Intelligent intersection multi-vehicle cooperative control method

Abstract

The invention relates to the field of intelligent traffic systems, in particular to an intelligent intersection multi-vehicle cooperative control method. The method comprises the following steps: S1, sensing a vehicle state and sharing vehicle node information by adopting a dual-mode communication mechanism; s2, fusing state data at edge computing nodes; establishing a dynamic traffic model according to the fused state data; s3, acquiring node data of the edge calculation node, calculating a vehicle passing priority index according to the node data, and performing vehicle priority decision according to the passing priority index; s4, performing grading processing on the vehicles with the overlapped tracks; s5, the RSU device broadcasts the control instruction, and if the vehicle-mounted unit receives the control instruction within the set time, an execution report is transmitted back to the RSU device; and if the vehicle-mounted unit does not receive the control instruction within the set time, a redundancy execution mechanism is started to detect and control the running state of the vehicle. According to the invention, millisecond-level synchronous transmission and high-precision perception of information are realized, and the real-time performance and accuracy of traffic environment modeling are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of intelligent transportation systems, in particular to a kind of intelligent intersection multi-vehicle coordination control method. BACKGROUND

[0002] The existing intersection control technology has systematic bottleneck, and the core is reflected in the four dimensions of communication, decision, architecture and security.Communication level, mainstream DSRC / LTE-V2X scheme is limited by bandwidth and high delay, cannot transmit high-precision data such as laser radar point cloud, leading to inaccurate real-time control;The decision mechanism relies on fixed weight distribution, lacks dynamic traffic flow adaptation ability, conflict detection accuracy is less than 90%, and a multi-factor coordinated traffic right model is not established;Architecture design, cloud centralized processing introduces more than 200ms delay, vehicle autonomous decision has blind area perception defects, and the lack of dynamic digital twin modeling further weakens the global situation awareness ability;Security redundancy mechanism is particularly weak, the degradation strategy is single when communication is interrupted, and the time-space conflict resolution scheme does not cover complex scenarios, making it difficult to respond to sudden risks.A smart intersection multi-vehicle coordination control method is needed to solve the problems of high multi-source data fusion delay, rigid decision mechanism, insufficient conflict resolution in complex scenarios and system efficiency bottleneck in existing intersection control. SUMMARY

[0003] In view of the problems existing in the prior art, the first aspect of the present application provides a kind of intelligent intersection multi-vehicle coordination control method, specific method includes: S1, adopt dual-mode communication mechanism to perceive vehicle state and share vehicle node information;Wherein, the dual-mode communication mechanism is used to transmit state data between vehicle and vehicle, vehicle and environment; S2, fuse the state data between vehicle and vehicle, vehicle and environment in edge computing node;According to the fused state data, a dynamic traffic model is established; S3, obtain the node data of the edge computing node, calculate the traffic priority index of the vehicle located at the intelligent intersection according to the node data, and make vehicle priority decision according to the traffic priority index; S4, after the priority decision of vehicle, the driving trajectory of each vehicle is predicted, and the vehicle with trajectory overlap is executed hierarchical processing; S5, the edge computing node generates control instruction, and transmits the control instruction to RSU multi-sensor device, the RSU device broadcasts the control instruction, if the vehicle-mounted unit receives the control instruction within the specified time, then to RSU device back transmission execution report;If the vehicle-mounted unit does not receive the control instruction within the specified time, then start redundant execution mechanism detection and control vehicle running state.

[0004] Optionally, in the dual-mode communication mechanism, the main link between the vehicle and the environment adopts NR-V2X Uu for state data communication; the standby link between the vehicles adopts NR-V2X PC5 for state data communication.

[0005] Optionally, the edge computing node fuses the state data between the vehicles and the environment by using an improved iterative closest point algorithm; wherein the state data between the vehicle and the environment is the full state data collected by the RSU multi-sensor device.

[0006] Optionally, the fusion process of the improved iterative closest point algorithm includes: converting the vehicle coordinate system into the coordinate system of the RSU multi-sensor; performing motion compensation on the RSU multi-sensor; and accurately aligning the vehicle state data and the RSU multi-sensor state data.

[0007] Optionally, a space-time grid index is constructed according to the node data of the edge computing node, and the specific construction process includes: dividing the intelligent intersection into a two-dimensional unit grid, and performing spatial discretization processing on the two-dimensional unit grid; performing time discretization processing on the time axis; generating a space-time grid by combining the two-dimensional unit grid subjected to the spatial discretization processing and the time axis subjected to the time discretization processing; predicting a future trajectory of a vehicle, and mapping a trajectory point on the future trajectory to the space-time grid; establishing an index optimization mechanism of the space-time grid, and updating the index process in real time.

