phase adjustment decision. Identify non-motorized vehicle groups in other directions where there is traffic conflict, and generate corresponding predicted traffic demand parameters for groups in the conflicting directions to obtain the traffic demand parameters of the conflicting parties; Establish a right-of-way decision-making model, and simultaneously input the predicted right-of-way demand parameters in this direction and the right-of-way demand parameters of the conflicting parties into the model to determine priority and generate priority ranking results; Based on the priority ranking results, the phase adjustment decision is adjusted to ensure that the passage demand of high-priority directions is met first; Based on the adjusted phase adjustment decision, the specific green light duration extension or phase start advance is calculated to formulate an optimization scheme and generate the final adaptive guidance strategy.

[0076] Specifically, the data sources for this step are the predicted traffic demand parameters output from step S4, the high-priority emergency safety strategy output from step S2, and the real-time feedback from the traffic signal controller regarding the signal light phase, remaining green light duration, remaining red light duration, planned green light duration, and maximum green light duration. The current signal light phase status is written as follows: ; In the formula, Indicates the first The current phase status of the traffic light in each acquisition cycle. Indicates the current signal phase type. This indicates the remaining green light duration for non-motorized vehicles. Indicates the remaining time of the red light for non-motorized vehicles. Indicates the planned green light duration for the current cycle. This indicates the maximum green light duration allowed for the non-motorized vehicle phase.

[0077] When there are non-motorized vehicle groups from other directions that conflict with this direction, predicted traffic demand parameters are also generated for the conflicting directions. Let... Indicates the direction type. Indicates this direction. Indicates the direction of conflict. The right-of-way decision-making model calculates priority scores for each direction: ; In the formula, Indicates the first In each collection cycle, the direction Priority rating, Indicates the weight of traffic demand. Indicates direction Predicted traffic demand parameters Indicates the weight of waiting time. Indicates direction Average waiting time Indicates the direction type. Indicates this direction. Indicates the direction of conflict.

[0078] Writing the priority sorting results: ; ; In the formula, Indicates the first Priority ranking results for each acquisition cycle. This indicates the priority score for this direction. Indicates the priority score for conflict direction. This indicates that this direction has priority. This indicates that the direction of conflict takes precedence.

[0079] For situations where non-motorized vehicles currently have a green light and their direction has priority, the extension of the green light duration is calculated based on the difference between the remaining green light capacity and the predicted traffic demand: ; In the formula, Indicates the first The green light duration extension calculated per data collection cycle Indicates the maximum green light duration. Indicates the planned green light duration. This represents the maximum value operation. This represents the predicted traffic demand parameters for this direction. This indicates the green light capacity for non-motorized vehicles. This indicates the remaining green light duration for non-motorized vehicles. This represents the minimum value operation.

[0080] For situations where the non-motorized vehicle lane is currently at a red light and has priority in its own direction, the phase start advance is calculated based on the priority ratio between the current direction and the conflicting direction: ; In the formula, Indicates the first Phase start advance calculated per acquisition cycle Indicates the remaining time of the red light for non-motorized vehicles. Indicates the maximum allowed early start time. This indicates the priority score for this direction. Indicates the priority score for conflict direction. This represents the minimum value operation.

[0081] Writing Adaptive Induced Strategies: ; In the formula, Indicates the first An adaptive guidance strategy is generated for each collection cycle. This indicates the priority sorting result. This indicates the amount of time the green light has been extended. Indicates the phase start advance. This indicates a high-priority emergency safety strategy. If If the abnormal movement trajectory is marked as 1, then the high-priority emergency safety strategy will take precedence over the regular green light extension or phase start advance in the execution phase.

[0082] For example, following step S4, the predicted traffic demand parameter for this direction is: The radar simultaneously detected non-motorized vehicle groups in the north-south conflict direction, and obtained the predicted traffic demand parameters for the conflict direction through the same prediction process. The average waiting time in this direction is taken as follows: The average waiting time in the conflict direction is taken as Traffic demand weighting Waiting time weighting The priority score for this direction is: ; The priority score for conflict direction is: ; because The priority sorting result is: ; The current green light phase for non-motorized vehicles is green, and the remaining green light duration is... The planned green light duration is The maximum green light duration is Non-motorized vehicle green light capacity The green light duration is extended by: ; The current situation is a green light extension, and the phase start time is measured in advance. If step S2 does not generate an abnormal motion trajectory identifier, then The adaptive induction strategy is as follows: ; This strategy suggests that the current green light for non-motorized vehicles should be extended. This is to absorb the demand from the group that is about to reach the stop line. If step S2 generates... In this case, the conventional green light extension strategy is suppressed by the security strategy, and subsequent execution steps are given priority. Full red protection avoids the risk of collisions between abnormal targets and conflict phases.

