Vehicle cooperative avoidance system, method, device and medium based on dynamic no-entry zone

Abstract

The application discloses a vehicle cooperative avoidance system and method based on a dynamic no-entry zone, equipment and a medium, and particularly relates to the field of intelligent driving, and comprises a road space-time data access module, a target intention and trajectory cluster analysis module, a dynamic no-entry zone construction module, a cooperative conflict identification module, a cooperative avoidance decision module, a feasible corridor planning module and a vehicle motion control module. The system converts the ego vehicle state, external target state, roadside cooperative information and communication delay information into a road space-time data frame, generates a target trajectory cluster, and encodes entity occupation, shadowing and cooperative reserved space as dynamic no-entry zone objects. The cooperative avoidance group is determined by detecting the space-time crossing relationship between the ego vehicle candidate action and the dynamic no-entry zone objects, vehicle trajectories are generated, and control instructions are output. The application can uniformly express multi-source risk constraints, so that the cooperative decision directly acts on the vehicle planning control process.

Description

Technical Field

[0001] This invention relates to the field of intelligent driving technology, and more specifically, to a vehicle cooperative avoidance system, method, device, and medium based on dynamic restricted areas. Background Technology

[0002] Existing intelligent driving vehicle avoidance systems typically operate primarily through single-vehicle perception and planning. During operation, the vehicle uses cameras, millimeter-wave radar, lidar, and positioning devices to acquire position and speed information of lane lines, vehicles ahead, vehicles to the sides, pedestrians, non-motorized vehicles, and temporary obstacles. A perception fusion algorithm then generates a target list and drivable areas. Subsequently, a trajectory prediction algorithm predicts the future trajectories of surrounding targets based on their current position, speed, lane topology, and traffic rules. A planning algorithm generates the vehicle's own trajectory based on the predicted trajectory, lane boundaries, and preset safety distances. A control algorithm then converts this trajectory into steering, braking, and drive control commands. Some existing systems also receive information on the position, speed, acceleration, and driving intentions of adjacent vehicles via vehicle-to-vehicle or vehicle-to-infrastructure communication, using this communication data as supplementary input for target prediction or path planning to assist the vehicle in following other vehicles, changing lanes, detouring, and controlling intersection passage.

[0003] However, existing technologies still have shortcomings in multi-vehicle interaction and avoidance scenarios. First, existing systems typically use predicted obstacle trajectories or fixed safety boundaries directly as the basis for avoidance, making it difficult to uniformly express occlusion areas, communication delays, yielding pressure from adjacent vehicles, and cooperative reserved space as spatiotemporal constraints that can be directly invoked by the planning algorithm. This forces vehicles to switch between multiple rules in complex intersection scenarios. Second, existing cooperative driving solutions mostly use vehicle intentions or future trajectories as communication objects, requiring the planning end to reinterpret the intentions of other vehicles, easily leading to inconsistencies between cooperative results and motion planning constraints. Third, existing systems are insufficient in identifying indirect conflicts between adjacent vehicles. They typically only determine whether the vehicle's trajectory overlaps with the target vehicle's predicted trajectory, but struggle to determine whether the vehicle's candidate actions will force adjacent vehicles to decelerate sharply, laterally avoid, or enter an uncontrollable state, resulting in some conflicts only being discovered during actual execution. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, the present invention provides a vehicle cooperative avoidance system, method, device and medium based on dynamic restricted areas, and solves the problems mentioned in the background art through the following solutions.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a vehicle cooperative avoidance system based on dynamic restricted areas, comprising: The vehicle cooperative avoidance system based on dynamic restricted areas is deployed within the vehicle domain controller and is connected to vehicle sensors, vehicle positioning equipment, V2X communication equipment, steering system, braking system and drive system; The vehicle cooperative avoidance system based on dynamic restricted areas includes a road spatiotemporal data access module, a target intent and trajectory cluster analysis module, a dynamic restricted area construction module, a cooperative conflict identification module, a cooperative avoidance decision module, a feasible corridor planning module, a vehicle motion control module, and an operation verification and parameter calibration module. The road spatiotemporal data access module is used to convert vehicle status, external target status, roadside coordination information and communication delay information into road spatiotemporal data frames under the same control cycle. The target intent and trajectory cluster analysis module is used to generate target trajectory clusters based on the road spatiotemporal data frames; The dynamic restricted area construction module is used to generate dynamic restricted area objects based on the target trajectory cluster, vehicle occupancy prediction, other vehicle occupancy prediction, occupancy area and communication delay status, and write the dynamic restricted area objects into the dynamic restricted area index table; The cooperative conflict identification module is used to detect the spatiotemporal travel relationship between the candidate actions of the vehicle and the dynamic restricted area object, and generate a cooperative avoidance group; The collaborative avoidance decision module is used to determine the main passing vehicle, the decelerating and yielding vehicle, and the lateral yielding vehicle according to the collaborative avoidance group, and convert the determined collaborative results into collaborative reserved dynamic no-entry zone objects; The feasible corridor planning module is used to generate an executable trajectory for vehicles under the constraints of the dynamic restricted area index table; The vehicle motion control module is used to convert the vehicle's executable trajectory into steering commands, braking commands, and driving commands. The operation verification and parameter calibration module is used to calibrate the boundary expansion, activation window, and confidence score boundary of the dynamic restricted area object based on the actual vehicle operation data.

[0006] Preferably, a vehicle cooperative avoidance method based on dynamic restricted areas includes: The system acquires the vehicle status, external target status, roadside cooperation information, and communication delay information, and converts the vehicle status, external target status, roadside cooperation information, and communication delay information into road spatiotemporal data frames under the same control cycle. A target trajectory cluster is generated based on the road spatiotemporal data frame, and a dynamic no-entry zone object is generated based on the target trajectory cluster, vehicle occupancy prediction, other vehicle occupancy prediction, occupancy area, and communication delay status. Write the dynamic restricted area object into the dynamic restricted area index table, and detect the spatiotemporal travel relationship between the candidate actions of the vehicle and the dynamic restricted area object; When a spatiotemporal travel relationship is detected between the candidate action of the autonomous vehicle and the object of the dynamic restricted area, a cooperative avoidance group is generated, and the main passing vehicle, the vehicle that slows down and yields, and the vehicle that yields laterally are determined according to the cooperative avoidance group. Based on the main passing vehicle, the decelerating and yielding vehicle, and the lateral yielding vehicle, a collaborative reserved dynamic restricted area object is generated, and an executable trajectory for the vehicle is generated under the constraints of the dynamic restricted area index table. The vehicle's executable trajectory is converted into steering, braking, and driving commands, and the boundary expansion, activation window, and confidence score boundary of the dynamic restricted area object are calibrated based on the vehicle's actual operating data.

[0007] Preferably, an electronic device includes a processor, a memory, a vehicle communication interface, and a chassis control interface; The memory stores a computer program, which, when executed by the processor, causes the electronic device to perform the vehicle cooperative avoidance method based on dynamic no-entry zones. The vehicle communication interface is used to receive vehicle status, external target status, roadside coordination information, and communication delay information. The chassis control interface is used to output steering commands, braking commands, and drive commands generated based on the vehicle's executable trajectory.

[0008] Preferably, a computer-readable storage medium stores a computer program therein; When the computer program is executed by the processor, it implements the vehicle cooperative avoidance method based on dynamic restricted areas.

[0009] The technical effects and advantages of this invention are as follows: This solution uses a dynamic restricted area object to uniformly write the entity-occupied area, the interactive shadow area caused by occlusion, the uncertain area caused by communication delay, and the collaborative reserved area into a dynamic restricted area index table. This allows the vehicle planning process to directly query the restricted space in three dimensions: road longitudinal coordinates, road lateral coordinates, and time. As a result, the system no longer relies solely on obstacle trajectory points or fixed safety distances for obstacle avoidance. Instead, it transforms the results of multi-source perception, roadside collaboration, communication compensation, and target prediction into a single executable constraint, enabling the feasible corridor planning module to generate executable vehicle trajectories based on a unified data structure. This solution uses a collaborative avoidance decision module to convert the collaborative results of vehicles passing through, vehicles slowing down to yield, and vehicles yielding laterally into collaboratively reserved dynamic restricted areas objects, and writes them back to the dynamic restricted area construction module and the dynamic restricted area index table. As a result, the collaborative intent no longer exists merely as a communication message, but directly becomes a constraint on subsequent path search and speed orchestration, enabling a consistent data interface between collaborative avoidance decisions and vehicle motion planning, and reducing the situation where vehicles have reached a collaborative intent but the planning end has not accurately executed the collaborative intent. This solution performs counterfactual occupancy analysis through an interactive shadow tagging unit. Under three action assumptions—candidate vehicle not yielding, candidate vehicle decelerating suddenly, and candidate vehicle continuing to accelerate—it determines whether adjacent vehicles will enter an unexecutable avoidance state due to the candidate vehicle's action. If such a state exists, an interactive shadow dynamic no-entry zone object is generated. Thus, the system not only determines the direct overlap between the vehicle's trajectory and the target trajectory but also identifies indirect conflicts induced by the vehicle's actions. This allows the cooperative conflict identification module to discover risk areas requiring collaborative handling before the vehicle actually performs an avoidance action and to classify the corresponding vehicle into a cooperative avoidance group. Attached Figure Description

[0010] Figure 1 This is a flowchart of the dynamic restricted area generation process of the present invention.

