Urban-rural joint multi-modal perception traffic flow cooperative optimization control system

Abstract

The application discloses a kind of urban-rural joint multi-modal perception traffic flow cooperative optimization control systems, it is related to traffic control technical field, including: data acquisition processing module, for obtaining point cloud data and image data collected by target intersection roadside sensing unit and carrying out space-time registration, the microcosmic trajectory aggregation data of non-motor vehicle group and the macroscopic space-time distribution data of adjacent motor vehicle lane are extracted from registration result.Non-motor vehicle's tendency to cross the border is converted into potential energy by equivalent borrowing potential energy physical field analogy model, and the passing condition of motor vehicle is converted into gap and repulsion degree, when accurately capturing that there is available space-time gap in motor vehicle lane, the system allows non-motor vehicle to borrow road legally under the premise of ensuring safety, realizes the dynamic sharing of space resources, effectively alleviates the non-motor vehicle congestion problem of urban-rural joint tidal peak.

Description

Technical Field

[0001] This invention relates to the field of traffic control technology, specifically to a multimodal sensing traffic flow collaborative optimization control system for urban-rural fringe areas. Background Technology

[0002] With the continuous advancement of urbanization, the urban-rural fringe areas, as hubs connecting urban areas and towns, are experiencing increasingly complex traffic flows. Due to the unique travel patterns, intersections in these areas generally exhibit a mixed traffic situation with a high degree of intermingling between motor vehicles and a large number of non-motorized vehicles, especially during peak tidal periods of daily commuting, when the flow of non-motorized vehicles often experiences explosive growth.

[0003] To regulate this complex traffic order, current conventional control measures typically involve setting up fixed physical barriers between motor vehicle and non-motor vehicle lanes, or using traffic signal phase duration allocation for basic traffic control, in order to rigidly separate different types of traffic participants in physical space or travel time.

[0004] However, this conventional approach aimed at ensuring safety is based on the assumption of relatively stable and homogeneous traffic flow. Once faced with the extreme asymmetrical road conditions typical of urban-rural fringe areas, the drawbacks of static boundaries become fully exposed. When the waiting space of non-motorized vehicle lanes is rapidly filled or even saturated by explosive traffic flow, the originally fixed physical barriers become obstacles to passage, causing large numbers of non-motorized vehicles to be trapped in confined spaces. Due to the lack of flexible spatial scheduling mechanisms, the crowded non-motorized vehicle groups inevitably spread outwards, forced to illegally encroach on adjacent motorized vehicle lanes. At this point, even if adjacent motorized vehicle lanes are extremely empty due to tidal phenomena, they cannot be legally and safely used. This severe imbalance in right-of-way allocation ultimately leads to severe disorderly mixing of motorized and non-motorized traffic and congestion. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a multimodal sensing traffic flow collaborative optimization control system for urban-rural fringe areas.

[0006] To achieve the above objectives, the technical solution of the present invention is as follows: A multimodal sensing traffic flow collaborative optimization control system for urban-rural fringe areas includes: The data acquisition and processing module is used to acquire point cloud data and image data collected by the roadside sensing unit at the target intersection and perform spatiotemporal registration. From the registration results, it extracts the micro-trajectory aggregation data of the non-motorized vehicle group and the macro-spatiotemporal distribution data of the adjacent motor vehicle lanes. The feature extraction interaction module is used to perform spatial vector analysis on micro-trajectory aggregation data, extract non-motorized vehicle overflow occupancy rate data and group state dispersion data, and extract available spatiotemporal gap data of motor vehicle lanes based on macro-spatiotemporal distribution data. The data coupling calculation module is used to perform cross-coupling calculations between the group state discreteness data and the available spatiotemporal gap data to generate boundary expansion impedance coefficient data. The boundary expansion impedance coefficient data is then used to perform damping attenuation processing on the non-motorized vehicle overflow occupancy rate data to obtain equivalent borrowing potential energy data. The data judgment and output module is used to determine whether the equivalent borrowed potential energy data is greater than the preset mapping threshold. If so, the optimal detour boundary coordinate data is generated based on the numerical difference between the equivalent detour potential energy data and the mapping threshold, and a light and shadow reconstruction collaborative control instruction set containing the optimal detour boundary coordinate data is output to the roadside light and shadow projection device. If not, then static right-of-way maintenance instruction data is generated.

[0007] A multimodal sensing traffic flow collaborative optimization control method for urban-rural fringe areas includes the following steps: The point cloud data and image data collected by the roadside sensing unit at the target intersection are acquired and spatiotemporally registered. From the registration results, the micro-trajectory aggregation data of the non-motorized vehicle group and the macro-spatiotemporal distribution data of the adjacent motor vehicle lanes are extracted. Spatial vector analysis was performed on the micro-trajectory aggregation data to extract non-motorized vehicle overflow occupancy rate data and group state dispersion data. Based on macro-spatiotemporal distribution data, available spatiotemporal gap data of motor vehicle lanes were extracted. The group state discreteness data and the available spatiotemporal gap data are cross-coupled to generate boundary expansion impedance coefficient data. The boundary expansion impedance coefficient data is then used to dampen and attenuate the non-motorized vehicle overflow occupancy rate data to obtain the equivalent borrowing potential energy data. Determine whether the equivalent borrowing potential energy data is greater than the preset mapping threshold; If so, the optimal detour boundary coordinate data is generated based on the numerical difference between the equivalent detour potential energy data and the mapping threshold, and a light and shadow reconstruction collaborative control instruction set containing the optimal detour boundary coordinate data is output to the roadside light and shadow projection device. If not, then static right-of-way maintenance instruction data is generated.

[0008] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention uses an equivalent potential energy physical field analogy model to transform the tendency of non-motorized vehicles to cross boundaries into potential energy, and the passage conditions of motorized vehicles into gaps and repulsion. When a usable spatiotemporal gap exists in the motorized vehicle lane, the system allows non-motorized vehicles to legally cross the lane under the premise of ensuring safety, realizing dynamic sharing of spatial resources and effectively alleviating the problem of non-motorized vehicle congestion during peak hours in urban-rural fringe areas. Furthermore, it goes beyond simple traffic statistics; it extracts the microscopic trajectory of non-motorized vehicles (overflow occupancy rate, group state dispersion) and the macroscopic state of motorized vehicles (3D bounding box, approximation momentum, etc.) through a multimodal sensing unit, and inputs both into a risk repulsion model for cross-coupling. By introducing environmental damping, dispersion gain weights, and a smoothing constant to prevent division by zero, the system can quantify the instantaneous risk of mixed flow conflicts with extreme precision, effectively avoiding blind lane crossing by non-motorized vehicles and ensuring the safety of traffic participants from the underlying algorithm level. Attached Figure Description

[0009] The disclosure of this 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 this invention. In the drawings, the same reference numerals are used to refer to the same parts. Wherein: Figure 1 This is a system module diagram of the present invention; Figure 2 This is a diagram illustrating the method steps of the present invention; Figure 3 This is a flowchart of the present invention. Detailed Implementation

[0010] It is readily understood that, based on the technical solution of this invention, those skilled in the art can propose various interchangeable structural methods and implementations without altering the essential spirit of the invention. Therefore, the following detailed embodiments and accompanying drawings are merely illustrative examples of the technical solution of this invention and should not be considered as the entirety of the invention or as limitations or restrictions on the technical solution of this invention.

[0011] Application Overview: In the field of traffic signal and right-of-way allocation control, especially in mixed traffic scenarios in urban-rural fringe areas, the smoothness of traffic flow and space utilization are regarded as key indicators for measuring the collaborative operation capability of road networks. This highly efficient traffic flow is essentially a process of non-destructive replacement of spatiotemporal resources at the traffic dynamics level. That is, through the full-domain monitoring of multiple sensors, the road plane is used as an elastic carrier to accurately match the traffic needs of different types of vehicles, thereby forming a safe movement trajectory with high physical continuity in the intersection area.

[0012] However, existing technologies lack a verification mechanism to verify the consistency of the state coupling between the microscopic anomalies of two-wheeled vehicle groups and the macroscopic evolution of vehicle traffic flow. This leads to an inability to accurately identify the disorderly cross-boundary compression and static space wastage problems that exist during tidal congestion periods. Disorderly cross-boundary compression manifests as two-wheeled vehicle groups, although constrained by guardrails, actually generating a strong lateral expansion impulse due to excessive internal density. Static space wastage, on the other hand, manifests as the potential carrying capacity of adjacent vehicle lanes being completely consumed by rigid rules even when large sections of adjacent lanes are vacant, when the system forcibly maintains the inherent lane divisions. As a result, a strict physical correspondence cannot be established between the lateral expansion of microscopic groups and the longitudinal idleness of macroscopic lanes, causing the system to misjudge potential conflicts or provide ambiguous feedback on traffic capacity, thereby affecting the accurate assessment of mixed flow dynamics characteristics and the targeted nature of control strategies.

[0013] For example, in the morning and evening commuting traffic in urban-rural fringe areas, when a large number of two-wheeled vehicles are queuing at intersections, conventional monitoring systems can only capture the macroscopic queue length on the non-motorized vehicle lane through ordinary visual analysis, but cannot distinguish whether there are any covert probing actions at the front of the queue that are ready to illegally cross the line at any time. Furthermore, when a large area of ​​blank space appears in an adjacent car lane due to the upstream red light blocking it, the system only records the appearance of each lane operating in isolation, failing to monitor the dynamic drift potential of the safety boundaries between different lanes and the efficiency of spatial replacement. Specifically, the system misjudges high-risk boundary crossing attempts as normal static waiting and does not intervene, or incorrectly classifies absolutely safe blank road sections as unusable physical no-go zones. As a result, multi-source traffic flows remain fixed in a rigid isolation mode, unable to form a spatially flexible shared trajectory that conforms to the principles of traffic dynamics.

[0014] If the above problems are not addressed, the control system will continue to lose its ability to objectively judge the transient supply and demand relationship of mixed traffic flow. Among them, the failure to identify disorderly cross-boundary squeezing will cause two-wheeled vehicles to excessively accumulate lateral crossing impulses, resulting in the group's movement trajectory deviating from the constraints of physical guardrails, thereby increasing the risk of violent collisions in mixed traffic. At the same time, the failure to correct the problem of static space waste will severely suppress the road network's throughput potential, preventing different types of traffic flow from exhibiting spatiotemporal complementary characteristics, and ultimately causing intersections to lose their due dynamic guidance capacity. As a result, the inaccuracy of right-of-way boundary feedback will systematically hinder the operation of the core mechanism of flexible spatiotemporal sharing of traffic flow, seriously affecting the achievement of the goal of refined governance of complex road networks.

[0015] like Figure 1-3 As shown, a multimodal sensing traffic flow collaborative optimization control system for urban-rural fringe areas includes: The data acquisition and processing module acquires point cloud data and image data collected by the roadside sensing units at the target intersection and performs spatiotemporal registration. From the registration results, it extracts microscopic trajectory aggregation data of non-motorized vehicle groups and macroscopic spatiotemporal distribution data of adjacent motorized vehicle lanes. The system collects point cloud data and image data in real time through sensing units (such as LiDAR and high-definition cameras) deployed on the roadside of the target intersection. To eliminate data barriers between heterogeneous sensors, the system presets a time deviation tolerance to complete timestamp alignment. Subsequently, based on the intrinsic and extrinsic parameter matrices of the sensors, the spatial coordinates of the three-dimensional point cloud are accurately projected and mapped to the pixel coordinate system of the two-dimensional image, thereby giving the image semantic contours absolute depth and physical location information. Based on this four-dimensional spatiotemporal structured data, the system performs structured separation of the hybrid stream through target attribute identification: on the one hand, the motion states of non-motorized vehicle targets are connected in time sequence to construct microscopic trajectory aggregation data reflecting their individual microscopic motion characteristics; on the other hand, the parameters and speeds of car surround boxes within the physical boundaries of adjacent motorized vehicle lanes are extracted to construct macroscopic spatiotemporal distribution data reflecting the macroscopic occupancy status of the lanes.

