Vehicle control method and device for construction risk area and vehicle

Abstract

The application provides a control method and device for a vehicle in a construction risk area and the vehicle, and belongs to the field of vehicle driving control. The method comprises the following steps: generating a planned driving track of a target control vehicle; if the planned driving track of the target control vehicle meets a preset conflict condition with a predicted driving track, adding the target control vehicle to a path conflict vehicle list, and the conflict condition comprises driving track overlap and driving time overlap; determining a conflict priority of each vehicle based on a collision time margin, a lane changing probability, a prediction confidence of each vehicle in the path conflict vehicle list, and a relative position relationship between each vehicle and a construction obstacle; taking a vehicle with the highest conflict priority as a main avoidance object, and generating a vehicle control instruction for the target control vehicle based on driving prediction data of the main avoidance object. The above scheme can effectively improve the driving safety of a vehicle driver.

Description

Technical Field

[0001] This invention relates to the field of vehicle driving control, and more specifically to control methods, devices, and vehicles for vehicles in construction risk areas. Background Technology

[0002] Existing intelligent driving technologies in construction risk areas of urban expressways and highways primarily focus on identifying lane markings and avoiding stationary obstacles. The conventional approach of existing technologies is to acquire the current-moment characteristic data of surrounding vehicles, directly determine the existence of collision risk based on this current-moment characteristic data, and then generate vehicle control commands for the vehicle itself.

[0003] However, relying on currently available data for vehicle control inherently delays the response time, potentially causing the best intervention window to be missed and leading to dangerous accidents such as collisions. This is especially true in complex traffic flow areas during construction, where narrower lanes and chaotic markings increase the risk of accidents such as scraping and collisions caused by lane-changing maneuvers. Summary of the Invention

[0004] To address the aforementioned issues, this application provides a method for controlling vehicles within construction risk areas, aimed at improving driving safety for drivers in these areas. The technical solution is as follows: In a first aspect, the present invention provides a method for controlling vehicles within a construction risk area, comprising: When the target controlled vehicle is in the construction risk area, obtain the temporal feature matrix of the target monitored vehicle in the construction risk area; The driving prediction data of the target monitoring vehicle is determined based on the time-series feature matrix. The driving prediction data includes at least the lane change probability and the prediction confidence level corresponding to the lane change probability. Based on the driving prediction data, the predicted driving trajectory of the target monitoring vehicle is generated. If the planned driving trajectory of the target control vehicle and the predicted driving trajectory meet the preset conflict conditions, the target monitoring vehicle is added to the path conflict vehicle list. The conflict conditions include overlapping driving trajectories and overlapping driving times. Based on the collision time margin, lane change probability, prediction confidence, and relative positional relationship between each vehicle and construction obstacle in the list of vehicles in path conflict, the conflict priority of each vehicle is determined. The vehicle with the highest conflict priority is selected as the primary avoidance target, and vehicle control commands are generated for the target vehicle based on the driving prediction data of the primary avoidance target.

[0005] Combining the first aspect and the above implementation methods, in some possible implementation methods, the temporal feature matrix of target monitoring vehicles in the construction risk area is obtained, including: Acquire feature vector data of the target monitored vehicle; The feature vector data is stored in the target queue buffer corresponding to the target monitoring vehicle in chronological order, and the feature vector data is normalized. The normalized feature vector data in the target queue buffer is extracted and arranged in chronological order to obtain the temporal feature matrix of the target monitoring vehicle.

[0006] Combining the first aspect and the above implementation methods, in some possible implementation methods, the feature vector data of the target monitoring vehicle is obtained, including: Acquire the intent micro-features, kinematic features, and interaction scene features of the target monitoring vehicles in the construction risk area. The intent micro-features include the instantaneous speed difference of the wheel set and the turn signal status. The kinematic features include relative longitudinal distance, relative lateral distance, relative speed, and relative acceleration. The interaction scene features include time margin, lane line position, and remaining passable road width. Based on intent micro-features, kinematic features, and interaction scene features, feature vector data of target monitoring vehicles is generated.