[0011] Figure 2 This is a flowchart of the collaborative avoidance negotiation process of the present invention.

[0012] Figure 3 This is a flowchart of the closed-loop execution verification process of the present invention. Detailed Implementation

[0013] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0014] refer to Figures 1-3The vehicle cooperative avoidance system based on dynamic restricted zones, as shown, is deployed within the vehicle domain controller and connected to the braking system, steering system, drive system, millimeter-wave radar, camera, lidar, and positioning equipment via the vehicle bus. It connects to adjacent vehicles and roadside units via V2X communication. The system operates according to a fixed control cycle. The road spatiotemporal data access module first acquires target data for the vehicle, other vehicles, the road, and roadside units. The target intent and trajectory cluster analysis module then transforms the target motion data into multiple candidate trajectory clusters. The dynamic restricted zone construction module transforms the trajectory clusters and vehicle occupancy boundaries into dynamic restricted zones with activation windows. The cooperative conflict identification module determines the crossing relationship between the vehicle's candidate behavior and the dynamic restricted zone. The cooperative avoidance decision module determines the avoidance role through intent interaction between vehicles. The feasible corridor planning module generates executable vehicle trajectories outside the dynamic restricted zone. The vehicle motion control module transforms the trajectories into steering, braking, and drive commands. The operation verification and parameter calibration module replays and analyzes actual operational deviations and updates the dynamic restricted zone parameter boundaries.

[0015] The road spatiotemporal data access module is used to form a unified spatiotemporal reference within each control cycle. The module first reads the vehicle's speed, yaw rate, steering angle, brake opening, drive torque, vehicle positioning, and vehicle attitude through the vehicle-side motion acquisition unit. Then, it receives the positions of vehicles, pedestrians, non-motorized vehicles, cones, and temporary obstacles detected by the vehicle-side sensors through the external target fusion unit. It also receives broadcast information from the roadside unit regarding obstructed areas, temporary control areas, and adjacent vehicles through the roadside collaborative access unit. Finally, it performs time compensation on various data types through the communication delay calibration unit to obtain a road spatiotemporal data frame consistent with the current control cycle. The road spatiotemporal data frame output by the road spatiotemporal data access module is directly sent to the target intent and trajectory cluster analysis module and provides a data source identifier for the dynamic restricted area construction module.

[0016] The vehicle-side motion acquisition unit reads the vehicle chassis status via the vehicle's CAN bus or in-vehicle Ethernet. Within each control cycle, the unit establishes a vehicle status record, including the sampling time, vehicle center point coordinates, vehicle orientation, vehicle speed, longitudinal acceleration, lateral acceleration, yaw rate, steering angle, brake opening, drive torque, wheel speed difference, and vehicle gear. The unit uses a circular buffer to store the vehicle status records from the most recent control cycles, sorted by sampling time, allowing the communication delay calibration unit to trace back to the actual generation time of the target data. For missing fields, the unit uses the available value from the previous control cycle and writes a missing flag in the corresponding field. The missing flag is included in the subsequent confidence score calculation; the vehicle status record is not directly deleted.

[0017] The external target fusion unit reads the target bounding boxes from the camera, target points from the millimeter-wave radar, clustered targets from the lidar, and the matching results from the vehicle-side positioning map. The external target fusion unit first transforms the coordinate systems of each sensor to the vehicle's coordinate system, and then transforms the vehicle's coordinate system to the road coordinate system. The road coordinate system uses the longitudinal coordinates and lateral offset coordinates of the road centerline to represent the target position. The external target fusion unit establishes external target records based on the target's dimensions, speed, orientation, and the number of consecutive occurrences. For multi-source detection results of the same target, the external target fusion unit correlates them based on spatial overlap, speed difference, orientation difference, and sensor source credibility markers. Successful correlation generates a fused target record; uncorrelated correlations are retained as unconfirmed target records. Unconfirmed target records do not directly enter the cooperative avoidance decision module, but form short-term low-confidence dynamic no-entry zones in the dynamic no-entry zone construction module.

[0018] The roadside cooperative access unit receives information from the roadside unit regarding road construction areas, accident-occupied lane areas, traffic light phases, obstruction boundaries, intersection conflict zones, and roadside sensing targets. The roadside cooperative access unit performs source authentication and time checks on roadside messages. Messages failing source authentication are discarded, and messages whose time lag exceeds the communication delay calibration unit's set window are moved to the history queue and do not participate in the current control cycle. The roadside cooperative access unit converts roadside coordinates to the road coordinate system and writes a roadside source identifier into the road spatiotemporal data frame, enabling the dynamic restricted area construction module to distinguish between dynamic restricted areas formed by vehicle-end sensing and those formed through roadside cooperation.

[0019] The communication delay calibration unit performs delay compensation on vehicle-to-vehicle and vehicle-to-infrastructure communication data. The unit reads the message generation time, message arrival time, and the vehicle's control cycle time, calculates the message arrival lag, and extrapolates the states of adjacent vehicles to the current control cycle based on the speed, acceleration, and steering intention carried in the message. During extrapolation, the unit does not directly rely on the future trajectory provided by the communicating vehicle; instead, it first calculates the motion residual based on the vehicle's two most recent broadcast states. As the motion residual continuously increases, the state confidence score of the communicating vehicle is reduced. The unit writes the compensated target state back to the road spatiotemporal data frame and provides the delay compensation amount to the target intention and trajectory cluster analysis module for trajectory cluster candidate weight calculation.

[0020] The road spatiotemporal data frames output by the road spatiotemporal data access module are managed using a unified frame number. Each road spatiotemporal data frame consists of a frame header, road coordinate reference, vehicle status record, fused target record, unconfirmed target record, roadside coordination record, communication compensation record, and field integrity flag. This road spatiotemporal data frame serves as the sole input for subsequent modules, preventing them from repeatedly interpreting the target status under different coordinate systems or time references.

[0021] The target intent and trajectory cluster analysis module is used to transform moving targets in road spatiotemporal data frames into analyzable candidate trajectory clusters. First, the module extracts continuous motion segments of each target from the road spatiotemporal data frames and historical queues using a trajectory segmentation unit. Then, an intent candidate generation unit generates multiple intent candidates based on lane topology, traffic light phase, distance to the vehicle ahead, and target lateral offset. Finally, a behavior credibility update unit corrects the credibility scores of each intent candidate based on target motion residuals, communication compensation, and roadside coordination records, ultimately forming a target trajectory cluster with a time window and credibility score. The target trajectory clusters output by the target intent and trajectory cluster analysis module are then fed into the dynamic restricted area construction module.

[0022] The trajectory segmentation unit reads fused target records, unconfirmed target records, and roadside collaborative records from the road spatiotemporal data frames output by the road spatiotemporal data access module, and reads historical states with the same target identifier from the historical queue. The trajectory segmentation unit performs continuity checks based on target identifier, road longitudinal coordinates, road lateral coordinates, and target dimensions. When the spatial transition between two consecutive states exceeds the reachable range of the target at the current speed, the trajectory segmentation unit splits the target into two trajectory segments and marks the latter as a new trajectory segment. The trajectory segmentation unit establishes reconnection candidates for targets that reappear after being occluded. Reconnection candidates must simultaneously satisfy three conditions: similar road lateral coordinates, consistent longitudinal velocity direction, and interpretable occlusion boundary. Once these conditions are met, the trajectory segments before and after the break are connected into a single target trajectory segment.