[0016] The feature extraction interaction module performs spatial vector analysis on the aggregated micro-trajectory data to extract non-motorized vehicle overflow occupancy rate data and group state dispersion data. Based on macro-spatiotemporal distribution data, it extracts usable spatiotemporal gap data for motorized vehicle lanes. After acquiring the above data, the system quantifies the tendency of non-motorized vehicles to cross boundaries and the carrying capacity of motorized vehicles. For non-motorized vehicle groups, the system performs vector comparison between their micro-trajectory coordinates and the existing physical boundaries of the non-motorized vehicle lanes, calculates the area occupancy rate of the boundary-crossing coordinate set, and extracts non-motorized vehicle overflow occupancy rate data. Simultaneously, by statistically analyzing the variance of the heading angle and the dispersion coefficient of instantaneous velocity within the group and performing weighted fusion, it quantifies and extracts group state dispersion data, which characterizes the degree of disorder in the group's movement. For adjacent motorized vehicle lanes, the system calculates the physical head-to-tail distance between adjacent vehicles and the approach speed of the following vehicle, and extracts vehicle-free passage segments within the future prediction window of the lane through kinematic deduction, i.e., absolutely safe usable spatiotemporal gap data.

[0017] The data coupling calculation module is used to cross-couple group state discreteness data with available spatiotemporal gap data to generate boundary expansion impedance coefficient data. This boundary expansion impedance coefficient data is then used to dampen and attenuate non-motorized vehicle overflow occupancy rate data, yielding equivalent lane-borrowing potential energy data. This module aims to mathematically reduce and integrate the subjective lane-borrowing needs of non-motorized vehicles with the objective safety environment. The system first performs a reciprocal operation on the available spatiotemporal gap data (introducing a minimal constant to prevent division by zero) to obtain the basic repulsion degree of the motorized vehicle lane. Subsequently, the basic repulsion degree, group state discreteness, and environmental state parameters are input into the risk repulsion model for coupling, and the evaluation results are normalized to generate the boundary expansion impedance coefficient characterizing the current comprehensive lane-borrowing risk. The system uses this impedance coefficient to construct an attenuation multiplier to dampen and attenuate the initial non-motorized vehicle overflow occupancy rate data. The attenuated eigenvalues ​​are then transformed into equivalent lane-borrowing potential energy data for the final right-of-way decision. This process ensures that lane-borrowing potential energy is forcibly suppressed when space is insufficient or the group is chaotic, guaranteeing the safety of the decision from the underlying logic.

[0018] The data judgment and output module is used to determine whether the equivalent lane-crossing potential energy data is greater than a preset mapping threshold. If so, it generates optimal lane-crossing boundary coordinate data based on the difference between the equivalent lane-crossing potential energy data and the mapping threshold, and outputs a light and shadow reconstruction collaborative control instruction set containing the optimal lane-crossing boundary coordinate data to the roadside light and shadow projection equipment. If not, it generates static right-of-way maintenance instruction data. The system does not use a static and fixed judgment standard, but dynamically generates a mapping threshold that matches the current spatiotemporal and meteorological situation by combining the current timestamp, historical periodic flow density, and real-time environmental risk weighting parameters such as illumination, precipitation, and road surface slippage. The system compares the real-time calculated equivalent lane-crossing potential energy data with this dynamic mapping threshold to determine whether the current conditions are safe for initiating right-of-way replacement. If the equivalent lane-crossing potential energy data is greater than the mapping threshold, it indicates that the non-motorized vehicle group's crossing potential energy has safely exceeded the current threshold. The system calculates the difference between the two values ​​and converts it into a lateral lane-crossing width according to a preset mapping relationship. The system overlays the wide-width data along the normal vector onto the original non-motorized vehicle lane boundary, generating new optimal lane-sharing boundary coordinates. These coordinates are then encapsulated into a light and shadow reconstruction collaborative control instruction set and sent to the roadside light and shadow projection equipment. The equipment then dynamically projects a bright virtual lane-sharing line onto the road surface to guide non-motorized vehicles to safely use the lane. Conversely, if the equivalent lane-sharing potential energy data fails to exceed the mapping threshold (i.e., a negative judgment), it indicates a high current risk or insufficient lane-sharing demand. The system directly extracts the existing physical boundary coordinates of the non-motorized vehicle lane to generate static right-of-way maintenance instruction data. Upon receiving this instruction, the lower-level equipment maintains the original physical boundary indication state, cuts off lane-sharing guidance, and ensures the normal traffic order at the intersection in the most secure way.

[0019] The core innovation of the aforementioned technology lies in breaking through the rigid limitations of traditional physical isolation. Through an analogy model of equivalent lane-borrowing potential energy, the tendency of non-motorized vehicles to cross boundaries is transformed into potential energy, and the passage conditions of motorized vehicles are transformed into gaps and repulsion. When a usable spatiotemporal gap is accurately detected in the motorized vehicle lane, the system allows non-motorized vehicles to legally borrow lanes under safe conditions, achieving dynamic sharing of spatial resources and effectively alleviating the congestion problem of non-motorized vehicles during peak hours in urban-rural fringe areas. Furthermore, it goes beyond simple traffic statistics; it extracts the microscopic trajectories of non-motorized vehicles (overflow occupancy rate, group state dispersion) and the macroscopic states of motorized vehicles (3D bounding box, approximation momentum, etc.) through a multimodal sensing unit, and inputs both into a risk repulsion model for cross-coupling. By introducing environmental damping, dispersion gain weights, and a smoothing constant to prevent division by zero, the system can quantify the instantaneous risk of mixed flow conflicts with extreme precision, effectively avoiding blind lane-borrowing by non-motorized vehicles and ensuring the safety of traffic participants from the underlying algorithmic level.

[0020] Because a single visual image sensor lacks depth information, it cannot accurately calculate the physical distance and space occupancy between vehicles; while a single point cloud sensor lacks rich texture information, it is difficult to accurately distinguish closely adjacent non-motorized vehicles from the environmental background in dense mixed traffic flow. Existing single-modal perception methods cannot provide high-precision micro-motion characteristics and macro-space occupancy status for complex mixed traffic flow, resulting in a lack of reliable data foundation for subsequent control systems. Therefore, this paper proposes: acquiring point cloud data and image data collected by the roadside perception unit at the target intersection and performing spatiotemporal registration; extracting micro-trajectory aggregation data of non-motorized vehicle groups and macro-spatiotemporal distribution data of adjacent motorized vehicle lanes from the registration results, specifically including: Extract pixel-level two-dimensional semantic contour data from image data and three-dimensional depth coordinate data from point cloud data; The pixel-level two-dimensional semantic contour data is mapped and matched with the three-dimensional depth coordinate data to generate a four-dimensional spatiotemporal matrix data containing target attributes and displacement vectors. From the four-dimensional spatiotemporal matrix data, the displacement vectors of non-motorized vehicle targets are clustered to generate micro-trajectory aggregation data; Vehicle bounding box sequence data located within the preset physical motor vehicle lane coordinate range are extracted from the four-dimensional spatiotemporal matrix data as macro-spatiotemporal distribution data.

[0021] Pixel-level two-dimensional semantic contour data: refers to the set of two-dimensional coordinates that reflect the shape of the outer edge of a target object (such as a person, bicycle, or car) and its category attributes after pixel-by-pixel classification and recognition of two-dimensional images captured by a high-definition camera.

[0022] Three-dimensional depth coordinate data: refers to a set of three-dimensional points in space generated by scanning equipment such as LiDAR, which contains the specific location of the target object in the real physical world (i.e., the distance in the horizontal, vertical and longitudinal directions).

[0023] Four-dimensional spatiotemporal matrix data refers to a multi-dimensional data structure formed by fusing the three-dimensional spatial coordinates of a target with a one-dimensional timestamp and the target's category attributes. It is used to uniformly describe the physical state of a target at a specific point in time.

[0024] Vehicle bounding box sequence data: refers to the set of geometric cuboid (length, width, height) parameters and their center coordinates arranged in chronological order to completely wrap the external contour of a single motor vehicle in three-dimensional space.

[0025] Step 1: Time Synchronization and Data Extraction: The system extracts image data and point cloud data from the camera and LiDAR respectively, and calculates the absolute difference between their timestamps to determine whether they belong to the same temporal frame. ;in, This represents the time synchronization deviation value. The system timestamp for acquiring the current point cloud frame for the lidar; The system timestamp for capturing the current image frame from the camera. The system has a preset time deviation threshold. (The value range is usually set between 0.01 seconds and 0.05 seconds). When At that time, the two sets of data are determined to be synchronous and valid. Because the sampling frequencies of heterogeneous sensors differ, direct matching can lead to spatial ghosting. By calculating the time deviation and setting a physical threshold, asynchronous and invalid data can be eliminated, ensuring the accuracy of subsequent spatial registration.

[0026] Step 2: Spatial mapping and matching to generate four-dimensional spatiotemporal matrix data: Under the premise of effective time synchronization, the system uses the pre-calibrated camera intrinsic and extrinsic parameter matrix to project the three-dimensional depth coordinates in the point cloud data into the two-dimensional image coordinate system, so that the physical coordinates are aligned with the pixel-level semantic categories.

[0027] The formula for spatial projection mapping is: ; in, This represents the depth value of the projection point in the camera coordinate system. The horizontal pixel coordinates of the projection point in the two-dimensional image; The vertical pixel coordinates of the projection point in the two-dimensional image; This is the intrinsic parameter matrix of the camera; This is the extrinsic parameter matrix between the lidar and the camera; , , These represent the horizontal, vertical, and lateral coordinates of the 3D depth coordinate data in the real-world coordinate system.

[0028] Through the above mapping, if the coordinates Within the semantic contour of "non-motorized vehicle" in a two-dimensional image, the "non-motorized vehicle" attribute is assigned the corresponding three-dimensional coordinates. And add a timestamp to generate a four-dimensional spatiotemporal matrix data unit: ;in, For the first A four-dimensional spatiotemporal matrix data unit of an object at this moment; This method uses semantic category attributes (such as non-motorized vehicles and motorized vehicles) obtained through mapping. By projecting the dot product of three-dimensional space onto a two-dimensional plane to find the corresponding category, it solves the physical defects of lidar lacking semantics and cameras lacking depth, and achieves accurate linkage between physical size and target attributes.

[0029] Step 3: Extracting micro-trajectory aggregation data: The system filters out category attributes from the four-dimensional spatiotemporal matrix data. For the target "non-motorized vehicle", calculate its displacement vector in two adjacent frames and perform clustering.

[0030] The formula for calculating the displacement vector of a single unit is: ; in, Let be the displacement vector magnitude of the non-motorized vehicle in a two-dimensional plane; and The physical coordinates of the target at the current moment; and The physical coordinates of the target at the previous moment.

[0031] Then, the physical distance between two adjacent non-motorized vehicle targets is calculated. If this distance is less than a preset group clustering threshold... When the value range is typically 0.5 meters to 1.5 meters, calibrated according to road congestion, it is determined that the two belong to the same group, generating micro-trajectory aggregation data containing all displacement vectors of the group. By calculating the time-series coordinate difference, accurate motion vectors are obtained. Clustering using distance thresholds can structure scattered individual data into group data, providing a mathematical basis for subsequent assessment of the group's dispersion.