[0008] Optionally, the conflict area of the intelligent intersection is detected according to the space-time grid index, and the specific detection process includes: traversing the space-time grid and retrieving a vehicle object list; determining a conflict type; evaluating a collision risk; clustering a same group of vehicle conflict events; and adjusting a conflict threshold according to a current traffic scene.

[0009] Optionally, the vehicle priority decision is calculated based on a multi-factor weighted formula; wherein the calculation steps of the multi-factor weighted formula include: basic factor calculation, LSTM dynamic weight adjustment, and calculation of the passing priority index.

[0010] Optionally, the basic factor calculation process includes: time factor calculation, direction weight calculation, and priority factor calculation; the LSTM dynamic weight adjustment process includes: input features, network structure construction, time sequence feature processing, and establishment of an online learning mechanism; and the passing priority index calculation process includes: multi-factor integration calculation, and priority sorting and grouping.

[0011] Optionally, the hierarchical processing steps performed on the vehicles with trajectory overlap include: conflict classification processing and elastic time window generation.

[0012] Optionally, the redundancy execution mechanism includes a normal execution mode and a degraded execution path mode.

[0013] Compared with the prior art, the application has the beneficial effects that: in the application, a dual-mode communication mechanism is adopted to perceive vehicle states and share vehicle node information, and through NR-V2X hybrid communication and multi-source data fusion technology, millisecond-level synchronous transmission and high-precision perception of vehicle state information are realized, so that the real-time performance and accuracy of traffic environment modeling are significantly improved. The edge computing node fuses state data and establishes a dynamic traffic model to avoid delay and interruption in the data transmission process, thereby improving the data transmission efficiency. Vehicle priority decision is made according to the traffic priority index, and the decision flexibility of multiple vehicles running at the intersection position is improved by establishing a reasonable decision scheduling. The vehicles with trajectory overlap are subjected to hierarchical processing to reduce the probability of accidents at the intersection. The redundancy execution mechanism is started to solve the vehicle running conflict problem in a complex scenario, thereby further improving the safety of intelligent vehicle driving. BRIEF DESCRIPTION OF DRAWINGS

[0014] The drawings constituting a part of the specification of the application are used to provide a further understanding of the application, and the illustrative embodiments of the application and the description thereof are used to explain the application and do not constitute an improper limitation on the application. In the drawings: Figure 1 A flowchart of a kind of intelligent intersection multi-vehicle cooperative control method of the application; Figure 2 A point cloud registration optimization registration process flowchart in a kind of intelligent intersection multi-vehicle cooperative control method of the application; Figure 3 A system architecture schematic diagram in a kind of intelligent intersection multi-vehicle cooperative control method of the application; Figure 4 A dynamic traffic modeling flowchart in a kind of intelligent intersection multi-vehicle cooperative control method of the application; Figure 5 A priority decision and conflict resolution schematic diagram in a kind of intelligent intersection multi-vehicle cooperative control method of the application; Figure 6 A redundancy execution mechanism state diagram in a kind of intelligent intersection multi-vehicle cooperative control method of the application. DETAILED DESCRIPTION

[0015] The application will be described in detail below with reference to the drawings and in combination with embodiments. It should be noted that the embodiments in the application and the features in the embodiments can be combined with each other without conflict.

[0016] The following detailed description is exemplary in nature and is intended to provide further description of the application. Unless otherwise defined, all technical terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in the description of the application herein is for describing particular embodiments only and is not intended to be limiting of the example embodiments in accordance with the application. The first aspect of the embodiment provides a method for intelligent intersection multi-vehicle cooperative control. Referring to Figure 1 and Figure 3 , the specific method comprises: S1, using a dual-mode communication mechanism to perceive vehicle state and share vehicle node information; wherein the dual-mode communication mechanism is used to transmit state data between vehicles and between vehicles and the environment; S2, fusing the state data between vehicles and between vehicles and the environment at the edge computing node; and establishing a dynamic traffic model according to the fused state data; S3, obtaining node data of the edge computing node, calculating a passing priority index of the vehicle located at the intelligent intersection according to the node data, and making a vehicle priority decision according to the passing priority index; S4, after making a priority decision for the vehicle, predicting the driving trajectory of each vehicle, and performing hierarchical processing on vehicles with overlapping trajectories; S5, the edge computing node generates a control instruction and transmits it to the RSU multi-sensor device, the RSU device broadcasts the control instruction, if the vehicle-mounted unit receives the control instruction within the specified time, it returns an execution report to the RSU device; if the vehicle-mounted unit does not receive the control instruction within the specified time, a redundant execution mechanism is started to detect and control the vehicle operating state.