[0023] The intent candidate generation unit reads the target trajectory segment output by the trajectory segmentation unit and combines it with lane centerlines, lane boundaries, intersection connections, no-lane-change lines, stop lines, yield lines, and traffic light phases from the high-precision map to generate intent candidates. For targets within a lane, the intent candidate generation unit constructs lane-keeping candidates, left lane entry candidates, right lane entry candidates, deceleration candidates, and acceleration candidates. For targets at intersection entrances, the intent candidate generation unit constructs straight-ahead candidates, left-turn candidates, right-turn candidates, and stop candidates. For unconfirmed target records, the intent candidate generation unit only generates low-speed-keeping candidates and stationary-occupancy candidates to prevent unconfirmed target records from receiving excessively high collaborative weights in subsequent modules. Each intent candidate is written with a start time, predicted end time, road longitudinal sequence, road lateral sequence, speed sequence, and candidate source.

[0024] The behavior credibility update unit performs progressive data analysis on the intent candidates output by the intent candidate generation unit. First, the behavior credibility update unit calculates the short-term motion residual between the target trajectory segment and each intent candidate. Then, it reads the communication compensation amount generated by the communication delay calibration unit, reducing the credibility score of candidates with larger communication compensation amounts. The behavior credibility update unit then reads the roadside coordination record. When the occlusion boundary in the roadside coordination record intersects with the target trajectory segment, the intent candidate in the occluded direction is retained but marked as occluded and awaiting verification. The behavior credibility update unit also analyzes the target's lateral offset trend. When the target's lateral offset continuously moves towards the centerline of the adjacent lane and the yaw rate is consistent with the direction of movement, the credibility score of the corresponding lane entry candidate is increased. The credibility score is not used as a single decision result but is passed to the dynamic restricted area construction module to determine the dynamic restricted area boundary expansion amount and activation window.

[0025] The target trajectory clusters generated by the target intent and trajectory cluster analysis module are managed according to target identifiers, and each target trajectory cluster stores multiple intent candidates. Each intent candidate in the target trajectory cluster is bound to the frame number of the road spatiotemporal data frame. The subsequent dynamic restricted area construction module retains this frame number when generating dynamic restricted area objects, enabling the operation verification and parameter calibration module to trace the source of the dynamic restricted area. The target intent and trajectory cluster analysis module does not directly output avoidance commands; all outputs are verifiable trajectory cluster data.

[0026] The dynamic restricted area construction module transforms target trajectory clusters, vehicle occupancy predictions, and road restriction areas into dynamic restricted area objects that can be used for planning and search. The module generates the vehicle's occupancy range under different candidate actions using a vehicle occupancy extrapolation unit, transforms the target trajectory cluster into time-varying occupancy ranges of other vehicles using a vehicle occupancy extrapolation unit, analyzes occupancy, communication delays, and adjacent vehicle avoidance chain relationships to form interactive shadow regions using an interactive shadow marking unit, and then uniformly writes the above occupancy ranges into the dynamic restricted area object using a dynamic restricted area encoding unit. The dynamic restricted area object output by the module then enters the cooperative conflict identification module and the feasible corridor planning module.

[0027] The vehicle occupancy extrapolation unit reads the vehicle status record output by the road spatiotemporal data access module and receives the steering, braking, and driving commands actually sent by the vehicle motion control module in the previous control cycle. Based on the current vehicle speed, yaw rate, steering angle, and vehicle dimensions, the unit generates vehicle occupancy ranges within a short-term prediction window, indicating whether the vehicle is maintaining current control, slightly decelerating, slightly accelerating, deviating to the left, or deviating to the right. The unit converts each vehicle occupancy range into a polygon sequence in the road coordinate system, with each polygon sequence corresponding to a future time point. The unit also considers vehicle actuator response delays, incorporating steering system response delays, braking system response delays, and driving system response delays into the vehicle occupancy ranges to prevent the dynamic restricted area construction module from misjudging areas that the vehicle cannot reach immediately as feasible areas.

[0028] The vehicle occupancy extrapolation unit reads the target trajectory cluster output by the target intent and trajectory cluster analysis module, and converts each intent candidate into a vehicle occupancy range. Based on the target's dimensions, orientation, speed sequence, and confidence score, the unit generates occupancy polygons along the longitudinal and transverse road sequences of the intent candidates. The boundary expansion of the occupancy polygon is jointly determined by the target speed, communication compensation, sensor source confidence marker, and the intent candidate's confidence score. When the confidence score is low but the target may conflict with the vehicle, the vehicle occupancy extrapolation unit does not delete the intent candidate but instead generates a dynamic restricted area object with a shorter activation time and a larger boundary expansion. When the confidence score is high and the target's movement is stable, the vehicle occupancy extrapolation unit generates a dynamic restricted area object with tighter boundaries and retains the release conditions.

[0029] The interactive shadow marking unit is used to identify and mark risky occupancy areas that cannot be directly obtained from a single target trajectory. The interactive shadow marking unit first reads the occlusion boundary provided by the roadside cooperative access unit, and generates an occlusion shadow region behind the occlusion boundary along the direction of possible entry into the road. The interactive shadow marking unit then reads adjacent vehicle trajectory clusters. When the intention candidates of two adjacent vehicles simultaneously point to the same lane gap, the interactive shadow marking unit constructs a gap squeezing region. The interactive shadow marking unit further performs counterfactual occupancy analysis: under three action assumptions—the candidate vehicle not yielding, the candidate vehicle rapidly decelerating, and the candidate vehicle continuing to accelerate—it calculates whether adjacent vehicles will be forced to enter the autonomous vehicle's planned corridor. When this situation exists, the interactive shadow marking unit marks the affected time segment in the autonomous vehicle's planned corridor as an interactive shadow region. The interactive shadow region is not a static obstacle, but a dynamic no-entry zone object with a source target and triggering action.

[0030] The dynamic restricted area coding unit writes the polygon sequences generated by the self-vehicle occupancy extrapolation unit, the other-vehicle occupancy extrapolation unit, and the interactive shadow marking unit into a unified data structure. The dynamic restricted area coding unit establishes a three-dimensional occupancy table indexed by the road longitudinal coordinates, road lateral coordinates, and time. Each cell in the three-dimensional occupancy table records the dynamic restricted area object number, activation start time, activation end time, restricted area type, and collaborative locking flag. When multiple dynamic restricted area objects occupy the same cell, the dynamic restricted area coding unit merges the boundaries based on activation time window overlap, confidence score, and source data category. After merging, all source target identifiers are retained for the collaborative conflict identification module to determine the set of conflicting vehicles. The dynamic restricted area coding unit also generates a dynamic restricted area index table, which the collaborative conflict identification module can quickly query through a time window to see if a candidate trajectory of a self-vehicle crosses a certain dynamic restricted area object.

[0031] The dynamic restricted area construction module generates three types of dynamic restricted area objects. The first type is entity-occupied dynamic restricted area objects, derived from actual vehicles, pedestrians, non-motorized vehicles, and temporary obstacles. The second type is collaboratively reserved dynamic restricted area objects, derived from avoidance intentions exchanged and locked between vehicles. The third type is interactive shadow dynamic restricted area objects, derived from occupancy analysis, communication delays, and counterfactual occupancy analysis. All three types of dynamic restricted area objects use the same data structure, but differ in restricted area type; subsequent modules employ different processing logics based on the restricted area type.

[0032] The cooperative conflict identification module is used to determine the spatiotemporal crossing relationship between the candidate actions of the autonomous vehicle and the objects in the dynamic restricted area, and to identify the set of vehicles that need to participate in cooperative avoidance. The cooperative conflict identification module first matches the candidate actions of the autonomous vehicle with the dynamic restricted area index table through the restricted area crossing detection unit. Then, it analyzes the deceleration, lateral offset, and time window compression that adjacent vehicles need to bear when the autonomous vehicle continues to execute the candidate action through the yield pressure calculation unit. Finally, it divides vehicles that have mutual influence into cooperative avoidance groups through the cooperative vehicle grouping unit. The cooperative avoidance groups and conflict causes output by the cooperative conflict identification module are then fed into the cooperative avoidance decision module.

[0033] The restricted area crossing detection unit reads the dynamic restricted area index table output by the dynamic restricted area construction module and the candidate occupancy range of the vehicle generated by the vehicle occupancy extrapolation unit. The unit queries whether the candidate occupancy range of the vehicle overlaps with the dynamic restricted area object in three dimensions: road longitudinal coordinates, road lateral coordinates, and time. For physical occupancy dynamic restricted area objects, the unit records an overlap as a physical crossing conflict. For collaboratively reserved dynamic restricted area objects, the unit also checks whether the collaborative locking flag is still valid; if the collaborative locking flag fails, the dynamic restricted area object is returned to the pending negotiation state. For interactive shadow dynamic restricted area objects, the unit reads their triggering action; if the triggering action matches the candidate action of the vehicle, it is recorded as an interactive crossing conflict; if the triggering action does not match the candidate action, it is only recorded as an observation conflict.