[0032] Step 4: Extract macroscopic spatiotemporal distribution data: The system presets the physical vehicle lane coordinate range, which is composed of the vertex coordinates of multiple reference polygons. The system filters category attributes. For the target "motor vehicle", extract its geometric dimensions and generate a bounding box. The formula for representing a motor vehicle bounding box sequence is: ; in, For the first Data on the body kit of a vehicle; , The physical coordinates of the geometric center of the motor vehicle; , , These are the physical dimensions of the vehicle, extracted from three-dimensional depth coordinates: length, width, and height. If... and If the data falls within the preset polygonal area of ​​the motor vehicle lane, it will be recorded as macroscopic spatiotemporal distribution data.

[0033] Example 1: Time Synchronization Determination: Assume that in a certain period, the timestamp of the point cloud collected by the lidar... Seconds, timestamps of the camera capturing images Second.

[0034] Calculate the time synchronization deviation value using the formula: Second.

[0035] Preset time deviation threshold Seconds. Due to The system determined that the two frames of data were time-synchronized and valid.

[0036] Spatial mapping matching: Extracting the 3D coordinates of a target point in a point cloud: rice, rice, Meters. Using the projection formula of the intrinsic and extrinsic parameter matrix, the pixel coordinates of this point in the two-dimensional image are calculated as follows: , The system queries the semantic segmentation results of the image and finds pixels. Located within the semantic outline of "non-motorized vehicles (electric bicycles)". The system assigns category attributes to this point, generating a four-dimensional spatiotemporal matrix data unit: .

[0037] Extracting micro-trajectory aggregation data: in the previous valid frame ( At (seconds), the coordinates of the non-motorized vehicle are rice, rice.

[0038] The system measured a physical distance of 0.8 meters between the system and another adjacent non-motorized vehicle, which is less than the preset group clustering threshold. Since the displacement vectors of the two are combined, microscopic trajectory aggregation data is generated.

[0039] Extracting macroscopic spatiotemporal distribution data: The system identifies a target set of category "motor vehicle" in the four-dimensional spatiotemporal matrix and calculates its geometric center coordinates as follows: rice, Meters, vehicle dimensions meters, width meters, height Meters. The system presets the longitudinal boundary of the motor vehicle lane as... Rice. Due to Meters, determining that the target is located within the vehicle lane, generating bounding box sequence data. And recorded as macroscopic spatiotemporal distribution data.

[0040] In the aforementioned technology, visual and lidar data from the same moment are acquired through timestamp registration. Three-dimensional point cloud coordinates are converted into two-dimensional pixel coordinates using intrinsic and extrinsic parameter matrices for spatiotemporal registration. The semantic category of the image is fused with the spatial coordinates of the point cloud to generate a four-dimensional spatiotemporal matrix containing category, coordinates, velocity, and heading. Based on this, the mixed traffic flow is structurally separated using category identifiers, constructing a micro-trajectory aggregation data matrix representing the instantaneous movement trend of individual non-motorized vehicles, and a macro-spatiotemporal distribution tuple set representing the space occupancy of motorized vehicle lanes. This eliminates the perception blind spots and measurement errors of single sensors, enabling the system to acquire structured traffic data with absolute physical scale and accurate semantic attributes. By reducing the dimensionality of the mixed traffic flow and splitting it into the micro-trajectories of non-motorized vehicles and the macro-distribution of motorized vehicles, direct and quantitative data input support is provided for subsequent calculations of non-motorized vehicle boundary crossing tendencies and motorized vehicle lane passage conditions, improving the targeting and computational efficiency of data processing in multimodal traffic scenarios.

[0041] Conventional traffic control systems typically rely solely on statistical data to determine road congestion, lacking the ability to analyze the stability of the movement within non-motorized vehicle (NMO) groups. When congestion occurs, the system cannot quantify the specific spatial proportion of NMOs crossing into motorized vehicle lanes, nor can it accurately assess the degree of directional confusion caused by mutual avoidance within the group. Furthermore, the system lacks precise temporal extrapolation of available insertion space in motorized vehicle lanes, resulting in the inability to provide physically corresponding quantitative safety indicators for lane-borrowing decisions. Therefore, this paper proposes: performing spatial vector analysis on micro-trajectory aggregation data to extract NMO overflow occupancy rate data and group state dispersion data; and extracting available spatiotemporal gap data for motorized vehicle lanes based on macro-spatiotemporal distribution data, as detailed below: Extracting non-motorized vehicle overflow occupancy rate data, population dispersion data, and available spatiotemporal gap data, specifically including: Calculate the area occupancy rate of the coordinate set that exceeds the physical boundary of the non-motorized vehicle lane in the micro-trajectory aggregation data, and generate non-motorized vehicle overflow occupancy rate data. Extract the instantaneous velocity vector data and heading angle data of each target in the micro-trajectory aggregation data; Calculate the statistical variance of the heading angle data and the coefficient of variation of the instantaneous velocity vector data, and then fuse the two to generate the group state dispersion data; Calculate the physical distance between the first and last vehicles in the macroscopic spatiotemporal distribution data, as well as the approach speed data of the following vehicle; Based on the physical distance data between the beginning and end of the lane and the approximation speed data, the time segment of the current motor vehicle lane without vehicles passing through in the future prediction time window is calculated as usable spatiotemporal gap data.

[0042] Physical boundary of non-motorized vehicle lane: refers to the fixed geometric dividing line on the actual road surface used to separate the driving areas of non-motorized vehicles and motorized vehicles, such as the bottom edge of a physical guardrail or traffic markings painted on the road surface.

[0043] Non-motorized vehicle overflow occupancy rate data: refers to the ratio between the actual road surface area occupied by non-motorized vehicle groups that illegally cross the above-mentioned physical boundaries and enter the motorized vehicle lane, and the area of ​​the benchmark monitoring area delineated by the system in the adjacent motorized vehicle lane.

[0044] Heading angle data: refers to the angle between the instantaneous motion direction vector of a non-motorized vehicle and the preset longitudinal axis of the road standard.

[0045] Group dispersion data: a comprehensive dimensionless numerical value used to quantitatively evaluate the degree of disorder and inconsistency in the driving direction and speed of individuals within a non-motorized vehicle group.

[0046] Available spatiotemporal gap data: refers to the absolute length of time without vehicles between two moving vehicles in adjacent motor vehicle lanes, which can provide safe passage for non-motorized vehicles to cross or use the lane.

[0047] Step 1: Calculate the non-motorized vehicle overflow occupancy rate data: The system extracts the lateral physical coordinates of all non-motorized vehicle targets from the micro-trajectory aggregation data and compares them with the physical boundary coordinates of the non-motorized vehicle lane. Targets that cross the boundary and enter the motorized vehicle lane are selected, and the total two-dimensional projected area of ​​these targets is calculated to determine the occupancy rate. The formula for calculating the non-motorized vehicle overflow occupancy rate is: ; in, This data represents the overflow occupancy rate of non-motorized vehicles. The total number of non-motorized vehicles that cross the physical boundary of the non-motorized vehicle lane; For the first The physical length of the non-motorized vehicle that crossed the boundary; For the first The physical width of a non-motorized vehicle that has crossed the boundary; This refers to the area of ​​the pre-defined baseline edge monitoring zone for adjacent motor vehicle lanes. Typically, a rectangular area extending backward from the start / stop line at the target intersection is selected. Its length usually ranges from 15 to 30 meters, and its width is the standard width of the adjacent motor vehicle lane (e.g., 3.5 meters). Discrete boundary-crossing coordinates are converted into a percentage value with physical area significance. This percentage directly reflects the actual extent to which non-motorized vehicles encroach on motor vehicle lane space, eliminating assessment errors caused by varying target volume.

[0048] Step 2: Extracting and calculating group dispersion data: The system performs statistical analysis on the heading angle and instantaneous velocity vector in the micro-trajectory aggregation data. First, the variance of the group heading angle is calculated to assess the degree of directional disorder. The formula for calculating the statistical variance of the heading angle is: ; in, This is the statistical variance data for the heading angle data; This represents the total number of individual non-motorized vehicles within the current statistical window. For the first Heading angle data for individual non-motorized vehicles; This is the arithmetic mean of the heading angles of the group.

[0049] Subsequently, the system calculates the discrete coefficients of the instantaneous velocity vector to assess the differences in acceleration and deceleration states of the group. The formula for calculating the velocity discrete coefficient data is as follows: ; in, The discrete coefficients of the instantaneous velocity vector data; The standard deviation of the instantaneous velocity of the group; It is the arithmetic mean of the instantaneous velocities of the group.

[0050] Finally, the system performs a weighted fusion of the chaotic features of direction and velocity. The formula for calculating the group state discreteness data is: ; in, This is group-state discreteness data; The preset heading angle variance weighting constant; This is a preset weighting constant for the velocity dispersion coefficients. The weight values ​​must satisfy... Because non-motorized vehicles are moving in groups, the probability of a collision caused by a sudden change in direction is higher than that caused by a change in speed. The value range is usually set to 0.6 to 0.8. The values ​​are set to 0.2 to 0.4. Variance quantifies the yaw amplitude, and the coefficient of variation eliminates the bias caused by different absolute speed bases. Combining the two into a single scalar can accurately describe the "internal friction" and disorderly movement tendency within the non-motorized vehicle group caused by congestion and avoidance.

[0051] Step 3: Calculate available spatiotemporal gap data: Based on macroscopic spatiotemporal distribution data, the system identifies adjacent targets within the same lane, extracts the rear coordinates of the preceding vehicle and the front coordinates of the following vehicle, and calculates the physical distance between them. The formula for calculating the physical distance between the two targets is: ; in, This refers to the physical distance between the front and rear ends of two adjacent vehicles. The longitudinal physical coordinates of the geometric center of the vehicle in front; This refers to the physical length parameter of the vehicle in front; The longitudinal physical coordinates of the geometric center of the rear vehicle; This refers to the physical length parameter of the following vehicle.

[0052] Subsequently, considering the speed of the following vehicle, the estimated duration for this spatial segment to become a car-free zone in the future is calculated. The formula for calculating this using spatiotemporal gap data is as follows: ; in, For usable spatiotemporal gap data; This involves calculating the approaching speed of the following vehicle. The pure spatial distance data is divided by the following vehicle's speed to convert it into a time-based buffer period. This calculation is based on rigorous kinematics, providing the system with a physically corresponding quantitative safety indicator for subsequently determining whether the vehicle can successfully change lanes before the other vehicle arrives.

[0053] Example 2: Continuing with the spatiotemporal matrix and bounding box data generated above. It is known that the longitudinal distribution of motor vehicle lanes is on the horizontal axis (road direction), and the physical boundary between non-motorized vehicle lanes and motor vehicle lanes is set as the horizontal coordinate. rice( (meters for motor vehicle lanes).

[0054] Calculating the non-motorized vehicle overflow occupancy rate: The system detects the lateral coordinates of a non-motorized vehicle from the micro-trajectory aggregation data. The three non-motorized vehicles were identified as exceeding the designated boundary by meters. They are respectively: , , The system presets the baseline monitoring area for motor vehicle lanes. Calculate the total overflow projected area using the formula: .

[0055] Substitute into the market share formula: (That is, the overflow occupancy rate is approximately 4.38%).

[0056] Extract and calculate group state discreteness: The system extracts the discreteness within the current window. The motion characteristics of the non-motorized vehicles. Their heading angles are as follows: , , , Arithmetic mean Substituting into the variance formula: .

[0057] Their instantaneous speeds are 3.0 m / s, 3.5 m / s, 4.0 m / s, and 5.5 m / s, respectively.

[0058] Arithmetic mean meters per second, calculate the standard deviation. meters per second. Substituting into the velocity dispersion coefficient formula: Preset weights , Substituting into the group state discreteness fusion formula: This figure indicates that there is significant disarray or shift within the current group.