[0017] In some embodiments, in the dual-mode communication mechanism, the communication between the vehicle and the environment is the main link, and NR-V2X Uu is used for state data communication; the communication between the vehicles is the backup link, and NR-V2X PC5 is used for state data communication.

[0018] Specifically, in this embodiment, the cooperation of the roadside unit RSU multi-sensor device and the on-board unit OBU is relied on. The RSU multi-sensor device is deployed on the top of the intersection signal lamp pole, with an installation height of 6 meters, and is equipped with an NR-V2X Uu communication interface. The on-board unit OBU is integrated into the front-mounted system of the autonomous vehicle, which contains an NR-V2X chip set and an RTK-GNSS positioning module. When the vehicle enters the range of 50 meters from the intersection, the system is initialized through the following process: Firstly, a communication link needs to be established. The RSU multi-sensor device broadcasts a beacon signal at a frequency of 5 Hz, mainly containing a location ID and a communication protocol version. After detecting the beacon signal, the on-board unit OBU sends an access request through the NR-V2X Uu interface. The RSU multi-sensor device allocates temporary communication resources, including time slots and frequency bands, and establishes a dedicated data transmission channel.

[0019] Secondly, state data collection and transmission are required. The vehicle broadcasts a structured data packet 10 times per second, mainly including vehicle ID, high-precision position information, speed and acceleration, target path, estimated time of arrival ETA, priority factor, and vehicle sensor summary.

[0020] Finally, the NR-V2X dual-mode communication mechanism needs to be ensured. The NR-V2X Uu interface is mainly used for communication between vehicles and RSU multi-sensor devices for transmitting full-state data. The NR-V2X PC5 interface is mainly used for direct vehicle-to-vehicle communication. In addition to enabling communication between vehicles, vehicles within the intersection range that receive RSU multi-sensor device broadcast information can also share information with vehicles outside the intersection coverage range through direct vehicle-to-vehicle communication, effectively implementing multi-vehicle cooperative control at the intersection and reducing the probability of traffic accidents.

[0021] In some embodiments, referring to Figure 4 , the edge computing node fuses the state data between the vehicles and the environment using an improved iterative closest point algorithm. The state data between the vehicles and the environment is the full-state data collected by the RSU multi-sensor device.

[0022] Specifically, the core task of the present embodiment is to fuse the vehicle-reported data and the RSU multi-sensor data to construct a real-time digital twin intersection model. Key data sources are collected synchronously at a period of 20 ms. The first type is dynamic data uploaded by vehicles. The RSU multi-sensor receives the structured state packet broadcast by the autonomous vehicle through the NR-V2X Uu interface and forwards it to the edge node, including vehicle ID, high-precision position (WGS84 coordinate system), real-time speed / acceleration, target path (straight / left turn / right turn), estimated time of arrival ETA, priority factor, and sensor summary. The second type is RSU multi-sensor data, including 10 Hz point cloud data output by the laser radar, with the coordinate system being the RSU multi-sensor local coordinate system, and video stream collected by the camera.

[0023] In some embodiments, the fusion process of the improved iterative closest point algorithm includes: converting the vehicle coordinate system to the RSU multi-sensor coordinate system; performing motion compensation on the RSU multi-sensor; and accurately aligning the vehicle state data and the RSU multi-sensor state data.

[0024] Before the fusion of multi-source perception data based on RSU multi-sensor, the WGS84 coordinate system position reported by the vehicle needs to be converted into the local coordinate system of the RSU multi-sensor. The specific method is to use the rotation matrix R and the translation vector T to perform coordinate transformation on the latitude and longitude difference to obtain the two-dimensional position in the RSU multi-sensor coordinate system , and the specific calculation formula is: wherein the rotation matrix R and the translation vector T can be obtained by installation position calibration of the RSU. After completing the coordinate system conversion, motion compensation is performed on the laser radar point cloud data to eliminate the distortion caused by the change of the vehicle position, laying a foundation for the subsequent precise alignment of multi-source data.

[0025] After completing the coordinate system processing, in order to realize the precise alignment of vehicle perception data and RSU multi-source sensor data, an improved ICP algorithm is used for point cloud registration optimization. The optimization target is: wherein, represents the point position reported by the vehicle, represents the corresponding point detected by the laser radar, is the weight based on dynamic calculation of point cloud density.

[0026] Here, referring to Figure 2 , the point cloud registration optimization registration process specifically includes the following steps: S20, quickly establish the nearest point correspondence based on KD-Tree; S21, calculate the rigid transformation matrix by singular value decomposition (SVD); S22, apply the transformation and update the registration error; S23, repeat iteration until the error is less than 0.1m or the iteration number reaches the upper limit of 20 times.