[0007] Combining the first aspect and the above implementation methods, in some possible implementation methods, the driving prediction data of the target monitoring vehicle is determined based on the time-series feature matrix, including: The temporal feature matrix is ​​input into the driving intention prediction model for temporal information encoding and intention decoding to obtain the driving prediction data of the target monitored vehicle.

[0008] In combination with the first aspect and the above implementation methods, in some possible implementation methods, if the planned driving trajectory of the target controlled vehicle and the predicted driving trajectory meet the preset conflict conditions, the target monitored vehicle is added to the path conflict vehicle list, including: Based on the lane change probability and prediction confidence in the driving prediction data, target monitoring vehicles with valid lane change intentions are identified. If the target monitoring vehicle with valid lane change intention and the target control vehicle have overlapping driving trajectories and overlapping driving times in their predicted and planned driving trajectories, then the target monitoring vehicle with valid lane change intention will be added to the list of vehicles with path conflict.

[0009] Combining the first aspect and the above implementation methods, in some possible implementation methods, the vehicle with the highest conflict priority is designated as the primary avoidance target, and vehicle control commands for the target control vehicle are generated based on the driving prediction data of the primary avoidance target, including: Traverse the list of vehicles with path conflicts. If the main avoidance target meets the preset avoidance conditions, generate an active deceleration and yield command as a vehicle control command. The avoidance conditions are that the lane change probability is greater than a preset threshold, the turn signal is continuously illuminated, and the intention micro-features and lateral acceleration are enhanced in a consistent manner. If the main obstacle to avoid does not meet the avoidance conditions but meets the preset following conditions, a following deceleration command is generated as a vehicle control command. The following conditions are that the lane change probability is within a preset range and the lateral movement is in an oscillating mode.

[0010] Combining the first aspect and the above implementation methods, in some possible implementation methods, after designating the vehicle with the highest conflict priority as the primary avoidance target and generating vehicle control commands for the target control vehicle based on the driving prediction data of the primary avoidance target, the method further includes: The decision objective is obtained by parsing the vehicle control commands; Based on the decision-making objective, the dynamic constraints of the target-controlled vehicle, and the road boundary constraints, trajectory planning is performed to obtain a reference driving trajectory; The vehicle control quantity is obtained by performing rolling optimization calculations based on the reference driving trajectory and the current state of the target controlled vehicle.

[0011] Combining the first aspect and the above implementation methods, some possible implementation methods for obtaining vehicle control quantities also include: The system responds to vehicle control inputs to control the vehicle's movement and acquires real-time status information after the vehicle has started moving. The real-time status information is fed back to the step of obtaining the feature vector data of the target monitoring vehicles in the construction risk area.

[0012] Secondly, the present invention also provides a control device for vehicles in construction risk areas, comprising: The data acquisition unit is used to acquire the time-series feature matrix of the target monitored vehicle in the construction risk area when the target controlled vehicle is in the construction risk area; The data processing unit is used to determine the driving prediction data of the target monitoring vehicle based on the time-series feature matrix, and obtain the time-series feature matrix of the target monitoring vehicle. The vehicle list generation unit generates the predicted driving trajectory of the target monitoring vehicle based on the driving prediction data. If the planned driving trajectory of the target control vehicle and the predicted driving trajectory meet the preset conflict conditions, the target monitoring vehicle is added to the path conflict vehicle list. The conflict conditions include overlapping driving trajectories and overlapping driving times. The instruction generation unit is used to determine the conflict priority of each vehicle based on the collision time margin, lane change probability, prediction confidence, and relative positional relationship between each vehicle and the construction obstacle in the list of vehicles in path conflict; the vehicle with the highest conflict priority is taken as the primary avoidance target, and vehicle control instructions for the target control vehicle are generated based on the driving prediction data of the primary avoidance target.