[0059] Calculate the available spatiotemporal gap: The system identifies the longitudinal coordinates of the geometric center of the following vehicle (vehicle 2) in the motor vehicle lane. Meters, length meters, approaching speed meters per second.

[0060] At the same time, the system detected another vehicle (vehicle 1) ahead in the same lane, whose geometric center longitudinal coordinates... Meters, length Meters. Substitute into the formula for the physical distance between the first and last ends: rice.

[0061] Substitute into the spacetime gap formula: Second.

[0062] Calculations show that the current lane can provide approximately [a certain amount of space] before vehicles behind arrive. Absolutely safe insertion time in seconds.

[0063] In the aforementioned technology, the coordinate set of non-motorized vehicles crossing physical boundaries is screened through spatial vector comparison, and their area ratios are calculated to quantify the degree of overflow. The chaotic movement characteristics of the group are quantified by extracting the variance of the group's heading angle and the discrete coefficient of its instantaneous velocity and performing weighted fusion. The time segments in the motor vehicle lane where no vehicles pass are obtained by extracting the physical distance between the head and tail of adjacent motor vehicles and the speed of the following vehicle, using kinematic division. This transforms the complex lateral crossing behavior of non-motorized vehicles into specific area ratio data, the disordered movement of the group into numerical discreteness indicators, and the remaining capacity of the motor vehicle lane into specific time gap data. The extraction of these quantitative data allows the system to objectively grasp the tendency of non-motorized vehicles to cross boundaries and the carrying capacity of the motor vehicle lane, providing accurate mathematical input parameters for subsequent coupled calculations of mixed flow conflict risk assessment values ​​and equivalent lane-borrowing potential energy.

[0064] In complex mixed traffic scenarios, the tendency of non-motorized vehicle groups to cross boundaries is a subjective expansion demand, while the objective traffic conditions of the motorized vehicle lanes and the degree of chaos within the group represent objective safety constraints. Existing traffic control algorithms typically treat these heterogeneous data as independent judgment conditions and make sequential judgments, lacking a mechanism to mathematically integrate expansion demand and safety constraints. This results in the system's inability to comprehensively assess the feasibility of "borrowing the current motorized vehicle lane gap under the current congestion and chaos," easily leading to overly conservative or potentially unsafe control commands. Therefore, this paper proposes: cross-coupling group state discreteness data with available spatiotemporal gap data to generate boundary expansion impedance coefficient data; and using this boundary expansion impedance coefficient data to dampen and attenuate the non-motorized vehicle overflow occupancy rate data to obtain equivalent lane-borrowing potential energy data, as detailed below: The steps for generating equivalent potential energy data are as follows: The basic repulsion data of the motor vehicle lane is obtained by calculating the reciprocal of the sum of the available spatiotemporal gap data and the preset minimum smoothing constant. The basic repulsion data and group state dispersion data are input into the preset risk repulsion model to calculate the mixed flow conflict risk assessment value. The mixed flow conflict risk assessment value is then normalized to generate boundary expansion impedance coefficient data within a set range. The difference between the boundary expansion impedance coefficient data and the unit constant 1 is calculated to obtain the attenuation multiplier; The non-motorized vehicle overflow occupancy rate data is multiplied with the attenuation multiplier to obtain the attenuated energy characteristic value; The decayed energy characteristic value is output as the equivalent borrowed potential energy data.

[0065] Basic rejection rate data: A quantitative indicator used to characterize the physical rejection strength of non-motorized vehicle lane borrowing behavior by the current traffic flow in the motor vehicle lane. This value is inversely proportional to the available idle time of the motor vehicle lane.

[0066] Environmental damping coefficient: a penalty multiplier that characterizes the external environmental conditions and is used to quantify the additional risk amplification effect of objective weather or physical environmental factors such as poor lighting and slippery road surfaces on detour behavior.

[0067] Discreteness gain weight: An exponential adjustment parameter used to exponentially amplify the sensitivity of the motor vehicle lane exclusion to the overall risk assessment when the movement of non-motorized vehicle groups is extremely disordered.

[0068] Mixed flow conflict risk assessment value: A raw calculation result without dimensionless standardization, used to comprehensively evaluate the probability and severity of spatial conflicts between motor vehicles and non-motor vehicles under current spatial, group order and environmental conditions.

[0069] Boundary expansion resistance coefficient data: The original risk assessment value is mapped to a normalized value within a standard unit range, representing the percentage of resistance preventing the physical boundary of the non-motorized lane from expanding outward.

[0070] Equivalent lane-borrowing potential energy data: The final output scalar of the system represents the actual physical tendency and feasibility of the non-motorized vehicle group to expand into adjacent lanes after deducting all objective safety risk resistance.

[0071] Step 1: Calculate basic repulsion data: The system obtains the available spatiotemporal gap data calculated in the upstream steps, adds it to a preset minimal constant, and then performs a reciprocal operation.

[0072] The formula for calculating basic rejection data is: ; in, Based on basic rejection data; For usable spatiotemporal gap data; A minimum smoothing constant is preset. The smoothing constant is typically set between 0.01 and 0.1. The larger the spatiotemporal gap, the smaller the repulsion should be; therefore, a reciprocal relationship is used. (Introduction...) In order to be in In the extreme case where the value equals 0 (i.e., the motor vehicle lane is completely congested with no gaps), this prevents the underlying algorithm from encountering a division-by-zero overflow error and ensures the continuous operation of the system.

[0073] Step 2: Calculate the risk assessment value of the mixed flow conflict and generate boundary expansion impedance coefficient data: The system extracts the basic repulsion data and group state dispersion data into the preset risk repulsion model for cross-product calculation. The specific formula of the risk repulsion model is as follows: ; in, This is the risk assessment value for mixed flow conflicts; This refers to the environmental damping coefficient. Based on basic rejection data; Weights for discreteness gain; This is group state dispersion data. Environmental damping coefficient. Based on meteorological sensor data, a value of 1.0 is used for good weather conditions, while the value ranges from 1.2 to 2.0 for severe weather conditions such as rain and snow; dispersion gain weight. The value typically ranges from 1.0 to 3.0, depending on the road speed limit level. This formula mathematically integrates three different dimensions of safety constraints (environment, lane margin, and internal group disorder). Using exponential calculations ensures that when the exclusion factor is high, the risk assessment value increases rapidly and non-linearly, thus providing a higher safety redundancy.

[0074] Subsequently, the system performs linear normalization on the risk assessment values. The normalization formula for the boundary expansion impedance coefficient data is: ; in, This is the boundary expansion impedance coefficient data (by algorithm constraints, if the calculated result is greater than 1, the output will be forcibly truncated to 1). This is the minimum risk calibration value (usually set to 0) preset by the system under safe conditions based on historical data. This is the maximum risk calibration value preset for the system under hazardous conditions. The original risk value with no upper limit is compressed into a standard range of 0 to 1, and it is transformed into a uniform percentage resistance form, providing a mathematically logical multiplier for subsequent attenuation calculations.

[0075] Step 3: Calculate the attenuation multiplier: The system uses the unit constant to subtract the impedance coefficient mentioned above to determine the allowable energy retention ratio through the bypass. The formula for calculating the attenuation multiplier is: ; in, For decay multipliers; This refers to the boundary expansion impedance coefficient data. The higher the impedance (risk), the smaller the attenuation multiplier after subtraction, indicating a lower proportion of allowable passage demand.

[0076] Step 4: Obtaining Equivalent Lane-Sharing Potential Energy Data: The system extracts the non-motorized vehicle overflow occupancy rate data calculated in the previous step and multiplies it by the attenuation multiplier. The formula for calculating the equivalent lane-sharing potential energy data is: ; in, This is equivalent potential energy data (i.e., the energy characteristic value after decay). This data represents the overflow occupancy rate of non-motorized vehicles. This is a damping multiplier. The subjective out-of-bounds demand (overflow rate) and objective safety constraints (damping multiplier) are physically reduced and coupled through multiplication. Through damping attenuation, a single-valued scalar is obtained, serving as the sole criterion for the subsequent control system to determine whether to issue an execution command.

[0077] Example 3: Continuing from the scene parameters in the previous step. It is known that the available spatiotemporal gap data of the currently adjacent motor vehicle lanes has been obtained through previous calculations. Seconds, non-motorized vehicle overflow occupancy rate data (i.e., 4.38%), group dispersion data .

[0078] Calculate the basic repulsion data: system preset smoothing constant .

[0079] Substitute into the formula: .

[0080] Calculate risk assessment value and impedance coefficient: The system detects that the current weather is cloudy with light rain, and the preset environmental damping coefficient is invoked. Road speed limit level corresponding to dispersion gain weight .

[0081] Substituting into the risk repulsion model formula: .

[0082] The system presets the historical minimum risk calibration value. Maximum risk rating .

[0083] Substitute into the normalization formula: .

[0084] Calculate the decay multiplier: The results indicate that, due to the current light rain environment and moderate disorder within the population, the system's safety margin only allows for the release of 63.5% of the detour demand.

[0085] Obtain equivalent bypass potential energy data: Combine the initial overflow occupancy rate with the decay multiplier: The system ultimately calculates the equivalent borrowing potential energy for the current cycle to be 0.0278. This value will then be compared with the dynamic mapping threshold to determine whether to generate the optimal borrowing boundary.

[0086] Existing traffic cooperative control systems often treat all motor vehicles as equal point masses when assessing mixed conflict risks, considering only speed and distance, neglecting the significant differences in physical momentum between large vehicles (such as dump trucks and buses) and small cars. Simultaneously, when assessing non-motorized vehicle groups, the system lacks consideration of the smoothness of movement within the group and fails to identify the reduced avoidance ability caused by congestion. This leads to the system still using conventional static weights to release right-of-way when large vehicles approach at high speed or when non-motorized vehicle groups move slowly, creating significant safety hazards. Therefore, this paper proposes that the discrete gain weights are generated through dynamic modulation, specifically including: Acquire the preset basic discrete gain constant data and the three-dimensional bounding box volume data and instantaneous approach velocity data of the motor vehicle target approaching the spatiotemporal gap within the corresponding time period of the available spatiotemporal gap data; The three-dimensional bounding box volume data and the instantaneous approximation velocity data are weighted and multiplied to generate the approximation momentum threat index data; Downsampling analysis is performed on the instantaneous velocity vector data of targets in the micro-trajectory aggregation data to extract the proportion of slow targets that are below the preset normal traffic speed threshold; Map the slow-moving target percentage data to the group vulnerability compensation coefficient data; Based on the approximation momentum threat index data and the population vulnerability compensation coefficient data, an amplification multiplier is generated; The discrete gain constant data is fused with the amplification multiplier to generate discrete gain weights.

[0087] Basic discrete gain constant data: The system pre-sets a benchmark value based on the physical shape of the intersection (such as lane width and turning radius) as the initial reference point for calculating the final discrete gain weight.

[0088] 3D bounding box volume data: The geometric volume obtained by extracting the physical length, width, and height of the target vehicle's bounding box and multiplying them. This volume is positively correlated to some extent with the vehicle's physical mass (e.g., a large truck has a much larger volume than a small car).

[0089] Approaching Momentum Threat Index: A quantitative indicator reflecting the potential destructive power of approaching vehicles. Since directly measuring the actual mass of a vehicle is difficult, the system uses bounding box volume to represent mass, combined with velocity, to simulate the concept of momentum in physics.

[0090] Slow-speed target percentage data: In the non-motorized vehicle group, the ratio of the number of target individuals whose instantaneous driving speed is lower than the normal traffic speed standard set by the system to the total number of samples in the group.