[0034] The yielding pressure calculation unit analyzes the impact of the vehicle's actions on the control boundaries of adjacent vehicles. It reads conflict records from the no-entry zone crossing detection unit and retrieves the target trajectory clusters of the conflicting vehicles from the target intent and trajectory cluster analysis module. For each conflicting vehicle, the yielding pressure calculation unit generates three executable corrective actions: maintain current trajectory, decelerate and yield, and lateral avoidance. The unit determines whether these corrective actions are still within the executable range based on the conflicting vehicle's current speed, lane boundaries, distance between vehicles, traffic light phase, and road curvature. If the conflicting vehicles can only resolve the conflict by rapidly decelerating or laterally avoiding the lane line, the yielding pressure calculation unit marks the corresponding candidate action as a high-pressure action. If the conflicting vehicles can resolve the conflict by slightly adjusting their speed, the unit marks the corresponding candidate action as a negotiable action.

[0035] The cooperative vehicle grouping unit transforms conflict records into cooperative avoidance groups. Using its own vehicle as the central node, the cooperative vehicle grouping unit adds adjacent vehicles that have physical crossing conflicts, interactive crossing conflicts, or high-pressure actions associated with its own vehicle to the same cooperative avoidance group. The cooperative vehicle grouping unit continues to check for mutual conflicts between vehicles within the cooperative avoidance group; if mutual conflicts exist, the relevant vehicles are added to the same cooperative avoidance group until no new vehicles are added. The cooperative vehicle grouping unit writes a group number, member vehicle identifiers, conflict dynamic restricted area object number, conflict type, conflict time window, conflict road coordinate range, and recommended negotiation start time to each cooperative avoidance group. The cooperative vehicle grouping unit does not generate a final avoidance trajectory, but instead provides the cooperative avoidance decision module with clear cooperative objects and conflict reasons.

[0036] The collaborative conflict identification module generates a conflict identification report for each control cycle. The conflict identification report records all inspected candidate actions of the autonomous vehicle, the corresponding dynamic restricted area objects, the conflict type, the yield pressure marker, and the collaborative avoidance group. The conflict identification report is also written to a history queue, allowing the runtime review and parameter calibration module to retrospectively investigate why a particular avoidance action was selected or prohibited based on the actual control results.

[0037] The collaborative avoidance decision module is used to determine vehicle avoidance roles within the collaborative avoidance group and convert the collaborative results into collaborative reserved dynamic restricted area objects. The collaborative avoidance decision module first sends its own candidate actions, yield pressure flags, and conflict time windows to vehicles within the collaborative avoidance group through the collaborative intent publishing unit. Then, through the avoidance role negotiation unit, it determines the primary passing vehicle, the decelerating yielding vehicle, and the lateral yielding vehicle based on right-of-way, time window margin, vehicle controllable boundaries, and historical negotiation records. The collaborative rule convergence unit writes consistent avoidance roles into the collaborative locking flag. When communication is interrupted or intents are inconsistent, the abnormal communication handling unit generates a conservative dynamic restricted area object and reverts to single-vehicle safe avoidance. The collaborative reserved dynamic restricted area object output by the collaborative avoidance decision module is returned to the dynamic restricted area construction module, while the collaborative avoidance roles are sent to the feasible corridor planning module.

[0038] The collaborative intent publishing unit reads the collaborative avoidance group output by the collaborative conflict identification module and the set of executable actions for the vehicle generated by the feasible corridor planning module in the previous control cycle. The collaborative intent publishing unit assembles a collaborative intent message from the vehicle's vehicle identifier, current position, current speed, candidate actions, conflict time window, yield pressure flag, and controllable boundary. The collaborative intent message is sent to vehicles within the collaborative avoidance group via V2X communication and is simultaneously forwarded to the roadside unit by the roadside collaborative access unit. The collaborative intent publishing unit writes a message sequence number and a valid time window to each collaborative intent message; collaborative intent messages that do not receive a response within the valid time window will no longer participate in role negotiation in the current control cycle.

[0039] The yielding role negotiation unit receives the collaborative intent message and acknowledgment returned by vehicles within the collaborative yielding group. The unit first determines the vehicle's right-of-way based on traffic rules, then calculates the yielding role based on the time difference between each vehicle's arrival at the conflict dynamic restricted area object, the controllable deceleration boundary, available lateral clearance, and the current control state. The primary through vehicle is allowed to occupy the collaboratively reserved dynamic restricted area object within the conflict time window. Vehicles decelerating to yield must complete speed reduction before the conflict time window, and vehicles yielding laterally must complete lateral road offset before the conflict time window. The yielding role negotiation unit does not use a single first-come, first-served rule, but instead uses the high-pressure action given by the yielding pressure calculation unit as a constraint. When a vehicle assuming a yielding role would cause it to enter an unexecutable control boundary, the yielding role negotiation unit reassigns roles until all vehicle roles are within the executable range, or it transitions to the abnormal communication handling unit.

[0040] The collaborative rule convergence unit is used to determine whether vehicles within the collaborative avoidance group have formed consistent roles. The unit reads the confirmed primary passing vehicle, yielding vehicle, and lateral yielding vehicle identifiers from each vehicle and checks whether the conflict time windows confirmed by each vehicle overlap. If all necessary vehicles return to consistent roles within the valid time window, the collaborative rule convergence unit writes the consistent roles into the collaborative locking flag and generates a collaborative reserved dynamic restricted area object. The collaborative reserved dynamic restricted area object reserves a short-term passage corridor for primary passing vehicles and marks prohibited entry time segments for yielding and lateral yielding vehicles. The collaborative rule convergence unit writes the collaborative reserved dynamic restricted area object back to the dynamic restricted area construction module, so that the dynamic restricted area index table in the next control cycle can reflect the collaborative results.

[0041] The abnormal communication handling unit processes communication interruptions, message conflicts, and coordination timeouts. It is activated when a coordination intent message is not acknowledged, adjacent vehicle role confirmations are inconsistent, roadside unit return status conflicts occur, or the communication compensation exceeds a set window. Instead of waiting for new coordination messages, the unit generates a conservative dynamic restricted area object based on existing road spatiotemporal data frames. This object covers the longitudinal and lateral coordinate ranges of the road that adjacent vehicles may occupy within the conflict time window, and has a short activation endpoint time. Simultaneously, the unit sends a single-vehicle safety avoidance marker to the feasible corridor planning module, ensuring that the module searches for trajectories only within the space occupied by the dynamic restricted area object and the conservative dynamic restricted area object without crossing them.

[0042] The cooperative avoidance decision module locks only short-term cooperative roles in each control cycle, and the locking duration must not exceed the time range that the vehicle controller can stably predict. In the next control cycle, the cooperative avoidance decision module rereads the outputs of the dynamic restricted area construction module and the cooperative conflict identification module. When an adjacent vehicle does not move according to the cooperative role, the cooperative avoidance decision module cancels the original cooperative locking mark, and the abnormal communication handling unit generates a new conservative dynamic restricted area object.

[0043] The feasible corridor planning module is used to generate an executable trajectory for the vehicle under the constraints of a dynamic restricted area object. The module first filters passable road time windows from the dynamic restricted area index table using a lane time window filtering unit. Then, it determines the lateral movement range of the vehicle within its lane or adjacent lanes using a lateral offset correction unit. Next, it generates a speed curve consistent with the cooperative avoidance role using a speed curve arrangement unit. Finally, it combines the lateral movement range and speed curve into an executable trajectory using a trajectory smoothing verification unit. The feasible corridor planning module outputs the executable trajectory to the vehicle motion control module and returns the set of executable actions to the cooperative avoidance decision module.

[0044] The lane time window filtering unit reads the dynamic restricted area index table output by the dynamic restricted area construction module and the cooperative avoidance role output by the cooperative avoidance decision module. The lane time window filtering unit scans the longitudinal coordinates of the road along the vehicle's target route and queries the corresponding dynamic restricted area objects within the lateral range of the road at future times. For objects occupying dynamic restricted areas, the lane time window filtering unit directly excludes overlapping areas. For collaboratively reserved dynamic restricted area objects, the lane time window filtering unit determines whether to allow entry into the corresponding time window based on whether the vehicle is the primary vehicle. For interactive shadow dynamic restricted area objects, the lane time window filtering unit determines whether to exclude them based on the triggering action; the corresponding time window is excluded only when the vehicle's candidate action will trigger the interactive shadow dynamic restricted area object. The lane time window filtering unit outputs a passable time window table, which includes the road's longitudinal start point, road's longitudinal end point, allowed lateral range, allowed entry time, and mandatory exit time.