[0083] S6. Execute the adaptive guidance strategy to adjust the timing scheme of non-motorized vehicle traffic lights.

[0084] In one embodiment of the present invention, step S6 includes the following steps: The adaptive guidance strategy is sent to the traffic signal controller; The traffic signal controller adjusts the green light duration or phase start time of the non-motorized vehicle traffic signal light according to the adaptive guidance strategy; After implementing the adjusted timing scheme, microwave radar data will continue to be collected to provide input for the generation of swarm intelligence flow parameters and the prediction of flow evolution in the next cycle.

[0085] Specifically, the data source for this step is the adaptive induction strategy output in step S5. The execution target is the timing parameters of traffic lights for non-motorized vehicles. If the adaptive guidance strategy... If the abnormal movement trajectory is marked as 1, the traffic signal controller will prioritize the execution of the safety protection phase and the safety protection duration; if If the abnormal movement trajectory is marked as 0, the standard guidance strategy will be executed according to the green light duration extension or phase start advance. In the case of a green light extension, the green light duration for the non-motorized vehicle phase will be updated, provided it does not exceed the maximum green light duration. ; In the formula, Indicates the first The duration of the new green light after each data collection cycle is completed. Indicates the maximum green light duration. Indicates the planned green light duration. This indicates the amount of time the green light has been extended. This represents the minimum value operation.

[0086] When the red light ends early and the green light for non-motorized vehicles begins, the phase start time should be updated as follows: ; In the formula, Indicates the first The start time of the new phase after each acquisition cycle. Indicates the originally planned phase start time. This indicates the phase start advance.

[0087] The execution instructions sent to the traffic signal controller are written as follows: ; In the formula, Indicates the first Each acquisition cycle sends an execution command to the traffic signal controller. This indicates the non-motorized vehicle phase number that needs adjustment. Indicates the duration of the new green light. Indicates the start time of the new phase. Indicates the safety protection phase. Indicates the duration of safety protection. This indicates the timestamp when the execution command was sent.

[0088] After the timing scheme is adjusted, the microwave radar continues to collect actual flow data across the stop line. The actual flow data is returned to step S4 for model feedback updates, forming a closed-loop induced control.

[0089] For example, the adaptive induction strategy obtained following step S5 is as follows: ; The current planned green light duration is The maximum green light duration is The new green light duration after implementation is: ; Because the phase start advance is If the original planned phase start time is ,but: ; Safety protection phase take Safety protection duration is taken Execution instructions: ; After receiving the instruction, the traffic signal controller will adjust the green light duration for the non-motorized vehicle phase at the east entrance from... Adjusted to The radar detected four non-motorized vehicles crossing the stop line sequentially at the end of the adjusted green light period, with no significant congestion in the waiting area. This indicates that the guidance strategy effectively absorbed the predicted traffic demand and reduced the risk of rushing at the end of the green light and congestion at the red light. If we proceed from step S5... The traffic signal controller will then take priority. Full red protection, suspend the regular green light extension action, and resume the adaptive guidance process after the abnormal target passes or stops.

[0090] Based on the same inventive concept, such as Figure 5 As shown, the present invention also provides a non-motorized vehicle swarm intelligent flow guidance system based on microwave radar, comprising: The acquisition module is used to acquire signals collected by microwave radar deployed at the intersection within the non-motorized vehicle monitoring area and generate microwave radar echo data. The processing module is used to process microwave radar echo data to identify individual non-motorized vehicle targets and generate individual non-motorized vehicle target parameters for each target; The aggregation module is used to aggregate the target parameters of all non-motorized vehicles in the current period to quantify the overall motion characteristics of the group and generate the intelligent flow parameters of the non-motorized vehicle swarm. The prediction module is used to input the non-motorized vehicle crowd flow parameters into the flow evolution prediction model to estimate the future traffic trend of the group and generate predicted traffic demand parameters. The strategy generation module is used to generate adaptive guidance strategies based on predicted traffic demand parameters and real-time traffic signal control information. The execution module is used to implement adaptive guidance strategies to adjust the timing scheme of non-motorized vehicle traffic lights.