[0091] Group vulnerability compensation coefficient data: This is a value generated based on the mapping of the proportion of slow-moving targets. The slower the overall speed of the non-motorized vehicle group, the more crowded or hesitant the group is, and the more vulnerable they are to injury when faced with approaching motor vehicles (i.e., higher vulnerability). This coefficient is needed for safety compensation.

[0092] Amplification multiplier: A dynamic adjustment scaling factor generated by combining the momentum threat of external motor vehicles with the vulnerability of the internal non-motorized vehicle group, used to amplify the basic constant.

[0093] Step 1: Calculate the Approach Momentum Threat Index Data: From the previously acquired macroscopic spatiotemporal distribution data, the system extracts the length, width, and height dimensions of the vehicle (following vehicle) approaching the spatiotemporal gap, and calculates its 3D bounding box volume data. Subsequently, it performs a weighted product calculation with the vehicle's instantaneous approach velocity data. The formula for calculating the Approach Momentum Threat Index data is: ; in, To approximate the momentum threat index data; The preset momentum dimension conversion weights typically range from 0.005 to 0.05. To approximate the physical length of a motor vehicle; To approximate the physical width of the motor vehicle; To approximate the physical height of motor vehicles; To approximate the instantaneous approach speed data of a motor vehicle. According to the principles of physics, the destructive force of a vehicle collision depends on its mass and velocity. Utilizing volume... As a reasonable approximation of mass, it is multiplied by velocity to simulate momentum characteristics. This formula enables the system to output a higher threat index when facing large or high-speed vehicles, thereby increasing the stringency of safety judgments.

[0094] Step 2: Extracting the proportion of slow-moving targets: To reduce system computational consumption, the system downsamples and extracts the instantaneous velocity vectors of non-motorized vehicles from the micro-trajectory aggregation data. The number of individuals with speeds below a set threshold in the sampled data is counted, and their proportion is calculated. The formula for calculating the proportion of slow-moving targets is: ; in, This data represents the percentage of slow-moving targets. For the sampled data, the instantaneous speed is lower than the preset normal passage speed threshold. The number of individual non-motorized vehicles; This represents the total number of individual non-motorized vehicle samples extracted through downsampling. Normal traffic speed threshold. The speed limit is typically set at 50% to 60% of the design speed limit for non-motorized lanes (e.g., 3.0 m / s to 4.0 m / s). Downsampling reduces computational load. Statistical analysis of the percentage of individuals traveling below normal speeds directly reflects whether the non-motorized vehicle group is currently experiencing congestion, stagnation, or hesitation.

[0095] Step 3: Generating Group Vulnerability Compensation Coefficient Data: The system converts the slow target proportion data calculated above into compensation coefficients through a linear mapping relationship. The formula for calculating the group vulnerability compensation coefficient data is as follows: ; in, Data for the group vulnerability compensation coefficient; This is a preset vulnerability mapping constant, typically ranging from 0.5 to 1.5; This data represents the proportion of slow-moving individuals. Using a base of 1, the coefficient is set to 1 when the group moves smoothly (slow-moving proportion is 0), without additional compensation. As the number of slow-moving individuals increases, the group's risk avoidance ability decreases. The proportion is linearly converted to a compensation coefficient greater than 1 through addition to quantify the group's vulnerability.

[0096] Step 4: Generating the Amplification Multiplier: The system multiplies the approximate momentum threat index data with the population vulnerability compensation coefficient data. The formula for calculating the amplification multiplier is: ; in, To amplify the multiplier; To approximate the momentum threat index data; This data represents the group vulnerability compensation coefficient. It combines the "external objective threat" posed by motor vehicles with the "internal subjective vulnerability" of the non-motorized vehicle group through a positive multiplication. If both are high, the multiplier will increase non-linearly, providing ample safety margin.

[0097] Step 5: Generate Discrete Gain Weights: The system performs a fusion operation between the pre-set basic discrete gain constant data and the calculated amplification multiplier. The formula for calculating the discrete gain weights is: ; in, The discrete gain weight for the final output (this sign is consistent with the risk repulsion model variables); The basic discrete gain constant data is calibrated according to the intersection safety level, and the value range is usually from 0.8 to 1.5; To amplify the multiplier, the static fundamental constants are dynamically modulated using the multiplier to generate weight parameters adapted to the current instantaneous road conditions. These parameters are then input into the preceding risk repulsion model as an exponential term, dynamically adjusting the model's sensitivity to repulsion.

[0098] By extracting the 3D bounding box volume of motor vehicles and weighting it with velocity, a momentum threat index simulating physical destructive forces is constructed. By downsampling and thresholding the speed of non-motorized vehicle groups, the proportion of slow-moving vehicles is extracted to quantify the group's vulnerability. Finally, these two heterogeneous risk characteristics are multiplied and fused to generate an amplifying multiplier, which is then used to dynamically modulate the basic discrete gain constant in real time. This gives the traffic control system the ability to perceive and judge at the physics level. When an approaching motor vehicle is large and fast, or when there is a large-scale slowdown and congestion in the non-motorized vehicle group, the system can quickly increase the discrete gain weight using this algorithm. Since this weight is exponential in the risk repulsion model, its increase will cause the system's risk assessment value to rise sharply, thereby cutting off lane-changing permission in advance and ensuring traffic safety under complex physical conditions.

[0099] Example 4: Continuing with the motor vehicle and non-motor vehicle data extracted earlier. The approximate motor vehicle dimension is known to be length. meters, width meters, height meters, approaching speed The instantaneous speeds of the sampled non-motorized vehicles were 3.0, 3.5, 4.0, and 5.5 m / s.

[0100] Calculate approximate momentum threat index data: preset momentum dimension transformation weights .

[0101] First, calculate the volume of the 3D bounding box: cubic meter.

[0102] Substitute into the formula: .

[0103] Extracting the percentage of slow-moving targets: Preset normal traffic speed threshold meters per second. During downsampling... Among the individuals, those with speeds of 3.0 m / s and 3.5 m / s are below the threshold, therefore the number of slow-moving targets is low. Substitute into the formula: (That is, the slow target accounts for 50%).

[0104] Mapping to generate population vulnerability compensation coefficient data: Preset vulnerability mapping constant .

[0105] Substitute into the formula: This figure indicates that, since half of the targets in the group move slowly, a 50% increase in security compensation is needed to increase their vulnerability.

[0106] Generate amplifying multipliers: Substitute into the formula: .

[0107] Generate discrete gain weights: Preset basic discrete gain constant data .

[0108] Substitute into the formula: .

[0109] The calculated discrete gain weight for the current period is 2.187. This value is consistent with the risk repulsion model parameters in the example above. This is consistent with the observation that, under conditions of rapid approach by medium-sized vehicles and partial congestion among non-motorized vehicles, the system significantly improves the sensitivity of risk weights through dynamic modulation.

[0110] In the aforementioned technology, a risk repulsion model is introduced to cross-product and couple the basic repulsion degree characterizing the traffic conditions of motor vehicle lanes with the dispersion degree characterizing the chaotic state of non-motorized vehicle groups, calculating a comprehensive impedance coefficient that includes environmental and group dynamic characteristics. Subsequently, this impedance coefficient is used to construct an attenuation multiplier, which is then used to mathematically dampen and reduce the dimensionality of the purely physical overflow occupancy rate of non-motorized vehicles, resulting in a single, comprehensive potential energy evaluation index. This achieves effective dimensionality reduction and feature fusion of multi-source heterogeneous traffic feature data. The previously incomparable "spatial congestion," "temporal gap," and "group directional chaos" are unified into a physically meaningful equivalent potential energy scalar. This damping attenuation mechanism ensures that even if non-motorized vehicle overflow is severe, as long as there is insufficient space in the motor vehicle lanes or the non-motorized vehicle group is extremely chaotic (impedance coefficient approaching 1), the final output potential energy data will be forcibly attenuated to below a safe threshold, guaranteeing the safety and reliability of the right-of-way dynamic replacement process from the algorithm's underlying structure.

[0111] Existing traffic control algorithms typically rely on a fixed, rigid set value as a trigger condition when allocating right-of-way or determining road conditions. This fixed value cannot adapt to the complex and variable environmental and meteorological factors in urban-rural fringe areas. For example, in rainy or snowy weather, or at night when visibility is low, slippery roads increase braking distances and obstruct visibility. If the fixed value used during sunny days is still applied to allow non-motorized vehicles to use the road, it could lead to traffic accidents. Conversely, in good weather and with good road conditions, the fixed value may be too conservative, resulting in underutilization of road space. Therefore, this paper proposes a method to determine whether the equivalent potential energy for road use exceeds a preset mapping threshold, as follows: The mapping threshold is dynamically generated and includes: The system obtains the timestamp data of the current processing time period and the preset initial threshold parameters, retrieves the corresponding historical periodic flow density distribution data, and generates the baseline saturation expected value data through time series weighted processing. The system acquires real-time ambient light intensity data, precipitation probability data, and road surface slippage classification data captured by the roadside sensing unit, performs feature fusion, and calculates environmental risk weighting coefficient data. Based on the baseline saturation expected value data and environmental risk weighted coefficient data, the initial threshold parameters are offset and gain compensated to generate a mapping threshold that matches the current road network physical environment and traffic situation in real time.

[0112] Timestamp data: Records the precise system time when the current data processing occurred, including the specific date, day of the week, and hour, minute, and second information, used to determine which stage of the traffic cycle is currently in (such as morning rush hour or off-peak hour).

[0113] Initial threshold parameter: A baseline safety limit constant pre-set by the system under ideal physical conditions (such as sunny days, sufficient sunlight, and dry road surface), representing the basic potential energy threshold required to allow non-motorized vehicles to use the road.

[0114] Historical periodic flow density distribution data: empirical data derived from statistical analysis of the spatial clustering of traffic participants with the same flow direction at the intersection over several past time periods (such as the same week or the same time period in the past four weeks).

[0115] Benchmark saturation expected value data: The benchmark expected value of the road network carrying capacity for the current period is calculated by combining the current time point with historical patterns.

[0116] Ambient light intensity data: The brightness values ​​of the external environment collected in real time by roadside sensing units (such as photosensitive sensors or vision sensors).

[0117] Precipitation probability data: Current rainfall intensity or probability index extracted based on real-time meteorological monitoring or visual water droplet recognition.

[0118] Road surface slipperiness grading data: The level of water film coverage and adhesion reduction on the road surface is assessed through visual reflective features or physical friction sensors.

[0119] Environmental risk weighted coefficient data: This is a comprehensive multiplier that represents the overall degree of external objective risk after the above-mentioned multi-dimensional environmental disadvantages are integrated and quantified.

[0120] Mapping threshold: The dynamic safety limit generated after being corrected by both historical experience and real-time environment, and used for decision-making in this processing cycle.

[0121] Step 1: Generate baseline saturation expected value data: The system obtains the current timestamp data and uses it as a search criterion to retrieve the historical periodic flow density distribution data for the corresponding time period. This data is then used as a weighting factor to perform time-series weighting processing on the preset initial threshold parameters. The formula for calculating the baseline saturation expected value data is as follows: ; in, This is the baseline saturation expected value data; This is a preset initial threshold parameter, typically set to a range of 0.05 to 0.15. The weighting factor is extracted based on historical periodic flow density distribution data, and its value typically ranges from 0.8 to 1.2. Traffic flow exhibits significant tidal periodicity. Using historical density as a weight to weight the initial threshold allows the system to anticipate the traffic saturation of the current period, avoiding the use of unchanging judgment criteria and giving the threshold a time-dimensional adaptability.