[0027] To ensure the consistency of multi-source perception data in the time dimension, first, all data are time-stamped based on the PTP (Precision Time Protocol) protocol, thereby reducing the synchronization error caused by clock deviation. For moving targets, a linear interpolation method is used for time compensation, and the calculation formula is: wherein, represents the position, v represents the speed, represents the time difference.

[0028] Through the weighted ICP motion compensation technology, the coordinate system difference problem between the vehicle and the roadside perception data is solved, laying a foundation for high-precision dynamic modeling.

[0029] In some embodiments, the spatio-temporal grid index is constructed according to the node data of the edge computing node, and the specific construction process includes: dividing the intelligent intersection into a two-dimensional unit grid, and performing spatial discretization processing on the two-dimensional unit grid; performing time discretization processing on the time axis; generating a spatio-temporal grid by combining the two-dimensional unit grid subjected to the spatial discretization processing and the time axis subjected to the time discretization processing; predicting the future trajectory of the vehicle, and mapping the trajectory points on the future trajectory to the spatio-temporal grid; establishing a spatio-temporal grid index optimization mechanism, and updating the index process in real time.

[0030] In this embodiment, the step of spatial discretization processing includes dividing the intersection area into two-dimensional grid units of 0.5 m*0.5 m to form a spatial matrix structure. The direction dimension is divided into 24 direction sectors at an interval of 15°, and a spatial hash function is established: , wherein, represents the floor function, is a two-dimensional plane coordinate, is a heading angle, in which the first term is discretized in the X direction at an interval of 0.5 m, the second term is discretized in the Y direction at an interval of 0.5 m, and the third term is discretized in the direction at an interval of 15°. The function can map a three-dimensional discrete space into a unique hash value, and realize fast query of complexity.

[0031] In this embodiment, the step of time discretization processing includes dividing the time axis of the next 5 seconds into time slots at an interval of 50 ms (0.05 s), and a total of 100 time units. A time index function is established: , wherein t is a predicted time point (range 0-5 s), each time slot corresponds to a unique integer index, thereby realizing discretization representation of the time dimension.

[0032] In this embodiment, a trajectory prediction and grid mapping process is established. For each vehicle, based on the current position, speed, acceleration, and driving intention (straight, left turn, or right turn), a three-order Bezier curve model is used to predict the future trajectory: , wherein, is the current position of the vehicle, is determined according to the driving intention of the vehicle, is the target position. In the prediction process, the trajectory points are calculated at a step of 50 ms, and the trajectory points are mapped to the corresponding spatio-temporal grid unit by using the spatial hash function and the time index function, and the vehicle ID is inserted into the object list of the unit.

[0033] ​In this embodiment, an index optimization mechanism is adopted, specifically including adopting a dynamic memory allocation strategy, activating only the grid cells where vehicles exist, reducing memory occupancy by 70%; implementing a time sliding window mechanism, removing outdated data every 50 ms The part that exceeds the range); for hash conflicts, the chain address method is adopted for processing, and a linked list is established in the conflict cell to store multiple objects. A two-level spatial index (R-tree) is established to speed up the range query operation and support fast retrieval of all vehicles in the specified spatio-temporal region.

[0034] Then a real-time update process is adopted, which is executed once every 50 ms, including receiving the latest vehicle state data, predicting future trajectories, updating the grid mapping, removing data of vehicles that have left, and optimizing the index structure. The update process adopts multi-thread parallel processing, vehicle-level prediction tasks are distributed for execution, and grid update tasks are processed centrally, thereby ensuring that the index update of 50+ vehicles is completed within 15 ms. The algorithm realizes efficient spatio-temporal relationship management by four-dimensional grid index , reducing the conflict detection complexity from to .

[0035] In some embodiments, the conflict area detection of intelligent intersections is performed according to the spatio-temporal grid index, and the specific detection process includes: traversing the spatio-temporal grid and retrieving the vehicle object list; determining the conflict type; evaluating the collision risk; clustering the same group of vehicle conflict events; and adjusting the conflict threshold according to the current traffic scene.

[0036] Specifically, the conflict area detection of intelligent intersections in the spatio-temporal grid index accurately predicts and processes intersection conflicts, merging conflicts, and splitting conflicts, and combines an elastic time window mechanism, so that the conflict avoidance rate in complex scenarios is greatly improved, effectively solving the problem of high blind collision risk in the prior art.