[0045] The lateral offset correction unit generates the lateral movement range based on the passable time window table. The lateral offset correction unit reads lane boundaries, lane width, road curvature, boundaries of adjacent vehicle dynamic restricted areas, and the vehicle's steering system response delay. The lateral offset correction unit first determines the safe lateral band within the vehicle's current lane. When the safe lateral band is compressed by a dynamic restricted area object, the lateral offset correction unit attempts to generate a lateral band for adjacent lanes. The adjacent lane lateral band must meet three conditions: lane lines allow crossing, vehicles behind the target lane do not enter within the conflict time window, and there is sufficient longitudinal clearance for vehicles ahead of the target lane. When generating the lateral band, the lateral offset correction unit retains the lateral movement start point, lateral movement end point, and lateral movement completion time for the speed curve arrangement unit to synchronize longitudinal speed.

[0046] The speed curve orchestration unit reads the passable time window table, the lateral movement range output by the lateral offset correction unit, and the cooperative avoidance role output by the cooperative avoidance decision module. If the vehicle is the primary through vehicle, the speed curve orchestration unit generates a speed curve that can pass through the conflict road coordinate range within the activation time window of the cooperative reserved dynamic restricted area object. If the vehicle is a decelerating yielding vehicle, the speed curve orchestration unit generates a speed curve that reduces speed before the conflict time window and maintains a recoverable driving speed curve. If the vehicle is a lateral yielding vehicle, the speed curve orchestration unit ensures that the lateral offset completion time is earlier than the start of the conflict time window and limits the longitudinal speed to within the range that the steering system can execute. The speed curve orchestration unit writes the start speed, end speed, maximum longitudinal acceleration, maximum deceleration, comfort limit marker, and actuator limit marker to each speed curve.

[0047] The trajectory smoothing verification unit combines the lateral movement range output by the lateral offset correction unit and the velocity curve output by the velocity curve arrangement unit into an executable trajectory for the vehicle. The trajectory smoothing verification unit generates a trajectory centerline along the road's longitudinal coordinates and converts the trajectory centerline into a sequence of target points in the vehicle coordinate system. The trajectory smoothing verification unit checks the curvature change, lateral acceleration, yaw rate, tire adhesion utilization, and steering angle change of the target point sequence. If the target point sequence triggers an actuator limit marker, the trajectory smoothing verification unit feeds back to the velocity curve arrangement unit that the speed needs to be reduced; if reducing the speed still fails to meet the actuator boundary requirements, the trajectory smoothing verification unit marks the vehicle's executable trajectory as unexecutable. The executable trajectory is written with the trajectory number, target point sequence, velocity sequence, activation time, applicable cooperative avoidance role, and associated dynamic restricted area object number.

[0048] The feasible corridor planning module outputs a vehicle executable trajectory divided into a main trajectory and a backup trajectory. The main trajectory is sent to the vehicle motion control module for execution, while the backup trajectory is activated when the vehicle motion control module detects excessive tracking deviation or when adjacent vehicles fail to move according to their cooperative roles. The feasible corridor planning module also returns the set of executable actions generated in this control cycle to the cooperative avoidance decision module as the message content for the cooperative intent release unit in the next control cycle.

[0049] The vehicle motion control module converts the vehicle's executable trajectory into chassis control commands and verifies whether the dynamic restricted area object still meets the control constraints during execution. The vehicle motion control module first calculates steering, braking, and drive feedforwards based on the vehicle's executable trajectory using a feedforward control quantity generation unit. Then, it corrects the control quantities based on the actual vehicle state using a feedback tracking correction unit. Finally, it determines whether the control quantities are within the vehicle's executable boundaries using an actuator constraint verification unit. When the vehicle's executable trajectory is truncated by a new dynamic restricted area object or the control deviation cannot be eliminated, a backup trajectory or minimum risk stop is initiated through a safety takeover trigger unit. The actual control results from the vehicle motion control module are transmitted back to the operation verification and parameter calibration module.

[0050] The feedforward control quantity generation unit reads the main trajectory output by the feasible corridor planning module and the vehicle status record output by the road spatiotemporal data access module. The feedforward control quantity generation unit generates a basic steering input based on the curvature change of the target point sequence and generates drive and braking feedforward inputs based on the speed sequence. The feedforward control quantity generation unit uses several future points in the target point sequence as control preview points; the number of preview points varies with vehicle speed, increasing the preview distance at higher speeds and decreasing it at lower speeds. The control quantity output by the feedforward control quantity generation unit carries a trajectory number and a control cycle marker, facilitating the feedback tracking correction unit and the operation verification and parameter calibration module to trace the control source.

[0051] The feedback tracking correction unit reads the control quantity output by the feedforward control quantity generation unit and the actual vehicle state provided by the vehicle-side motion acquisition unit. The feedback tracking correction unit calculates the lateral deviation, heading deviation, and speed deviation between the vehicle's current position and the main trajectory target point, and then corrects the steering, braking, and driving quantities. During the correction process, the feedback tracking correction unit reads the latest dynamic restricted area index table from the dynamic restricted area construction module. When the vehicle's actual position is close to the boundary of a physical object occupying a dynamic restricted area or a conservative dynamic restricted area object, the feedback tracking correction unit restricts further increases in steering quantity towards that boundary, preventing control correction from pushing the vehicle into the restricted space.

[0052] The actuator constraint verification unit receives the control quantity output by the feedback tracking correction unit and reads the vehicle chassis status. The actuator constraint verification unit checks the steering angular velocity, brake pressure changes, drive torque changes, wheel speed difference, tire adhesion utilization, and vehicle yaw stability. When the control quantity exceeds the actuator's executable boundary, the actuator constraint verification unit first trims the control quantity and sends a control-limited flag to the feasible corridor planning module. If the trimmed control quantity still cannot keep the vehicle within the passable time window corresponding to the main trajectory, the actuator constraint verification unit marks the main trajectory as failed and requests the safety takeover trigger unit to activate the backup trajectory.

[0053] The safety takeover trigger unit activates under three conditions. The first condition is that the dynamic restricted zone construction module generates a new entity occupying a dynamic restricted zone object during the current control cycle, which truncates the main trajectory. The second condition is that the cooperative avoidance decision module cancels the cooperative locking marker, and the vehicle no longer has the entry conditions for the original cooperative reserved dynamic restricted zone object. The third condition is that the actuator constraint verification unit determines that the main trajectory is unexecutable. The safety takeover trigger unit first determines whether the backup trajectory output by the feasible corridor planning module still does not cross the entity occupying the dynamic restricted zone object and the conservative dynamic restricted zone object. If the condition is met, it switches to the backup trajectory. If the backup trajectory is also unexecutable, the safety takeover trigger unit generates a minimum risk stopping command, causing the vehicle to decelerate within the available space of the current lane and maintain a controllable attitude.

[0054] The vehicle motion control module writes the actual steering commands, braking commands, drive commands, trajectory numbers, dynamic restricted zone object numbers, control restriction markers, and safety takeover trigger reasons sent to the chassis into the control execution record. The control execution record enters the operation review and parameter calibration module for subsequent analysis of whether the dynamic restricted zone boundary is too wide or too narrow, or whether the activation window is offset.

[0055] The operation verification and parameter calibration module is used to perform closed-loop calibration of the dynamic restricted area parameter boundaries based on actual operation data. First, the module replays the actual distance changes between the vehicle's executable trajectory and the dynamic restricted area object through the restricted area hit playback unit. Then, it calculates the deviation between the cooperative avoidance role and the actual behavior of adjacent vehicles through the negotiation deviation statistics unit. Finally, it updates the dynamic restricted area boundary expansion, activation window, and confidence score boundaries through the parameter boundary update unit, and seals abnormal fragments into reproducible test scenarios through the scene fragment sealing unit. The parameter boundaries output by the operation verification and parameter calibration module are returned to the dynamic restricted area construction module, the cooperative conflict identification module, and the feasible corridor planning module.

[0056] The restricted area hit playback unit reads the control execution record output by the vehicle motion control module and reads the dynamic restricted area objects, road spatiotemporal data frames, and vehicle executable trajectories from the historical queue. The restricted area hit playback unit reconstructs the vehicle's actual motion trajectory according to the control cycle and calculates the minimum distance between the vehicle's actual body boundary and the boundaries of each dynamic restricted area object. The restricted area hit playback unit correlates the minimum distance change with the confidence score, activation window, and restricted area type of the dynamic restricted area object at that time to form a restricted area hit record. If the vehicle's actual body boundary continuously approaches the boundary of a certain type of dynamic restricted area object, the restricted area hit playback unit marks that type of dynamic restricted area object as a candidate with a narrow boundary. If the vehicle's actual body boundary is consistently far from the boundary of a certain type of dynamic restricted area object and the vehicle exhibits obvious detours, the restricted area hit playback unit marks that type of dynamic restricted area object as a candidate with a wide boundary.