[0122] Step 2: Calculate the environmental risk weighting coefficient data: The system extracts real-time ambient light intensity data, precipitation probability data, and road surface slippage level classification data captured by the roadside sensing unit. First, these heterogeneous data are normalized into individual risk indices between 0 and 1 (higher values ​​indicate more severe environmental conditions), and then feature fusion is performed. The formula for calculating the environmental risk weighting coefficient data is as follows: ; in, This is data on environmental risk weighting coefficients; A light risk index mapped from ambient light intensity data; A rainfall risk index mapped from precipitation probability data; A wet skid risk index mapped to road surface wetness classification data; , , The three environmental risks mentioned above are respectively weighted constants. The weights of the three factors must satisfy... Because slippery road surfaces have the most direct physical impact on braking distance, The value is usually the highest, for example, set to . , , Different meteorological conditions affect traffic safety through different mechanisms. Using a linear weighted fusion formula, the system transforms multi-source, heterogeneous meteorological sensing data into a unified comprehensive risk coefficient, achieving mathematical quantification of environmental hazard levels.

[0123] Step 3: Generating the dynamically matched mapping threshold: The system uses environmental risk weighted coefficient data to offset and compensate for the gain of the baseline saturation expected value data obtained in Step 1 (i.e., reverse suppression), thus obtaining the final dynamic threshold. The formula for calculating the mapping threshold is: ; in, The mapping threshold is dynamically generated; This is the baseline saturation expected value data; This is data on environmental risk weighting coefficients. When environmental risk increases (... When (increases), The value decreases, resulting in a lower final calculated mapping threshold. The threshold decreases. A smaller threshold means that the system's requirements for safety thresholds become more stringent (requiring higher potential energy to break through), thus compensating for the risks and hidden dangers brought about by the physical environment through mathematical logic.

[0124] Step 4: Execute the logical judgment of equivalent borrowing potential energy: The system obtains the equivalent borrowing potential energy data calculated in the previous step and compares it with the real-time generated mapping threshold. The logical judgment conditions are as follows: ; in, This is equivalent to borrowing potential energy data; This is a dynamically generated mapping threshold. This inequality is the core judgment node of the system control flow. If it holds, it is determined that the current borrowing potential energy is sufficient to overcome the comprehensive risks under the current environment and spatiotemporal conditions, allowing the system to generate the borrowing boundary; if it does not hold, it is determined that the risk is too high or the potential energy is insufficient, maintaining the static right-of-way.

[0125] Example 5: Continuing from the previous text, it is known that the current equivalent potential energy data calculated in the previous steps under the "light rain, cloudy weather" environment is... Generate baseline saturation expected value data: System preset initial threshold parameters The current timestamp indicates the early morning commute period. Historical data shows that this period is characterized by chronic congestion. The extracted time-series weighting factors... Substitute into the formula: .

[0126] Calculation of environmental risk weighting coefficient data: The sensing unit measures insufficient current lighting, resulting in a light risk index. There is currently light rain; the risk of rainfall is low. The road surface is slightly damp, indicating a high risk of slipperiness. .

[0127] The preset fusion weight is , , Substitute into the formula: .

[0128] This figure indicates that the current meteorological environment presents a relatively high proportion (51%) of additional risks and hidden dangers.

[0129] Generate a dynamically matched mapping threshold: fuse the above benchmark value with the risk coefficient. Substitute into the formula: Calculations show that, due to the influence of light rain, the current detour mapping threshold is strictly compressed to 0.0353.

[0130] Execution logic judgment: Apply the equivalent potential energy data obtained in the previous step. With the current mapping threshold Compare them.

[0131] because Logical judgment conditions This is not true.

[0132] The system's decision was "No". Although there was some spillover from non-motorized vehicles, the environmental risk of the current light rain raised the safety threshold (leading to a tighter mapping threshold), and the approaching motor vehicles caused excessive impedance (the attenuation multiplier reduced the potential energy). Therefore, the system ultimately determined that it was not allowed to allow lane sharing and instead switched to executing the static right-of-way maintenance command.

[0133] In the aforementioned technology, by introducing timestamps and historical periodic flow density distribution data, the static initial setpoint is transformed into a benchmark expected value with temporal characteristics. Simultaneously, three environmental state parameters—light intensity, precipitation, and road surface slippage—are introduced, and an environmental risk weighting coefficient is calculated through weighted fusion. This environmental risk coefficient is used to correct the benchmark expected value in real time, generating a dynamically fluctuating mapping threshold that changes with the external environment. Finally, this threshold is compared with the potential energy of non-motorized vehicles using the lane. This effectively solves the problem of rigid judgment conditions and endows the control system with the ability to adaptively adjust to the external physical environment. When there is severe weather such as precipitation or poor visibility, the environmental risk coefficient increases, prompting the system to automatically lower the mapping threshold, making the lane-using triggering conditions more stringent and improving traffic safety margins under severe weather conditions. When environmental conditions are good, the threshold returns to a normal level, ensuring the smoothness of spatiotemporal resource replacement and avoiding decision-making errors caused by a single static indicator.

[0134] Existing road lane-sharing schemes typically rely on static road markings or physical barriers, making dynamic adjustments impossible based on real-time traffic flow changes. When the control system determines that lane sharing is permissible through algorithms, the lack of dynamic execution equipment in the spatial dimension often necessitates relying solely on traffic lights for time allocation, failing to define new safe passage boundaries in real-time within the existing physical space. This results in non-motorized vehicles lacking clear visual references when sharing lanes, increasing the risk of lateral collisions with motorized vehicles, and making it difficult to accurately control the lane width of non-motorized vehicles, posing a risk of encroaching on the entire motorized vehicle lane. To address this, a proposed solution is to generate optimal lane-sharing boundary coordinate data based on the numerical difference between equivalent lane-sharing potential energy data and a mapping threshold, and output a light and shadow reconstruction collaborative control instruction set containing the optimal lane-sharing boundary coordinate data to the roadside light and shadow projection equipment, as detailed below: Optimal boundary coordinate data for the bypass is generated based on the numerical difference between the equivalent bypass potential energy data and the mapping threshold, specifically including: Calculate the potential energy difference data when the equivalent borrowed potential energy data exceeds the mapping threshold; Based on the preset potential energy-space mapping relationship table, query the horizontal borrowing width data that matches the potential energy difference data; The horizontal lane width data is vector-overlaid with the basic outer edge coordinate data of the non-motorized vehicle lane to generate the optimal lane boundary coordinate data.

[0135] Potential energy difference data: This refers to the specific numerical portion of the equivalent borrowing potential energy data for the current period that exceeds the dynamic mapping threshold. This value represents the net passage demand that the non-motorized vehicle group can legally and safely convert into a spatial expansion request after deducting all safety margins.

[0136] Lateral lane width data: refers to the specific physical distance (usually in meters) that non-motorized vehicles are allowed to extend laterally into the adjacent motorized vehicle lane, as calculated by the system.

[0137] Potential Energy-Space Mapping Table: A pre-defined data conversion benchmark within the system, used to linearly or piecewise convert dimensionless abstract potential energy values ​​into specific spatial distance values. For example:

[0138] Basic outer edge coordinate data: refers to the spatial coordinate set of static physical boundary lines used to divide motor vehicle lanes and non-motor vehicle lanes in the original high-precision map of the intersection.

[0139] Optimal lane-borrowing boundary coordinate data: A new set of virtual boundary coordinates obtained after vector calculation, which defines the absolutely safe zone for non-motorized vehicles to borrow lanes within the current period.

[0140] Light and shadow reconstruction collaborative control instruction set: The above-mentioned optimal bypass boundary coordinates are converted into communication protocol messages that can be recognized by the underlying hardware, which are used to drive the roadside optical projection equipment to adjust the deflection angle and focal length, thereby generating light and shadow markings of a specified shape on the road surface.

[0141] The first step is to calculate the potential energy difference data: When the system concludes in the previous judgment step that the equivalent borrowed potential energy data is greater than the mapping threshold, the system extracts these two data points and subtracts them to extract the net potential energy exceeding the safety limit. The formula for calculating the potential energy difference data is: ; in, This represents the potential energy difference data; This is equivalent to borrowing potential energy data; The mapping threshold is dynamically generated. The equivalent lane-borrowing potential energy represents the total outward expansion impulse, while the mapping threshold represents the safety tolerance limit allowed by the current environment and time and space. The difference between the two directly reflects the congestion pressure that the system needs to alleviate by giving up physical space, ensuring that only the part exceeding the safety limit is converted into the actual lane-borrowing width calculation, thus avoiding excessive occupation of motor vehicle lane space.

[0142] The second step is to query and generate the horizontal bandwidth data: The system retrieves a preset potential energy-space mapping table. In the specific calculations, the system converts the potential energy difference data into physical width using a preset mapping coefficient. The formula for calculating the lateral access width is as follows: ; in, This refers to the width of the horizontal data channel. The preset potential energy to space mapping coefficient; This represents the potential energy difference data. Mapping coefficients. The value needs to be calibrated based on the standard width of the motor vehicle lane, and is usually set in the range of 10 to 30, representing that every 0.01 potential energy difference corresponds to a lane borrowing width of 0.1 meters to 0.3 meters. At the same time, the system will... A maximum output limit (e.g., 1.5 meters) is set to prevent the shared lane area from encroaching on the entire motor vehicle lane. The abstract potential energy value at the algorithmic level is converted into geometric dimensions at the physical spatial level. This linear proportional conversion enables the allocation of shared lane width to smoothly change with the degree of non-motorized vehicle congestion, achieving fine-grained scheduling of spatial resources.

[0143] The third step is to generate the optimal lane-sharing boundary coordinate data: The system extracts the basic outer edge coordinate data of the non-motorized vehicle lane and calculates the unit normal vector perpendicular to this boundary and pointing inwards towards the motorized vehicle lane. Then, along the direction of this normal vector, the lateral lane-sharing width data is superimposed as a scalar onto the basic coordinate system. The formula for calculating the optimal lane-sharing boundary coordinate data is as follows: ; in, This provides the optimal boundary coordinates for the bypass. Based on the outer edge coordinate data; This refers to the width of the horizontal data channel. This is a lateral unit normal vector perpendicular to the base boundary and pointing towards the side of the driveway. Traditional numerical widths cannot be directly used for controlling roadside equipment. Through a vector superposition formula, the system assigns specific geospatial attributes to the width value, generating a new boundary coordinate sequence with absolute geographical location, providing precise spatial anchor points for subsequent light and shadow projection.

[0144] The fourth step is to output the light and shadow reconstruction collaborative control instruction set: The system formats and encodes the optimal lane boundary coordinate data sequence generated above, combines it with the physical installation coordinates of the light and shadow equipment at the target intersection, calculates the deflection angle and projection intensity of the equipment, packages it to generate the light and shadow reconstruction collaborative control instruction set, and sends it to the roadside light and shadow projection equipment through the communication network. The equipment then projects a bright temporary boundary line on the road surface.

[0145] Example 6: Given that the system determined not to use the route due to light rain in the previous cycle, and assuming that precipitation has stopped and sunlight has returned, the dynamic mapping threshold is... It decreased to 0.0200. Meanwhile, the influx of a large number of non-motorized vehicles caused the equivalent potential energy of lane-sharing to rise to... .at this time The system determines that passage is permitted.

[0146] Given that the coordinates of the outer edge of the non-motorized vehicle lane foundation are parallel to the road's horizontal axis, and the vertical coordinate is... rice( (Meters represent the motor vehicle lane). The unit normal vector of this straight line points inward into the motor vehicle lane, i.e. .

[0147] Calculate the potential energy difference data: This difference represents 0.0350 units of potential energy that needs to be released through physical space.

[0148] Generate horizontal bandwidth data: System preset mapping coefficients (That is, each unit difference is converted into 25 meters of width).

[0149] Substitute into the formula: rice.

[0150] Calculations show that the system allows non-motorized vehicles to borrow 0.875 meters of width from the motorized vehicle lane.