[0037] In this embodiment, the spatio-temporal grid is traversed and the vehicle object list is retrieved, and the specific process includes: taking 50 ms as the time step, the system traverses all cells of the spatio-temporal grid, each cell contains a 0.5m×0.5m space range and a 50ms time window. For each spatio-temporal cell, retrieve the list of vehicle objects contained therein. When detecting that there are ≥2 vehicles in a single cell, trigger the conflict analysis process. The system prioritizes high severity areas (such as intersection center areas) and optimizes the traversal order using a breadth-first search strategy.

[0038] In this embodiment, the specific process of determining the conflict type includes conflict classification based on vehicle motion characteristics, calculation of the included angle of the driving directions of the two vehicles, determination of intersection conflict when the included angle is greater than 90°, determination of merging conflict when the included angle is less than 30° and the relative distance is less than 10 meters, detection of the overlapping area of the vehicle lane-changing trajectory in the space-time grid, determination of diverging conflict when the lane-changing sign bit is activated, the trajectory overlap duration is greater than 200 ms, and the lateral distance is less than 1.5 m.

[0039] In this embodiment, the specific process of evaluating the collision risk includes calculating a risk indicator for each potential conflict, including the time to collision (TTC): wherein, is the relative position vector, is the relative velocity vector.

[0040] Conflict severity:

[0041] wherein, the weight coefficient ; According to the vehicle type, the value of a sedan is 0.5, the value of a passenger car is 1.0, and the value of a truck is 1.5.

[0042] Risk level: When , it is determined as low risk; When , it is determined as medium risk; When , it is determined as high risk.

[0043] In this embodiment, the same group of vehicles in the conflict event is aggregated. For the same group of vehicles detected in consecutive space-time units, i.e., adjacent space grids and consecutive time slots, the conflict events are aggregated. The merging rule is that if the time interval of two conflict events is less than 100 ms and the spatial position is adjacent, then they are merged into one conflict event, and the severity is taken as the maximum value in the merged event.

[0044] Severity integration : wherein, n is the number of sub-conflicts included in the aggregated event.

[0045] Dynamic threshold adjustment. The conflict filtering threshold is dynamically adjusted according to the current traffic scene. The severity threshold is lowered when the vehicle density is high to ensure that no conflicts are missed, and the threshold is increased in low-density scenarios to reduce false positives. Finally, only the conflict events with severity exceeding the threshold are output.

[0046] In some embodiments, reference is made to Figure 5The vehicle priority decision is based on a multi-factor weighted formula calculation; wherein the calculation steps of the multi-factor weighted formula include: basic factor calculation, LSTM dynamic weight adjustment and passing priority index calculation.

[0047] Specifically, by adopting LSTM dynamic weight adjustment to adaptively optimize the passing right allocation, the limitations of the traditional fixed weight mechanism are broken through, and therefore the application makes effective progress in key indicators such as peak period passing efficiency and special vehicle priority guarantee.

[0048] In some embodiments, the basic factor calculation process includes: time factor calculation, direction weight calculation and priority factor calculation; the LSTM dynamic weight adjustment process includes: input features, network structure construction, time sequence feature processing and online learning mechanism establishment; the passing priority index calculation process includes: multi-factor integration calculation and priority sorting and grouping.

[0049] In this embodiment, the specific calculation process of the basic factor calculation includes: Time factor calculation.

[0050] Time factor Based on the estimated arrival time of the vehicle , the principle of "the earlier the arrival, the higher the priority" is embodied. When , the vehicle is considered to have entered the conflict area, and the system upper limit value = 10.0 is forcibly set; For normal value, the ; is calculated and normalized: ; wherein and are dynamic adjustment parameters, which are updated every 5 minutes according to historical traffic data to ensure that the factor distribution is close to the normal characteristic. The factor calculation is executed every 100 ms, which is synchronized with the vehicle state update.

[0051] Direction weight calculation. The direction weight is determined according to the vehicle driving intention, reflecting the priority principle of "straight> right turn> left turn".

[0052] The system presets the basic weight: straight = 1.0, right turn = 0.7, left turn = 0.5; Dynamic adjustment in special scenarios: When the left turn traffic density > 20 vehicles / minute, congestion compensation is started: ; When there is a special right turn lane at the intersection: ; The adjustment is based on real-time traffic flow analysis, using a sliding window (30 seconds) for statistics, with a response delay control within 200 ms. The maximum weight is output by the direction decision module, with a confidence score (0 to 1 range).

[0053] Priority factor calculation. Priority factor Integrate vehicle type and passenger status calculation: Ambulance / fire truck: ; Bus: ; Ordinary passenger car: ; Cargo vehicle: ; Passenger rate is obtained by on-board sensors; when data is missing, the default value is 0.7 for buses and 0.5 for ordinary passenger cars.

[0054] All calculation results are subject to boundary constraints: .