[0057] The negotiation deviation statistics unit reads the cooperative avoidance roles and the actual trajectories of adjacent vehicles output by the cooperative avoidance decision module. The unit calculates the actual arrival time, actual speed change, and actual lateral offset for vehicles passing through, vehicles slowing down to yield, and vehicles yielding laterally. If an adjacent vehicle continuously fails to move according to its confirmed role, the negotiation deviation statistics unit reduces the credibility weight of that vehicle's source message in the behavior credibility update unit. If vehicles in a certain scenario generally slow down early or delay yielding, the negotiation deviation statistics unit writes the corresponding scenario label into the parameter boundary update unit, causing the dynamic no-entry zone construction module to adjust its activation window in the same scenario.

[0058] The parameter boundary update unit updates system parameters based on the data output by the restricted area hit playback unit and the negotiation deviation statistics unit. The parameter boundary update unit does not directly change vehicle control commands; instead, it updates the upper and lower limits of boundary expansion, activation window advance, activation window delay, communication compensation penalty weight, and intent candidate confidence score boundaries used by the dynamic restricted area construction module. The parameter boundary update unit employs scenario-specific parameter table management, with different parameter table entries corresponding to different road curvatures, vehicle speed ranges, communication delay ranges, and occlusion intensities. After each update, the parameter boundary update unit generates a parameter version number, which must be recorded by the dynamic restricted area construction module, the cooperative conflict identification module, and the feasible corridor planning module.

[0059] The scene fragment sealing unit is used to save abnormal operation fragments. The scene fragment sealing unit is activated when the safety takeover trigger unit starts, the cooperative locking flag is canceled, a candidate for a narrow dynamic restricted area object boundary appears, or adjacent vehicles do not move according to their cooperative roles. The scene fragment sealing unit seals the road spatiotemporal data frames, target trajectory clusters, dynamic restricted area objects, cooperative avoidance groups, vehicle executable trajectories, and control execution records before and after the anomaly occurs into scene fragments. Each scene fragment includes a scene number, road type, number of targets, conflict type, communication delay range, and triggering reason. Scene fragments can be reproduced in an offline simulation platform to verify whether the new parameter version generated by the parameter boundary update unit changes the dynamic restricted area object boundary and the vehicle executable trajectory.

[0060] The operational verification and parameter calibration module employs a combination of offline and online lightweight calibration. Online lightweight calibration only allows adjustments to confidence scores and short-term activation windows, without altering vehicle dynamics boundaries. Offline calibration updates the sub-scenario parameter tables based on archived scene fragments and releases a new parameter version after simulation verification. After the vehicle is powered on, the road spatiotemporal data access module reads the current parameter version number and transmits it to subsequent modules with each road spatiotemporal data frame, ensuring that all modules use the same parameter version.

[0061] This embodiment provides a vehicle cooperative avoidance method based on dynamic restricted areas, which can be executed by the aforementioned vehicle cooperative avoidance system based on dynamic restricted areas. This method uses the vehicle control cycle as the execution unit, continuously processing road environment data, target motion data, communication coordination data, and vehicle control data within each control cycle, and using the processing results for vehicle cooperative avoidance control.

[0062] Specifically, the system first acquires the vehicle's status, external target status, roadside cooperative information, and communication delay information. The vehicle status includes the vehicle's center point coordinates, heading, speed, longitudinal acceleration, lateral acceleration, yaw rate, steering angle, brake opening, drive torque, wheel speed difference, and gear position. External target status includes the position, speed, orientation, and dimensions of vehicles, pedestrians, non-motorized vehicles, and temporary obstacles detected by vehicle-mounted sensors. Roadside cooperative information includes information from roadside units regarding road construction areas, accident-occupied lane areas, traffic light phases, obstruction boundaries, intersection conflict zones, and roadside sensing targets. Communication delay information includes the generation and arrival times of messages sent by adjacent vehicles or roadside units, and the corresponding communication compensation amounts.

[0063] Subsequently, the vehicle status, the external target status, the roadside coordination information, and the communication delay information are converted into road spatiotemporal data frames under the same control cycle. During the conversion process, data from different sources are unified to the road coordinate system, and time delay compensation is performed on the status of adjacent vehicles based on the message generation time, message arrival time, and the current control cycle time, so that the data within the same road spatiotemporal data frame has a consistent time and coordinate reference.

[0064] Then, target trajectory clusters are generated based on the road spatiotemporal data frames. When generating target trajectory clusters, trajectory segments of each target in a continuous control cycle are first extracted. Then, lane centerlines, lane boundaries, intersection connections, no-change lines, stop lines, yield lines, and traffic light phases are combined to generate intent candidates. The confidence scores of each intent candidate are updated based on target motion residuals, communication compensation, occlusion boundaries, target lateral offset trends, and target yaw rates. Each target trajectory cluster includes at least one intent candidate with a time window, road longitudinal sequence, road lateral sequence, speed sequence, and confidence score.

[0065] After obtaining the target trajectory cluster, a dynamic restricted area object is generated based on the target trajectory cluster, the predicted occupancy of the vehicle itself, the predicted occupancy of other vehicles, the occupancy area, and the communication delay status. When generating the dynamic restricted area object, the vehicle occupancy range corresponding to the candidate actions of the vehicle itself, the vehicle occupancy range corresponding to the candidate intentions of other vehicles, and the interactive shadow area formed by occupancy, communication delay, and counterfactual occupancy analysis are uniformly encoded as the dynamic restricted area object. The dynamic restricted area object has a restricted area number, activation start time, activation end time, road longitudinal coordinate range, road lateral coordinate range, boundary polyline, source vehicle identifier, source data category, restricted area type, confidence score, cooperative locking flag, release condition, and control cycle flag.

[0066] Next, the dynamic restricted area object is written into the dynamic restricted area index table, and the spatiotemporal travel relationship between the candidate vehicle action and the dynamic restricted area object is detected. During detection, the candidate vehicle's occupied area corresponding to the candidate vehicle action is matched with the dynamic restricted area object in three dimensions: road longitudinal coordinate, road lateral coordinate, and time. When the candidate vehicle action overlaps with an entity occupying a dynamic restricted area object, a collaboratively reserved dynamic restricted area object, or an interactive shadow dynamic restricted area object, the corresponding conflict type and conflict time window are recorded.

[0067] When a spatiotemporal crossing relationship is detected between the candidate vehicle action and the dynamic restricted area object, a cooperative avoidance group is generated, and the primary through vehicle, the decelerating yielding vehicle, and the lateral yielding vehicle are determined based on the cooperative avoidance group. When determining the yielding roles, the traffic right-of-way, the time difference between each vehicle's arrival at the conflicting dynamic restricted area object, the controllable deceleration boundary, the available lateral clearance, and the current control state are considered. If vehicles in the cooperative avoidance group return a consistent yielding role within the effective time window, a cooperative locking flag is generated; if the cooperative intent message is not acknowledged, the roles of adjacent vehicles are inconsistent, the roadside unit returns a conflicting status, or the communication compensation exceeds the set window, a conservative dynamic restricted area object is generated.

[0068] Then, based on the primary passing vehicle, the decelerating yielding vehicle, and the lateral yielding vehicle, a collaboratively reserved dynamic restricted area object is generated, and an executable vehicle trajectory is generated under the constraints of the dynamic restricted area index table. When generating the executable vehicle trajectory, firstly, passable time windows are filtered according to the dynamic restricted area index table; then, the lateral movement range is generated based on lane boundaries, road curvature, the boundaries of adjacent vehicle dynamic restricted area objects, and the vehicle's steering system response delay. Subsequently, speed curves are arranged according to the yielding role, and the lateral movement range and speed curves are combined into the executable vehicle trajectory. After the executable vehicle trajectory is generated, its curvature change, lateral acceleration, yaw rate, tire adhesion utilization, and steering angle change are verified.

[0069] Finally, the vehicle's executable trajectory is converted into steering, braking, and driving commands, and the boundary expansion, activation window, and confidence score boundary of the dynamic restricted area object are calibrated based on the vehicle's actual operating data. During vehicle execution, if a new entity occupies the dynamic restricted area object, truncating the vehicle's executable trajectory, the cooperative locking marker is canceled, or the vehicle's executable trajectory cannot meet the actuator boundaries, a backup trajectory is switched or a minimum risk parking command is generated.