[0091] All equivalent changes and modifications made in accordance with the teachings of this invention are still within the scope of this invention. Those skilled in the art will readily conceive of other embodiments of this invention upon considering the specification and the disclosure of practical truth. This application is intended to cover any variations, uses, or adaptations of this invention that follow the general principles of this invention and include common knowledge or conventional techniques in the art not described herein.

Claims

1. A non-motorized vehicle swarm intelligent flow guidance method based on microwave radar, characterized in that, Includes the following steps: The system acquires signals collected by microwave radar deployed at the intersection within the non-motorized vehicle monitoring area and generates microwave radar echo data. The microwave radar echo data is processed to identify individual non-motorized vehicle targets and generate individual non-motorized vehicle target parameters for each target; Aggregate all the individual target parameters of the non-motorized vehicles in the current period to quantify the overall motion characteristics of the group and generate the intelligent flow parameters of the non-motorized vehicle swarm. The non-motorized vehicle group flow parameters are input into the flow evolution prediction model to estimate the future travel trend of the group and generate predicted travel demand parameters. Based on the predicted traffic demand parameters and combined with the real-time traffic signal control information, an adaptive guidance strategy is generated. The adaptive guidance strategy is implemented to adjust the timing scheme of non-motorized vehicle traffic lights.

2. The non-motorized vehicle swarm intelligent flow guidance method based on microwave radar according to claim 1, characterized in that, The process of processing the microwave radar echo data to identify individual non-motorized vehicle targets and generating individual non-motorized vehicle target parameters for each target includes: The microwave radar echo data is subjected to time-frequency analysis and filtering to extract effective target signals and generate radar point clouds; A spatial clustering algorithm is applied to the radar point cloud to distinguish different physical entities and generate multiple target clusters; State estimation is performed on each of the target clusters to calculate its kinematic information, generating individual non-motorized vehicle target parameters that include position, velocity, and heading information.

3. The non-motorized vehicle swarm intelligent flow guidance method based on microwave radar according to claim 1, characterized in that, The aggregation of all individual non-motorized vehicle target parameters within the current period to quantify the overall motion characteristics of the group and generate non-motorized vehicle swarm intelligence flow parameters includes: The location coordinates of all targets are extracted from the individual non-motorized vehicle target parameters to construct the group spatial distribution and generate a group location dataset; The average distance and distribution density between targets are calculated based on the group location dataset to assess the degree of group aggregation and generate a group density index. Extract the speed vectors of all targets from the individual non-motorized vehicle target parameters, calculate the group average speed and speed variance to evaluate the group's speed coordination, and generate a group speed dispersion index. By integrating the population density index and the population speed dispersion index, a comprehensive description of the current state of the population is formed, and the non-motorized vehicle swarm intelligent flow parameters are generated.

4. The non-motorized vehicle swarm intelligent flow guidance method based on microwave radar according to claim 1, characterized in that, The step of inputting the non-motorized vehicle group flow parameters into the flow evolution prediction model to estimate the future travel trend of the group and generate predicted traffic demand parameters includes: The non-motorized vehicle crowd flow parameters are input as the initial state into an evolutionary model based on time series analysis to predict its state at the next time step and generate flow characteristics for future time steps. Based on the flow characteristics at the future moment and the geometric position of the intersection stop line, the time for the group's centroid to reach the stop line is calculated, and a predicted arrival time window is generated. By combining the population size information in the future flow characteristics with the predicted arrival time window, the traffic pressure is quantified, and the predicted traffic demand parameters are generated.