[0055] In this embodiment, the specific calculation process of LSTM dynamic weight adjustment includes: Input features. The system constructs a 10-dimensional feature vector as the input of LSTM, including real-time traffic status and historical patterns: Traffic density (vehicles / km2), average speed (km / h), straight / left / right vehicle ratio, emergency vehicle proportion, weather index (0 to 1, 0 for sunny), visibility (m), number of conflicts in the past 5 minutes, historical same period traffic efficiency, road slip coefficient (0-1), time band (early / late peak marker).

[0056] All features are standardized: where, and Based on past 7-day data dynamic calculation.

[0057] Network structure. Adopt a double-layer LSTM network architecture: 32 neurons in the first layer, 32 neurons in the second layer, and 3 neurons in the fully connected output layer. The specific processing flow of the network structure is: The input layer receives the standardized feature vector; The first LSTM layer calculates the hidden state: ; The second LSTM layer receives Output ; The fully connected layer applies Softmax activation: ; And meet the constraint condition: ; The network performs inference every 100 ms, with a processing delay of less than 5 ms.

[0058] Temporal feature processing. The algorithm uses a sliding time window mechanism to handle temporal dependencies, with a window length of 10 seconds and a feature vector sampled every second; the data of the last 10 time steps (t-9 to t) are retained.

[0059] Apply weighted processing to historical data: where k is the time reverse index, and the recent feature influence is enhanced. The LSTM hidden state is transmitted across cycles to achieve long-term memory function, and the memory decay factor is set to 0.85.

[0060] Online learning mechanism. The system triggers model updates every 10 minutes, including data collection: accumulate 1000 feature-delay records; loss calculation: Backpropagation: use Adam optimizer, learning rate is 0.001; weight clipping: gradient norm is limited within 1.0; Model verification: reserve 20% data for cross-validation, stop updating when the test set loss is greater than 120% of the training set.

[0061] In this embodiment, the calculation of the traffic priority index mainly includes multi-factor integration calculation and priority sorting and grouping.

[0062] Multi-factor integration calculation. The system uses a weighted fusion model to calculate the traffic priority index: where, is the dynamically generated weight of LSTM (satisfies ), is the AI adjustment factor (range [-0.1, 0.1]), used for special scene compensation.

[0063] Priority sorting and grouping. After calculation, intelligent sorting is performed, including sorting the value according to complexity; fine-tuning within the same direction vehicle group: straight-ahead group is sorted by ETA, emergency vehicle group is forced to the top, and freight vehicle group is placed at the bottom; smoothing processing: where, is the exponential moving average of the last 5 indices.

[0064] In some embodiments, the hierarchical processing step for vehicles with trajectory overlap includes conflict classification processing and elastic time window generation.

[0065] ​​In this embodiment, the conflict classification processing step specifically includes: when it is detected that the driving paths of two vehicles form an included angle greater than 90°, the intersection conflict is determined, and the core solution strategy is the time-space right allocation, and the specific calculation steps include: Calculate the priority difference: ; Determine the time adjustment amount: ( ); Adjust the speed of the low-priority vehicle: Here, represents the distance to the conflict point.

[0066] Special case cooperative travel mode: when , a time-space intersection point is generated to guide the synchronous travel of both vehicles.

[0067] The system monitors the adjustment effect in real time, and if the speed deviation exceeds 15%, a secondary optimization is triggered. For vehicles that are in the same direction and are about to merge, if the included angle is less than 30° and the distance is less than 10 meters, a safe time-distance control strategy is adopted, and the specific calculation steps include: Calculate the basic safety time distance: ; Where, is 5m, is 2m.

[0068] According to the dynamic compensation of weather, specifically including: Increase in rainy weather; increase in foggy weather.

[0069] Control the speed of the following vehicle:

[0070] When it is detected that the lane change trajectories of vehicles overlap, the split conflict processing is triggered, and the specific calculation steps include: Calculate the lane allocation probability: ; select the lane as the recommended lane; The key strategies for sending lane change instructions 200 meters in advance include: buffer zone dynamic management, lane change success rate prediction model, and alternative path generation. Among them, the buffer zone dynamic management can accommodate a maximum of 5 vehicles; the lane change success rate prediction model is calculated based on historical data, and the alternative path generation generates a maximum of 3 alternative paths.

[0071] In this embodiment, the specific steps of generating an elastic time window include calculating the time window basic parameters based on the dynamic attributes of the vehicle, including the start time, the basic window width, and the priority expansion amount.

[0072] Determine the start time: ; wherein, MinResponseTime represents the minimum response time, default is 0.5s, ETA is the estimated time of arrival of the vehicle.