[0151] Generate optimal boundary coordinate data for the bypass: Take a point on the outer edge of the foundation and perform vector superposition calculation, setting the coordinate form as follows: .

[0152] Substitute into the formula: The calculation shows that the optimal bypass boundary is translated to the vertical coordinate. Meters away.

[0153] Generate lighting and shadow control commands: The system will use the coordinate line equation Data is packaged to generate a set of light and shadow reconstruction collaborative control instructions, which is then sent to the light and shadow projection equipment on the streetlight pole via the roadside communication unit. After receiving the instructions, the equipment adjusts the projection lens to accurately project a continuously lit light and shadow marking line onto the asphalt pavement of the motor vehicle lane, 0.875 meters outside the original non-motorized vehicle lane boundary.

[0154] In the aforementioned technology, the abstract traffic congestion potential energy is transformed into a specific physical width value by calculating the potential energy difference and combining it with a mapping coefficient. Vector superposition calculations are used to generate a new coordinate set extending towards the motor vehicle lane based on the original non-motorized vehicle lane boundary. Finally, the calculated coordinate data is converted into control commands and sent to the roadside light and shadow projection equipment, generating dynamic virtual markings on the road surface through optical projection technology. This achieves precise execution of traffic control algorithms in the physical space dimension. Transforming lane-sharing decisions into visualized road surface light and shadow boundaries provides non-motorized vehicle drivers with intuitive route guidance and clarifies the lateral safety boundaries during right-of-way replacement. This light and shadow reconstruction method can adjust the lane-sharing width in real time according to the instantaneous state of traffic flow without altering the road infrastructure, improving the utilization rate of intersection space resources and reducing the probability of conflict during lane-sharing by mixed traffic flows.

[0155] During dynamic right-of-way assignment, non-motorized vehicles often use motorized vehicle lanes. When the signal phase changes, this group is in a physical space where their right-of-way does not originally belong. If the signal timing changes according to conventional methods, lateral traffic may enter the conflict zone before the non-motorized vehicles return to their original lanes, leading to serious safety accidents. Therefore, it is proposed that the light and shadow reconstruction cooperative control instruction set also include signal phase compensation instruction data, as follows: Obtain the current traffic signal timing cycle data at the target intersection; Calculate the additional clearing time compensation data based on the optimal bypass boundary coordinate data; The extra clearing time compensation data is written as an incremental parameter into the next execution phase of the traffic signal timing cycle data to generate signal phase compensation command data.

[0156] Traffic signal timing cycle data: refers to the complete cycle plan currently running by the signal controller at the target intersection, including the green light time, yellow light time, and red light clearing time for each phase.

[0157] Additional clearing time compensation data: Because non-motorized vehicles have borrowed space in the motorized vehicle lane, the physical path required for them to clear the conflict area at the intersection has shifted or increased. The system calculates the additional safety time increment that needs to be reserved based on this.

[0158] Signal phase compensation command data: A real-time generated command message used to modify the duration of the next execution phase of the signal controller to ensure that the group using the passage can completely evacuate before the conflict phase begins.

[0159] Step 1: Obtain current traffic signal timing cycle data: The system extracts the currently executing timing scheme through a real-time communication interface with the roadside signal controller. The formula for calculating the remaining duration of the current phase is: ; in, This represents the total remaining time for the current phase distance transition; The preset green light duration for the current phase; The preset yellow light duration for the current phase; This represents the duration of the current phase execution. It clarifies the time base of the current control state, providing a basis for the timing of subsequent compensation commands and preventing timing logic disruptions caused by command delays.

[0160] Step 2: Calculate the additional clearing time compensation data: Based on the optimal lane-borrowing boundary coordinates generated in the previous steps, the system calculates the lateral offset of the lane-borrowing area relative to the original boundary, and, combined with the average clearing speed of the non-motorized vehicle group, derives the safety compensation time. The formula for calculating the additional clearing time compensation is as follows: ; in, Data to compensate for additional clearing time; The lateral bypass width data calculated in the previous steps; The average clearing speed for non-motorized vehicle groups is typically set between 3.0 m / s and 5.0 m / s. The angle of the trajectory for non-motorized vehicles returning to their original lane is preset based on the intersection geometry, typically ranging from 30 to 60 degrees. The act of using another lane alters the physical spatial position of non-motorized vehicles when the red light is on. Through kinematic calculations of lateral offset and return velocity, the additional time required for the group using another lane to return to the safe area is quantified, thereby eliminating the collision risk caused by spatial positional shifts.

[0161] Step 3: Write incremental parameters and generate compensation instructions: The system will use the calculated compensation amount as the time increment and write it into the start delay parameter of the next execution phase, or directly extend the full red clearing time of the current phase.

[0162] The formula for the total duration after the next phase adjustment is: ; in, The adjusted execution duration for the next phase; The preset duration of the next phase in the original timing scheme; To compensate for the additional clearing time data, by dynamically fine-tuning the timing cycle, the spatial "borrowing weight" is transformed into temporal "safety protection," ensuring that non-motorized vehicles have sufficient physical avoidance time before the motor vehicle conflict flow begins.

[0163] Example 7: The lateral bypass width has been calculated. rice.

[0164] Get current timing: The system reads signal data: current phase green light Seconds, yellow light seconds, executed Second.

[0165] Calculate remaining time: Second.

[0166] Calculate additional compensation: Set the average speed of non-motorized vehicles meters per second, angle of return Spend( Substitute into the formula: Seconds. Rounding down, for decimals, directly convert the decimal part to 1, and finally... 1 Generate phase compensation command: original duration of the next phase Second.

[0167] Calculate the adjusted duration: Second.

[0168] The system encapsulates this parameter, generates an instruction, and sends it to the signal controller. The controller then postpones the start time of the red light in the next phase by 1 second.

[0169] In the aforementioned technology, by acquiring the real-time timing cycle and combining it with the lateral offset distance of the optimal lane-crossing boundary, the additional time increment required for safe clearance is calculated, and this increment is used to dynamically correct the execution sequence of the signal phase. This achieves deep coordination between the "light and shadow spatial boundary" and the "signal time dimension." The compensation command ensures that the lane-crossing behavior has a complete and closed safe clearance cycle in the time dimension, eliminating the clearance time blind spot caused by changes in spatial location, and ensuring the operational safety of intersections in urban-rural fringe areas under complex right-of-way changes.

[0170] In systems that dynamically allocate right-of-way based on multimodal data, when the road traffic flow does not meet the conditions for safely borrowing adjacent lanes (e.g., dense traffic flow in motor vehicle lanes, adverse environmental conditions, or no significant overflow demand from non-motorized vehicles), the lack of a clear backoff mechanism or default state maintenance mechanism can lead to roadside enforcement equipment (such as light projection devices or traffic lights) being in a random state without instruction control. This state can cause flashing or misleading guidance of road surface boundaries, causing visual confusion for traffic participants and increasing the risk of lateral collisions in mixed traffic flows. Therefore, this paper proposes generating static right-of-way maintenance instruction data, as follows: Static right-of-way maintenance instruction data: refers to a set of data messages generated when the system determines that the current traffic situation or environmental conditions do not meet the safe lane-changing standards. This is used to control the roadside to perform hardware restoration or maintain the original physical lane division state. This instruction is designed to cut off any dynamic spatial expansion behavior.

[0171] The first step involves evaluating the logical condition and triggering a branch: the system extracts the equivalent potential energy data for the current processing cycle and the dynamically generated mapping threshold, comparing their values. When the system detects that the potential energy fails to exceed the safety threshold, it triggers a static maintenance process. The mathematical expression for the logical condition is as follows: ; in, This is equivalent to borrowing potential energy data; This is a dynamically generated mapping threshold. It acts as a safety shut-off valve for the system. When harsh environmental conditions cause the threshold to be lowered... Increased demand, or the expansion of the non-motorized vehicle group. At lower values, this inequality holds. It ensures that the system does not blindly allocate space resources when potential security risks exist.

[0172] The second step is to generate static boundary coordinate data: If the above logical conditions are met, the system discards all calculations regarding the lateral lane width and directly extracts the pre-set coordinates of the outer edge of the non-motorized lane in the high-precision road network map, assigning them as the current control boundary. The formula for assigning static boundary coordinates is as follows: ; in, This is static boundary coordinate data; This provides the basic coordinate data for the outer edge of the non-motorized vehicle lane. Direct data assignment eliminates any spatial offset. This ensures that, even when the lane-sharing conditions are not met, the target boundary calculated by the system absolutely coincides spatially with the physical markings or guardrails applied during road construction.

[0173] The third step is to extract the default timing recovery parameters: Since no lane-changing behavior occurred, non-motorized vehicles do not need to cross the additional space of the motorized vehicle lane. Therefore, the system cancels any additional clearing time compensation and directly calls the original timing duration. The formula for calling the default timing is as follows: ; in, This refers to the standard execution time data issued during this processing cycle; This is the preset duration for the next phase in the original traffic signal timing cycle data. It avoids adding unnecessary safety buffer time to the traffic lights when there is no lane-changing behavior, thus ensuring the traffic efficiency of the vehicle phase and preventing an increase in overall intersection delays.

[0174] The fourth step is to encapsulate and output static right-of-way maintenance command data: The system encapsulates the above static boundary coordinate data, default timing recovery parameters, and device reset status words in a structured manner to generate the final command message.

[0175] The set expression for instruction data is as follows: ; in, This is static right-of-way maintenance instruction data; This is static boundary coordinate data; This is standard execution time data; This sets a preset device reset status word (usually a Boolean value of 0 or a specific hexadecimal shutdown code). It integrates discrete control parameters into a single communication data packet. After this command is issued, the projection device... and Cancel the virtual road markings, and the traffic signal will... The system operates according to standard logic, restoring the entire intersection to its basic static right-of-way state.

[0176] Example 8: Continuing from the previous description of the environment and situation. Given that the previous cycle was affected by light rain and slippery road conditions, the system calculated the equivalent potential energy for lane borrowing as follows: The mapping threshold generated after environmental dimensionality reduction compensation is .

[0177] Evaluation of execution logic conditions: The system compares the two sets of data above, because... If the logical inequality holds, the system determines that the current risk outweighs the benefits of using the detour, and enters the static right-of-way maintenance process.

[0178] Generate static boundary coordinate data: The system extracts the longitudinal coordinates of the basic outer edge of the non-motorized vehicle lane from the high-precision map. Meters. Substitute into the assignment formula: That is, the control target boundary for this period is strictly maintained at 3.0 meters on the longitudinal coordinate, without any numerical deviation towards the motor vehicle lane side.

[0179] Extracting default timing recovery parameters: The system reads the original preset green light duration for the next phase from the signal controller. Second.

[0180] Substitute into the formula: Seconds. Maintain normal operation for 30 seconds.

[0181] Encapsulated output instructions: The system detected a historical device status of "Device is off / standby". Device reset status word. The output should be a shutdown code of 0xFF. The system will assemble the data into... The static right-of-way maintenance instruction data is then issued. The roadside light projection equipment remains on standby and does not project out-of-bounds markings, the traffic signal controller executes the next phase every 30 seconds, and the intersection operates in a normal physical isolation state.

[0182] In the aforementioned technology, unsafe lane-changing requests are intercepted through logical comparison of numerical values. When the equivalent lane-changing potential energy data is less than or equal to the mapping threshold, the system directly calls the basic outer edge coordinate data of the non-motorized vehicle lane without performing any lateral width superposition calculations. Subsequently, this basic coordinate data is packaged with regular signal timing data to generate standard static right-of-way maintenance instruction data, which is then sent to the underlying optical and signal execution hardware to force the right-of-way to be restored or maintained within the reference range defined by physical isolation facilities. This provides a safety net mechanism for the entire collaborative optimization control system. When the risk of mixed flow conflict is high or the demand for spatiotemporal displacement is insufficient, the system can promptly cut off lane-changing logic, relying on explicit static instructions to maintain the original physical boundary order of the road. This approach eliminates the ambiguity of boundary indication status, providing clear and consistent right-of-way feedback for both non-motorized and motorized vehicle drivers, ensuring the seriousness of basic traffic rules and traffic safety in intersection areas.