[0070] This embodiment provides an electronic device that can be used as a vehicle domain controller, autonomous driving controller, intelligent driving computing platform, or on-board central computing unit. The electronic device includes a processor, memory, a vehicle communication interface, and a chassis control interface.

[0071] The processor executes a computer program stored in the memory. When executing the computer program, the processor implements the aforementioned vehicle cooperative avoidance method based on dynamic restricted zones. The processor can be an automotive-grade central processing unit, graphics processing unit, neural network processor, microcontroller, or a computing unit formed by combining the above processors. When executing the computer program, the processor calls program instructions corresponding to road spatiotemporal data access, target intent and trajectory cluster analysis, dynamic restricted zone construction, cooperative conflict identification, cooperative avoidance decision-making, feasible corridor planning, vehicle motion control, and operation verification and parameter calibration according to the vehicle control cycle.

[0072] The memory is used to store computer programs, road spatiotemporal data frames, target trajectory clusters, dynamic restricted area objects, dynamic restricted area index tables, cooperative avoidance groups, vehicle executable trajectories, control execution records, and parameter version data. The memory may include volatile memory and non-volatile memory, wherein the volatile memory is used to store data caches for the current control cycle and historical control cycles, and the non-volatile memory is used to store program files, parameter tables, scene fragments, and calibration results.

[0073] The vehicle communication interface is used to receive data from vehicle sensors, vehicle positioning equipment, V2X communication equipment, and roadside units. The data received by the vehicle communication interface includes the vehicle's own status, external target status, roadside cooperation information, and communication delay information. The vehicle communication interface is also used to send cooperation intent messages, cooperation confirmation messages, and abnormal communication handling messages to neighboring vehicles or roadside units.

[0074] The chassis control interface is used to output control commands to the vehicle's steering, braking, and drive systems. These control commands include steering, braking, and drive commands generated based on the vehicle's executable trajectory. The chassis control interface also receives chassis execution feedback, which includes actual steering angle, braking pressure, drive torque, wheel speed difference, and vehicle stability. The processor tracks and corrects the vehicle's executable trajectory based on the chassis execution feedback, and generates backup trajectory switching commands or minimum risk stopping commands when the trajectory cannot be executed.

[0075] This embodiment provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the aforementioned vehicle cooperative avoidance method based on dynamic no-entry zones.

[0076] The computer program includes program instructions for acquiring the vehicle's status, external target status, roadside cooperation information, and communication delay information; program instructions for converting the vehicle's status, external target status, roadside cooperation information, and communication delay information into road spatiotemporal data frames; program instructions for generating target trajectory clusters based on the road spatiotemporal data frames; program instructions for generating dynamic restricted area objects based on target trajectory clusters, vehicle occupancy prediction, other vehicle occupancy prediction, obstruction areas, and communication delay status; program instructions for writing dynamic restricted area objects into a dynamic restricted area index table; program instructions for detecting the spatiotemporal crossing relationship between candidate vehicle actions and dynamic restricted area objects; program instructions for generating cooperative avoidance groups and determining the primary passing vehicle, decelerating and yielding vehicles, and lateral yielding vehicles; program instructions for generating cooperative reserved dynamic restricted area objects and vehicle executable trajectories; and program instructions for converting vehicle executable trajectories into steering commands, braking commands, and driving commands.

[0077] The computer-readable storage medium can be a non-volatile memory inside an electronic device or an external storage medium that can be read by the electronic device. The computer program stored in the computer-readable storage medium can be organized in a modular manner, so that the road spatiotemporal data access module, target intent and trajectory cluster analysis module, dynamic restricted area construction module, cooperative conflict identification module, cooperative avoidance decision module, feasible corridor planning module, vehicle motion control module, and operation verification and parameter calibration module correspond to different program segments, or it can be organized into continuous program segments according to the execution order of the vehicle control cycle.

[0078] When the computer program is executed by the processor, the processor can read road spatiotemporal data frames in each control cycle, generate target trajectory clusters and dynamic restricted area objects, complete cooperative conflict identification and cooperative avoidance decisions based on the dynamic restricted area index table, and generate vehicle executable trajectories and chassis control commands. Thus, the computer-readable storage medium enables an electronic device carrying the computer program to execute a vehicle cooperative avoidance control process based on dynamic restricted areas.

[0079] Secondly: The accompanying drawings of the embodiments disclosed in this invention only involve the structures involved in the embodiments disclosed in this invention. Other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of this invention can be combined with each other. In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A vehicle cooperative avoidance system based on dynamic restricted zones, characterized in that, The vehicle cooperative avoidance system based on dynamic restricted areas is deployed within the vehicle domain controller and is connected to vehicle sensors, vehicle positioning equipment, V2X communication equipment, steering system, braking system and drive system; The vehicle cooperative avoidance system based on dynamic restricted areas includes a road spatiotemporal data access module, a target intent and trajectory cluster analysis module, a dynamic restricted area construction module, a cooperative conflict identification module, a cooperative avoidance decision module, a feasible corridor planning module, a vehicle motion control module, and an operation verification and parameter calibration module. The road spatiotemporal data access module is used to convert vehicle status, external target status, roadside coordination information and communication delay information into road spatiotemporal data frames under the same control cycle. The target intent and trajectory cluster analysis module is used to generate target trajectory clusters based on the road spatiotemporal data frames; The dynamic restricted area construction module is used to generate dynamic restricted area objects based on the target trajectory cluster, vehicle occupancy prediction, other vehicle occupancy prediction, occupancy area and communication delay status, and write the dynamic restricted area objects into the dynamic restricted area index table; The cooperative conflict identification module is used to detect the spatiotemporal travel relationship between the candidate actions of the vehicle and the dynamic restricted area object, and generate a cooperative avoidance group; The collaborative avoidance decision module is used to determine the main passing vehicle, the decelerating and yielding vehicle, and the lateral yielding vehicle according to the collaborative avoidance group, and convert the determined collaborative results into collaborative reserved dynamic no-entry zone objects; The feasible corridor planning module is used to generate an executable trajectory for vehicles under the constraints of the dynamic restricted area index table; The vehicle motion control module is used to convert the vehicle's executable trajectory into steering commands, braking commands, and driving commands. The operation verification and parameter calibration module is used to calibrate the boundary expansion, activation window, and confidence score boundary of the dynamic restricted area object based on the actual vehicle operation data.

2. The vehicle cooperative avoidance system based on dynamic restricted areas according to claim 1, characterized in that, The road spatiotemporal data access module includes a vehicle motion acquisition unit, an external target fusion unit, a roadside cooperative access unit, and a communication delay calibration unit. The vehicle-end motion acquisition unit is used to acquire the vehicle's center point coordinates, vehicle orientation, vehicle speed, longitudinal acceleration, lateral acceleration, yaw rate, steering angle, brake opening, drive torque, wheel speed difference, and vehicle gear. The external target fusion unit is used to uniformly convert the camera target frame, millimeter-wave radar target point, lidar clustered target and map matching results to the road coordinate system, and generate fused target record and unconfirmed target record; The roadside collaborative access unit is used to receive information such as road construction areas, accident-occupied areas, traffic light phases, obstruction boundaries, intersection conflict zones, and roadside sensing targets sent by the roadside unit, and convert the received content into roadside collaborative records. The communication delay calibration unit is used to calculate the communication compensation amount based on the message generation time, message arrival time and the control cycle time of the vehicle, and to extrapolate the status of adjacent vehicles to the current control cycle. The road spatiotemporal data access module writes the vehicle status, the fused target record, the unconfirmed target record, the roadside collaboration record, and the communication compensation amount into the road spatiotemporal data frame.

3. The vehicle cooperative avoidance system based on dynamic restricted areas according to claim 1, characterized in that, The target intent and trajectory cluster analysis module includes a trajectory segmentation unit, an intent candidate generation unit, and a behavior credibility update unit; The trajectory segmentation unit is used to extract target trajectory segments from the road spatiotemporal data frame according to the target identifier, road longitudinal coordinates, road transverse coordinates and target shape dimensions, and to split the target trajectory segments when the target space transition exceeds the corresponding speed reachable range. The intent candidate generation unit is used to generate lane-keeping candidates, left lane entry candidates, right lane entry candidates, deceleration candidates, acceleration candidates, straight-ahead candidates, left turn candidates, right turn candidates, and stop candidates based on lane center lines, lane boundaries, intersection connections, no-lane-change lines, stop lines, yield lines, and traffic light phases. The behavior credibility update unit is used to update the credibility score of each intention candidate based on the motion residual between the target trajectory segment and each intention candidate, the communication compensation amount, the occlusion boundary, the target lateral offset change trend and the target yaw rate. The target intent and trajectory cluster analysis module will compose the target trajectory cluster into intent candidates with confidence scores, time windows, road longitudinal sequences, road lateral sequences, and speed sequences.