[0073] BaseWindowWidth: ; wherein, density unit is vehicle per square kilometer, lower limit is 1.5s, upper limit is 3.0s.

[0074] PriorityExtension: (k=0.5); wherein, is the passing priority index.

[0075] Secondly, the window parameters are dynamically optimized based on real-time environmental factors, mainly including weather compensation and road condition adjustment.

[0076] Weather compensation includes that when the weather condition is sunny, the compensation coefficient is 1.0, and the window adjustment amount is 0; when the weather condition is light rain, the compensation coefficient is 1.1, and the window adjustment amount is +10%; when the weather condition is heavy rain, the compensation coefficient is 1.3, and the window adjustment amount is +30%; when the weather condition is thick fog, the compensation coefficient is 1.5, and the window adjustment amount is +50%.

[0077] Road condition adjustment includes, ; wherein, the friction coefficient ranges from 0.2 (ice surface) to 0.8 (dry asphalt road).

[0078] Final time window calculation: .

[0079] In some embodiments, referring to Figure 6 , the redundant execution mechanism includes a normal execution mode and a degraded execution path mode, wherein, In this embodiment, the core of the normal execution mode is the closed loop of "instruction generation-broadcasting-execution-feedback".

[0080] The edge computing node generates structured instructions based on the conflict resolution solution, and the core fields include: Space-time window: the passing time window accurate to milliseconds , wherein: ; ; Speed instruction: recommended speed value and allowed tolerance , slope section additional acceleration compensation.

[0081] The RSU multi-sensor device issues instructions through dual-mode communication concurrently, wherein the main link is NR-V2X Uu, and the broadcast period is 100ms, and NR-V2X PC5 is used as a backup link to realize message sharing between vehicles.

[0082] When the vehicle-mounted unit receives the instruction, the control instruction generated by the edge computing node and issued via the RSU multi-sensor device is executed. Finally, the vehicle sends an execution report to the RSU multi-sensor device every 100 ms, so that the edge node can perceive abnormalities in a timely manner, and the actual speed / position in the feedback content is used to correct the prediction model.

[0083] In this embodiment, the core of the degraded execution path mode is that the system monitors the communication state and risk indicators in real time. If no instruction is received within 200 ms, the first level of degradation is preferred to start V2V direct connection negotiation. If the V2V direct connection negotiation fails, the second level of degradation is activated to activate local perception decision. When the system detects that the TTC is less than 1.5 seconds, the third level of degradation is started to trigger AEB emergency braking regardless of the communication state, forming a three-level redundant safety mechanism.

[0084] Specifically, based on the three-level redundant execution architecture, the application can still maintain safe operation under abnormal conditions such as communication interruption. Since the architecture combines the global vision of vehicle-road cooperation and the advantages of single vehicle autonomous decision, the reliability of the system in extreme scenarios such as bad weather and mixed traffic flow has been significantly improved. At the same time, the single intersection investment return period is compressed to 1 / 3 of the traditional scheme, providing a cooperative control paradigm for intelligent transportation systems with high performance and economy.

[0085] Here, the V2V direct connection negotiation is that the vehicle broadcasts real-time state and passage request through the NR-V2X PC5 interface. At this time, the neighboring vehicles calculate the avoidance strategy based on the interaction information in real time, and both parties execute the agreed action within 10 ms to realize conflict resolution without central coordination. The real-time state includes position, speed, steering intention, etc. The local perception decision is that when the coordination fails, the vehicle activates multi-sensor fusion for environment perception and executes a conservative passage strategy, limiting the maximum vehicle speed to 50% of the original speed and not more than 30 km / h, while dynamically maintaining a safe distance, the calculation formula is ; wherein, is the current vehicle speed. The AEB emergency braking is that when the predicted TTC is lower than the 1.5 second critical value, the system immediately triggers automatic emergency braking to continuously brake at a deceleration of not less than 4 m / s² until the vehicle completely stops, giving priority to personal safety.

[0086] From the common general knowledge, the application can be realized by other embodiments without departing from the spirit or essential characteristics thereof. Therefore, the above disclosed embodiments are merely illustrative in all aspects and are not the only ones. All changes within the scope of the application or within the scope equivalent to the application are included in the application.

[0087] Those skilled in the art will appreciate that embodiments of the application can be devised for a method, a system, or a computer program product. Accordingly, the present application can be embodied in the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application can take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage devices, etc.) embodying computer readable program code.

[0088] The present application is described in reference to the flowchart and / or block diagrams of the method, apparatus (system) and computer program product according to embodiments of the application. It will be understood that each block of the flowchart and / or block diagrams, and combinations of blocks in the flowchart and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general purpose computer, special purpose computer, embedded processing device or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks.