5. The non-motorized vehicle swarm intelligent flow guidance method based on microwave radar according to claim 1, characterized in that, The step of generating an adaptive guidance strategy based on the predicted traffic demand parameters and in conjunction with real-time traffic light control information includes: Request and receive current traffic light phase and duration information from the traffic signal controller, and generate the current traffic light phase status; The predicted traffic demand parameters are compared with the current traffic light phase status to determine whether the existing timing scheme meets the predicted demand and generate a phase adjustment decision. Based on the phase adjustment decision, the specific green light duration extension or phase start advance is calculated to formulate an optimization scheme and generate the adaptive guidance strategy.

6. The non-motorized vehicle swarm intelligent flow guidance method based on microwave radar according to claim 5, characterized in that, Before generating the adaptive induction strategy, the following is also included: Identify non-motorized vehicle groups in other directions where there is traffic conflict, and generate corresponding predicted traffic demand parameters for groups in the conflicting directions to obtain the traffic demand parameters of the conflicting parties; Establish a right-of-way decision-making model, and simultaneously input the predicted right-of-way demand parameters in this direction and the right-of-way demand parameters of the conflicting parties into the model to determine priority and generate priority ranking results; Based on the priority ranking results, the phase adjustment decision is adjusted to ensure that the passage demand of high-priority directions is met first, and the final adaptive guidance strategy is generated based on the adjusted decision.

7. The non-motorized vehicle swarm intelligent flow guidance method based on microwave radar according to claim 4, characterized in that, After generating the predicted traffic demand parameters, the process also includes: After implementing the guidance strategy, microwave radar data is continuously acquired and processed to obtain the actual situation of non-motorized vehicle groups crossing the stop line and generate actual traffic flow data. The actual traffic flow data is compared with the predicted traffic demand parameters generated before induction to quantify the accuracy of the prediction model and generate the prediction model bias. By utilizing the bias in the prediction model, the internal parameters of the flow state evolution prediction model are adjusted through a feedback mechanism to improve the accuracy of subsequent predictions and generate a self-learning prediction model.

8. The non-motorized vehicle swarm intelligent flow guidance method based on microwave radar according to claim 1, characterized in that, The process of acquiring signals collected by microwave radar deployed at the intersection within the non-motorized vehicle monitoring area and generating microwave radar echo data includes: Based on the physical layout of the intersection and the direction of the non-motorized vehicle lanes, a dynamic key monitoring area covering the waiting area and the upstream road section is defined. Based on historical traffic flow data and real-time detected target density, the radar's transmit power and beam scanning mode are dynamically adjusted to focus on the dynamic key monitoring area, generating an optimized beam scanning strategy. The command microwave radar performs detection according to the optimized beam scanning strategy to enhance the signal-to-noise ratio of targets in key areas and generate microwave radar echo data.

9. A non-motorized vehicle swarm intelligent flow guidance method based on microwave radar according to claim 2, characterized in that, After generating the individual target parameters for the non-motorized vehicles, the method further includes: The motion history of each non-motorized vehicle individual target is established by performing time-series trajectory tracking on the individual target parameters for multiple consecutive cycles, thereby generating individual historical trajectory data. The individual's historical trajectory data is analyzed to identify any deviations from the normal driving pattern, such as sudden changes in speed or direction, and an abnormal motion trajectory identifier is generated. When the abnormal motion trajectory identifier is detected, a high-priority emergency safety strategy is generated and used to replace the originally planned adaptive guidance strategy in order to deal with potential traffic safety incidents.

10. A non-motorized vehicle swarm intelligent flow guidance system based on microwave radar, characterized in that, include: The acquisition module is used to acquire signals collected by microwave radar deployed at the intersection within the non-motorized vehicle monitoring area and generate microwave radar echo data. The processing module is used to process the microwave radar echo data to identify individual non-motorized vehicle targets and generate individual non-motorized vehicle target parameters for each target; The aggregation module is used to aggregate all the individual target parameters of the non-motorized vehicles in the current period to quantify the overall motion characteristics of the group and generate the intelligent flow parameters of the non-motorized vehicle group. The prediction module is used to input the non-motorized vehicle group intelligent flow parameters into the flow evolution prediction model to predict the future travel trend of the group and generate predicted travel demand parameters. The strategy generation module is used to generate an adaptive guidance strategy based on the predicted traffic demand parameters and the real-time traffic signal control information. An execution module is used to execute the adaptive guidance strategy to adjust the timing scheme of non-motorized vehicle traffic lights.

Images