[0183] A multimodal sensing traffic flow collaborative optimization control method for urban-rural fringe areas includes the following steps: The point cloud data and image data collected by the roadside sensing unit at the target intersection are acquired and spatiotemporally registered. The micro-trajectory aggregation data of the non-motorized vehicle group and the macro-spatiotemporal distribution data of the adjacent motor vehicle lanes are extracted from the registration results. Spatial vector analysis was performed on the micro-trajectory aggregation data to extract non-motorized vehicle overflow occupancy rate data and group state dispersion data. Based on macro-spatiotemporal distribution data, available spatiotemporal gap data of motor vehicle lanes were extracted. The group state discreteness data and the available spatiotemporal gap data are cross-coupled to generate boundary expansion impedance coefficient data. The boundary expansion impedance coefficient data is then used to dampen and attenuate the non-motorized vehicle overflow occupancy rate data to obtain the equivalent lane borrowing potential energy data. Determine whether the equivalent borrowing potential energy data is greater than the preset mapping threshold; If so, the optimal boundary coordinate data of the borrowing lane is generated based on the numerical difference between the equivalent borrowing potential energy data and the mapping threshold, and a set of light and shadow reconstruction collaborative control instructions containing the optimal borrowing lane boundary coordinate data is output to the roadside light and shadow projection device. If not, then static right-of-way maintenance instruction data is generated.

[0184] The technical scope of this invention is not limited to the content described above. Those skilled in the art can make various modifications and variations to the above embodiments without departing from the technical concept of this invention, and all such modifications and variations should fall within the protection scope of this invention.

Claims

1. A multimodal sensing traffic flow collaborative optimization control system for urban-rural fringe areas, characterized in that, include: The data acquisition and processing module is used to acquire point cloud data and image data collected by the roadside sensing unit at the target intersection and perform spatiotemporal registration. From the registration results, it extracts the micro-trajectory aggregation data of the non-motorized vehicle group and the macro-spatiotemporal distribution data of the adjacent motor vehicle lanes. The feature extraction interaction module is used to perform spatial vector analysis on micro-trajectory aggregation data, extract non-motorized vehicle overflow occupancy rate data representing the area occupancy of the coordinate set that exceeds the physical boundary of the non-motorized vehicle lane in the micro-trajectory aggregation data, and group state dispersion data representing the degree of disorder of group movement. Based on macro-spatiotemporal distribution data, it extracts available spatiotemporal gap data representing the time segment without vehicles passing through the current motorized vehicle lane in the future prediction time window. The data coupling calculation module is used to cross-couple the group state discreteness data with the available spatiotemporal gap data to generate boundary expansion impedance coefficient data, which represents the percentage of resistance that prevents the physical boundary of the non-motorized vehicle lane from expanding outward. The boundary expansion impedance coefficient data is used to perform damping attenuation processing on the non-motorized vehicle overflow occupancy rate data to obtain equivalent lane borrowing potential energy data, which represents the actual physical tendency and feasibility of the non-motorized vehicle group to expand into adjacent lanes after deducting all objective safety risk resistances. The data judgment and output module is used to determine whether the equivalent borrowed potential energy data is greater than the preset mapping threshold. If so, the optimal detour boundary coordinate data is generated based on the numerical difference between the equivalent detour potential energy data and the mapping threshold, and a light and shadow reconstruction collaborative control instruction set containing the optimal detour boundary coordinate data is output to the roadside light and shadow projection device. If not, then static right-of-way maintenance instruction data is generated.

2. The multimodal sensing traffic flow collaborative optimization control system for urban-rural fringe areas according to claim 1, characterized in that: The steps for generating the equivalent potential energy data are as follows: The basic repulsion data of the motor vehicle lane is obtained by calculating the reciprocal of the sum of the available spatiotemporal gap data and the preset minimum smoothing constant. The basic repulsion data and group state dispersion data are input into the preset risk repulsion model to calculate the mixed flow conflict risk assessment value. The mixed flow conflict risk assessment value is then normalized to generate boundary expansion impedance coefficient data within a set range. The difference between the boundary expansion impedance coefficient data and the unit constant 1 is calculated to obtain the attenuation multiplier; The non-motorized vehicle overflow occupancy rate data is multiplied with the attenuation multiplier to obtain the attenuated energy characteristic value; The decayed energy characteristic value is output as the equivalent borrowed potential energy data.

3. The multimodal sensing traffic flow collaborative optimization control system for urban-rural fringe areas according to claim 2, characterized in that: The specific formula for the risk repulsion model is as follows: ； in, This is the risk assessment value for mixed flow conflicts; This refers to the environmental damping coefficient. Based on basic rejection data; Weights for discreteness gain; This is group dispersion data.

4. The multimodal sensing traffic flow collaborative optimization control system for urban-rural fringe areas according to claim 3, characterized in that: The discrete gain weights are generated through dynamic modulation, specifically including: Acquire the preset basic discrete gain constant data and the three-dimensional bounding box volume data and instantaneous approach velocity data of the motor vehicle target approaching the spatiotemporal gap within the corresponding time period of the available spatiotemporal gap data; The three-dimensional bounding box volume data and the instantaneous approximation velocity data are weighted and multiplied to generate the approximation momentum threat index data; Downsampling analysis is performed on the instantaneous velocity vector data of targets in the micro-trajectory aggregation data to extract the proportion of slow targets that are below the preset normal traffic speed threshold; Map the slow-moving target percentage data to the group vulnerability compensation coefficient data; Based on the approximation momentum threat index data and the population vulnerability compensation coefficient data, an amplification multiplier is generated; The discrete gain constant data is fused with the amplification multiplier to generate discrete gain weights.

5. The multimodal sensing traffic flow collaborative optimization control system for urban-rural fringe areas according to claim 1, characterized in that: The point cloud data and image data collected by the roadside sensing unit at the target intersection are acquired and spatiotemporally registered. From the registration results, microscopic trajectory aggregation data of the non-motorized vehicle group and macroscopic spatiotemporal distribution data of adjacent motorized vehicle lanes are extracted, specifically including: Extract pixel-level two-dimensional semantic contour data from image data and three-dimensional depth coordinate data from point cloud data; The pixel-level two-dimensional semantic contour data is mapped and matched with the three-dimensional depth coordinate data to generate a four-dimensional spatiotemporal matrix data containing target attributes and displacement vectors. From the four-dimensional spatiotemporal matrix data, the displacement vectors of non-motorized vehicle targets are clustered to generate micro-trajectory aggregation data; Vehicle bounding box sequence data located within the preset physical motor vehicle lane coordinate range are extracted from the four-dimensional spatiotemporal matrix data as macro-spatiotemporal distribution data.

6. The multimodal sensing traffic flow collaborative optimization control system for urban-rural fringe areas according to claim 1, characterized in that: Extracting non-motorized vehicle overflow occupancy rate data, population dispersion data, and available spatiotemporal gap data, specifically including: Calculate the area occupancy rate of the coordinate set that exceeds the physical boundary of the non-motorized vehicle lane in the micro-trajectory aggregation data, and generate non-motorized vehicle overflow occupancy rate data. Extract the instantaneous velocity vector data and heading angle data of each target in the micro-trajectory aggregation data; Calculate the statistical variance of the heading angle data and the coefficient of variation of the instantaneous velocity vector data, and then fuse the two to generate the group state dispersion data; Calculate the physical distance between the first and last vehicles in the macroscopic spatiotemporal distribution data, as well as the approach speed data of the following vehicle; Based on the physical distance data between the beginning and end of the lane and the approximation speed data, the time segment of the current motor vehicle lane without vehicles passing through in the future prediction time window is calculated as usable spatiotemporal gap data.

7. The multimodal sensing traffic flow collaborative optimization control system for urban-rural fringe areas according to claim 1, characterized in that: The mapping threshold is dynamically generated, specifically including: The system obtains the timestamp data of the current processing time period and the preset initial threshold parameters, retrieves the corresponding historical periodic flow density distribution data, and generates the baseline saturation expected value data through time series weighted processing. The system acquires real-time ambient light intensity data, precipitation probability data, and road surface slippage classification data captured by the roadside sensing unit, performs feature fusion, and calculates environmental risk weighting coefficient data. Based on the baseline saturation expected value data and environmental risk weighted coefficient data, the initial threshold parameters are offset and gain compensated to generate a mapping threshold that matches the current road network physical environment and traffic situation in real time.

8. The multimodal sensing traffic flow collaborative optimization control system for urban-rural fringe areas according to claim 1, characterized in that: Optimal boundary coordinate data for the bypass is generated based on the numerical difference between the equivalent bypass potential energy data and the mapping threshold, specifically including: Calculate the potential energy difference data when the equivalent borrowed potential energy data exceeds the mapping threshold; Based on the preset potential energy-space mapping relationship table, query the horizontal borrowing width data that matches the potential energy difference data; The horizontal lane width data is vector-overlaid with the basic outer edge coordinate data of the non-motorized vehicle lane to generate the optimal lane boundary coordinate data.

9. The multimodal sensing traffic flow collaborative optimization control system for urban-rural fringe areas according to claim 1, characterized in that: The light and shadow reconstruction collaborative control instruction set also includes signal phase compensation instruction data, as follows: Obtain the current traffic signal timing cycle data at the target intersection; Calculate the additional clearing time compensation data based on the optimal bypass boundary coordinate data; The extra clearing time compensation data is written as an incremental parameter into the next execution phase of the traffic signal timing cycle data to generate signal phase compensation command data.

10. A multimodal sensing traffic flow collaborative optimization control method for urban-rural fringe areas, characterized in that, Includes the following steps: The point cloud data and image data collected by the roadside sensing unit at the target intersection are acquired and spatiotemporally registered. From the registration results, the micro-trajectory aggregation data of the non-motorized vehicle group and the macro-spatiotemporal distribution data of the adjacent motor vehicle lanes are extracted. Spatial vector analysis is performed on the micro-trajectory aggregation data to extract non-motorized vehicle overflow occupancy rate data, which represents the area occupancy rate of the coordinate set that exceeds the physical boundary of the non-motorized vehicle lane, and group state dispersion data, which represents the degree of disorder in group movement. Based on macro-spatiotemporal distribution data, usable spatiotemporal gap data, which represents the time segment of no vehicle passage in the current motorized vehicle lane within the future prediction time window, are extracted. By cross-coupling the group state discreteness data with the available spatiotemporal gap data, boundary expansion impedance coefficient data is generated, which represents the percentage of resistance that prevents the physical boundary of the non-motorized vehicle lane from expanding outward. The boundary expansion impedance coefficient data is then used to dampen and attenuate the non-motorized vehicle overflow occupancy rate data, resulting in equivalent lane borrowing potential energy data that represents the actual physical tendency and feasibility of the non-motorized vehicle group to expand into adjacent lanes after deducting all objective safety risk resistance. Determine whether the equivalent borrowing potential energy data is greater than the preset mapping threshold; If so, the optimal detour boundary coordinate data is generated based on the numerical difference between the equivalent detour potential energy data and the mapping threshold, and a light and shadow reconstruction collaborative control instruction set containing the optimal detour boundary coordinate data is output to the roadside light and shadow projection device. If not, then static right-of-way maintenance instruction data is generated.

Images