[0089] These computer program instructions can also be stored in a computer- readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the function specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks.

[0090] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks.

[0091] Finally, it should be noted that the above-mentioned embodiments are merely intended for describing the technical solutions of the present application, but not for limiting it. Although the present application is described in detail with reference to the above embodiments, those skilled in the field should understand that the specific embodiments of the present application can be modified or replaced equivalently without departing from the spirit and scope of the present application, and any modification or equivalent replacement without departing from the spirit and scope of the present application should be covered in the protection scope of the claims of the present application.

Claims

1. A method for multi-vehicle cooperative control at an intelligent intersection, characterized in that, Specific methods include: S1. A dual-mode communication mechanism is used to sense vehicle status and share vehicle node information; wherein, the dual-mode communication mechanism is used to transmit status data between vehicles and between vehicles and the environment. S2. At the edge computing node, fuse the state data between vehicles and between vehicles and the environment; establish a dynamic traffic model based on the fused state data; S3. Obtain the node data of the edge computing node, calculate the traffic priority index of the vehicle located at the smart intersection based on the node data, and make a vehicle priority decision based on the traffic priority index. S4. After making the priority decision for the vehicles, predict the driving trajectory of each vehicle, and perform hierarchical processing on vehicles with overlapping trajectories. S5. The edge computing node generates control commands and transmits the control commands to the RSU multi-sensor device. The RSU device broadcasts the control commands. If the vehicle unit receives the control commands within a specified time, it sends an execution report back to the RSU device. If the vehicle unit does not receive the control commands within a specified time, it activates a redundant execution mechanism to detect and control the vehicle's operating status.

2. The intelligent intersection multi-vehicle cooperative control method according to claim 1, characterized in that, In the dual-mode communication mechanism, the primary link between the vehicle and the environment uses NR-V2X Uu for status data communication; the backup link between vehicles uses NR-V2X PC5 for status data communication.

3. The intelligent intersection multi-vehicle cooperative control method according to claim 1, characterized in that, The edge computing node uses an improved iterative nearest point algorithm to fuse the state data between vehicles and between vehicles and the environment; wherein, the state data between vehicles and the environment is the full state data collected by the RSU multi-sensor device.

4. The intelligent intersection multi-vehicle cooperative control method according to claim 3, characterized in that, The fusion process of the improved iterative nearest point algorithm includes: converting the vehicle coordinate system into the coordinate system of the RSU multi-sensor; performing motion compensation on the RSU multi-sensor; and accurately aligning the vehicle state data with the RSU multi-sensor state data.

5. The intelligent intersection multi-vehicle cooperative control method according to claim 3, characterized in that, A spatiotemporal grid index is constructed based on the node data of the edge computing nodes. The specific construction process includes: dividing the intelligent intersection into two-dimensional cell grids and performing spatial discretization processing on the two-dimensional cell grids; performing temporal discretization processing on the time axis; combining the spatially discretized two-dimensional cell grids and the temporally discretized time axis to generate a spatiotemporal grid; predicting the future trajectory of the vehicle and mapping the trajectory points on the future trajectory onto the spatiotemporal grid; establishing the spatiotemporal grid index optimization mechanism and updating the index process in real time.

6. The intelligent intersection multi-vehicle cooperative control method according to claim 5, characterized in that, The collision area detection of the intelligent intersection is performed based on the spatiotemporal grid index. The specific detection process includes: traversing the spatiotemporal grid and retrieving the vehicle object list; determining the collision type; assessing the collision risk; aggregating vehicle collision events; and adjusting the collision threshold according to the current traffic scenario.

7. The intelligent intersection multi-vehicle cooperative control method according to claim 1, characterized in that, The vehicle priority decision is calculated based on a multi-factor weighted formula; wherein the calculation steps of the multi-factor weighted formula include: basic factor calculation, LSTM dynamic weight adjustment and traffic priority index calculation.

8. The intelligent intersection multi-vehicle cooperative control method according to claim 7, characterized in that, The basic factor calculation process includes: time factor calculation, directional weight calculation, and priority factor calculation; the LSTM dynamic weight adjustment process includes: input features, network structure construction, temporal feature processing, and establishment of an online learning mechanism; the passage priority index calculation process includes: multi-factor integrated calculation and priority ranking and grouping.

9. The intelligent intersection multi-vehicle cooperative control method according to claim 1, characterized in that, The hierarchical processing steps for vehicles with overlapping trajectories include: conflict classification processing and flexible time window generation.

10. The intelligent intersection multi-vehicle cooperative control method according to claim 1, characterized in that, The redundant execution mechanism includes a normal execution mode and a degraded execution path mode.

Images