4. The vehicle cooperative avoidance system based on dynamic restricted areas according to claim 1, characterized in that, The dynamic restricted area construction module includes a vehicle occupancy extrapolation unit, a other vehicle occupancy extrapolation unit, an interactive shadow marking unit, and a dynamic restricted area coding unit; The vehicle occupancy extrapolation unit is used to generate a candidate occupancy range for the vehicle based on the vehicle speed, yaw rate, steering angle, vehicle dimensions, steering system response delay, braking system response delay, and drive system response delay. The other vehicle occupancy extrapolation unit is used to generate the other vehicle occupancy range based on the intention candidates, target dimensions, target orientation, speed sequence and confidence score in the target trajectory cluster; The interactive shadow marking unit is used to generate interactive shadow regions based on occlusion boundaries, adjacent vehicle trajectory clusters, lane gaps, and counterfactual occupancy analysis results. The counterfactual occupancy analysis results are determined by the controllable state of adjacent vehicles under three action assumptions: candidate vehicle does not yield, candidate vehicle decelerates suddenly, and candidate vehicle continues to accelerate. The dynamic restricted area coding unit is used to encode the candidate occupancy range of the self-vehicle, the occupancy range of the other vehicle, and the interactive shadow area into the dynamic restricted area object, and to establish the dynamic restricted area index table according to the longitudinal coordinates of the road, the transverse coordinates of the road, and time; The dynamic restricted area object must have at least the restricted area number, activation start time, activation end time, road longitudinal coordinate range, road lateral coordinate range, boundary polyline, source vehicle identifier, source data category, restricted area type, confidence score, collaborative locking mark, release condition, and control cycle mark.

5. The vehicle cooperative avoidance system based on dynamic restricted areas according to claim 1, characterized in that, The cooperative conflict identification module includes a restricted area crossing detection unit, a yielding pressure calculation unit, and a cooperative vehicle grouping unit. The restricted area crossing detection unit is used to detect whether the occupancy range of the candidate vehicle corresponding to the candidate vehicle action overlaps with the dynamic restricted area object in three dimensions: road longitudinal coordinate, road lateral coordinate, and time. The restricted area crossing detection unit is also used to generate entity crossing conflicts, collaborative reserved crossing conflicts, and interactive crossing conflicts based on entity occupying dynamic restricted area objects, collaborative reserved dynamic restricted area objects, and interactive shadow dynamic restricted area objects, respectively. The yielding pressure calculation unit is used to generate three types of executable corrective actions for conflicting vehicles: maintaining the current trajectory, slowing down to yield, and lateral avoidance. Based on the current speed of the conflicting vehicles, lane boundaries, gaps between vehicles, traffic light phases, and road curvature, it determines whether the three types of executable corrective actions are within the executable range. The cooperative vehicle grouping unit is used to classify adjacent vehicles that have physical crossing conflicts, cooperative reserved crossing conflicts, interactive crossing conflicts, or high-pressure actions associated with the vehicle into the cooperative avoidance group. The cooperative vehicle grouping unit is also used to write the group number, member vehicle identifier, conflict dynamic restricted area object number, conflict type, conflict time window, conflict road coordinate range, and recommended negotiation start time into the cooperative avoidance group.

6. The vehicle cooperative avoidance system based on dynamic restricted areas according to claim 1, characterized in that, The collaborative avoidance decision-making module includes a collaborative intent publishing unit, an avoidance role negotiation unit, a collaborative rule convergence unit, and an abnormal communication handling unit. The cooperative intent publishing unit is used to send the vehicle's vehicle identifier, current position, current speed, candidate actions, conflict time window, yield pressure marker, and controllable boundary to the vehicles in the cooperative avoidance group. The avoidance role negotiation unit is used to determine the main passing vehicle, the decelerating yielding vehicle, and the lateral yielding vehicle based on traffic right-of-way, the time difference between each vehicle's arrival at the conflict dynamic restricted area object, the controllable deceleration boundary, the available lateral clearance, and the current control state. The collaborative rule convergence unit is used to generate a collaborative locking flag when vehicles in the collaborative avoidance group return to a consistent avoidance role and a consistent conflict time window, and write the collaborative locking flag into the collaborative reserved dynamic restricted area object; The abnormal communication handling unit is used to generate a conservative dynamic restricted area object when there is no reply to the collaborative intent message, inconsistent confirmation of adjacent vehicle roles, conflicting return status of roadside units, or communication compensation amount exceeds the set window, and send a single vehicle safety avoidance mark to the feasible corridor planning module.

7. The vehicle cooperative avoidance system based on dynamic restricted areas according to claim 1, characterized in that, The feasible corridor planning module includes a lane time window filtering unit, a lateral offset correction unit, a speed curve arrangement unit, and a trajectory smoothing verification unit. The vehicle motion control module includes a feedforward control quantity generation unit, a feedback tracking and correction unit, an actuator constraint verification unit, and a safety takeover trigger unit. The lane time window filtering unit is used to filter passable time windows based on the dynamic restricted area index table and the cooperative avoidance role; The lateral offset correction unit is used to generate the lateral movement range based on the lane boundary, road curvature, boundary of adjacent vehicle dynamic restricted area objects, and vehicle steering system response delay. The speed curve arrangement unit is used to generate speed curves based on the avoidance roles corresponding to the main passing vehicle, the decelerating yielding vehicle, or the lateral yielding vehicle. The trajectory smoothing verification unit is used to combine the lateral movement range and the velocity curve into the vehicle's executable trajectory, and to verify the curvature change, lateral acceleration, yaw rate, tire adhesion utilization rate and steering angle change of the vehicle's executable trajectory. The feedforward control quantity generation unit is used to generate steering feedforward, braking feedforward and driving feedforward based on the vehicle's executable trajectory. The feedback tracking correction unit is used to correct the control quantity based on the lateral deviation, heading deviation and speed deviation between the current position of the vehicle and the executable trajectory of the vehicle. The actuator constraint verification unit is used to check the steering angular velocity, braking pressure change, driving torque change, wheel speed difference, tire adhesion utilization rate and vehicle body yaw stability. The safety takeover trigger unit is used to switch to a backup trajectory or generate a minimum risk parking instruction when a new entity occupies a dynamic restricted area object that truncates the vehicle's executable trajectory, the cooperative locking mark is canceled, or the vehicle's executable trajectory cannot meet the actuator boundary.

8. A vehicle cooperative avoidance method based on dynamic restricted areas, characterized in that, The vehicle cooperative avoidance system based on dynamic restricted areas, applied to any one of claims 1 to 7, wherein the vehicle cooperative avoidance method based on dynamic restricted areas comprises: The system acquires the vehicle status, external target status, roadside coordination information, and communication delay information, and converts the vehicle status, external target status, roadside coordination information, and communication delay information into road spatiotemporal data frames under the same control cycle. A target trajectory cluster is generated based on the road spatiotemporal data frame, and a dynamic no-entry zone object is generated based on the target trajectory cluster, vehicle occupancy prediction, other vehicle occupancy prediction, occupancy area, and communication delay status. Write the dynamic restricted area object into the dynamic restricted area index table, and detect the spatiotemporal travel relationship between the candidate actions of the vehicle and the dynamic restricted area object; When a spatiotemporal travel relationship is detected between the candidate action of the autonomous vehicle and the object of the dynamic restricted area, a cooperative avoidance group is generated, and the main passing vehicle, the vehicle that slows down to give way, and the vehicle that gives way laterally are determined according to the cooperative avoidance group. Based on the main passing vehicle, the decelerating and yielding vehicle, and the lateral yielding vehicle, a collaborative reserved dynamic restricted area object is generated, and an executable trajectory for the vehicle is generated under the constraints of the dynamic restricted area index table. The vehicle's executable trajectory is converted into steering, braking, and driving commands, and the boundary expansion, activation window, and confidence score boundary of the dynamic restricted area object are calibrated based on the vehicle's actual operating data.

9. An electronic device, characterized in that, This includes the processor, memory, vehicle communication interface, and chassis control interface; The memory stores a computer program, which, when executed by the processor, causes the electronic device to perform the vehicle cooperative avoidance method based on dynamic restricted areas as described in claim 8. The vehicle communication interface is used to receive vehicle status, external target status, roadside coordination information, and communication delay information. The chassis control interface is used to output steering commands, braking commands, and drive commands generated based on the vehicle's executable trajectory.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program; When the computer program is executed by the processor, it implements the vehicle cooperative avoidance method based on dynamic restricted areas as described in claim 8.

Images