[0013] Thirdly, the present invention also provides a vehicle, a vehicle data acquisition unit, a memory, and a processor, wherein... The data acquisition unit is used to acquire the time-series feature matrix of the target monitoring vehicle in the construction risk area when the target control vehicle is in the construction risk area; Memory, used to store executable program code; The processor is used to determine the driving prediction data of the target monitoring vehicle based on the time-series feature matrix; generate the planned driving trajectory of the target control vehicle based on the real-time vehicle operation data of the target control vehicle, the target driving position and the current road data, and generate the predicted driving trajectory of the target monitoring vehicle based on the driving prediction data. If the planned driving trajectory and the predicted driving trajectory of the target control vehicle meet the preset conflict conditions, the target monitoring vehicle is added to the path conflict vehicle list. The conflict conditions include overlapping driving trajectories and overlapping driving times. Based on the collision time margin, lane change probability, prediction confidence and the relative positional relationship between each vehicle and the construction obstacle in the path conflict vehicle list, the conflict priority of each vehicle is determined. The vehicle with the highest conflict priority is taken as the primary avoidance target, and vehicle control commands for the target control vehicle are generated based on the driving prediction data of the primary avoidance target.

[0014] Fourthly, the present invention also provides a computer-readable storage medium for storing a computer-readable program or instructions, which, when executed by a processor, can implement the steps of the control method for vehicles within a construction risk area described in any of the above implementations.

[0015] The beneficial effects of this invention are as follows: By acquiring the feature vector data of the target monitored vehicle in the construction risk area when the target controlled vehicle is in the construction risk area, the limitation of existing technologies that only rely on data at the current moment is overcome. By performing temporal arrangement and normalization processing on the feature vector data, a temporal feature matrix of the target monitored vehicle is obtained, effectively capturing the temporal change pattern of the state. The temporal feature matrix is ​​input into a preset driving intention prediction model to obtain the driving prediction data of the target monitored vehicle. The driving prediction data includes at least the lane change probability and the prediction confidence level corresponding to the lane change probability, thereby achieving accurate prediction of dynamic intentions. Based on the real-time vehicle operation data and driving prediction data of the target controlled vehicle, a list of vehicles with path conflicts is generated, and based on the conflict priority of each vehicle in the list, vehicle control commands are generated for the target controlled vehicle, avoiding the risk of vehicle collisions caused by misjudgment of intentions and improving the driving safety of vehicle drivers. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart illustrating the vehicle control method for construction risk areas provided by the present invention. Figure 2 For the present invention Figure 1 A schematic diagram of an embodiment of S102; Figure 3 For the present invention Figure 1 A schematic diagram of an embodiment of S104; Figure 4 For the present invention Figure 1 A schematic diagram of an embodiment of S105; Figure 5 For the present invention Figure 1 A schematic diagram of an embodiment of S105; Figure 6 This is a schematic diagram of the structure of the control device for vehicles in construction risk areas provided by the present invention; Figure 7 This is a structural schematic diagram of the vehicle provided by the present invention. Detailed Implementation

[0018] 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 a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0019] In the description of the embodiments of the present invention, unless otherwise stated, "multiple" means two or more. "And / or" describes the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can represent three situations: A exists alone, A and B exist simultaneously, and B exists alone.

[0020] The terms "first," "second," etc., used in the embodiments of this invention are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a technical feature defined with "first" or "second" may explicitly or implicitly include at least one of that feature.

[0021] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0022] In construction risk areas, vehicles in adjacent lanes frequently exhibit dynamic behaviors such as tentatively cutting in or forcibly changing lanes without using turn signals. Current technologies only acquire the current-moment feature data of surrounding vehicles, failing to perform temporal arrangement and normalization of the feature vector data to obtain a temporal feature matrix. Therefore, they cannot capture the historical trajectory changes of vehicle driving states. Furthermore, current technologies cannot predict lane-changing intentions in advance, nor can they generate vehicle control commands based on conflict priorities, resulting in rigid and passive decision-making by the target vehicle when facing multi-vehicle conflicts.

[0023] For example, when a vehicle slowly drifts into the lane for several consecutive frames but has not yet crossed the line and has not activated its turn signal, the existing technology determines that there is no risk of collision based solely on the current state of not crossing the line. However, if the vehicle suddenly turns sharply to cut in, it is very likely to cause the vehicle to brake suddenly or even cause a collision.

[0024] To address the aforementioned technical problems, this application provides a method for controlling vehicles within a construction risk area. The method is implemented by a device for controlling vehicles within a construction risk area or a vehicle equipped with such a device. The following provides a detailed description; however, it should be noted that the order of description in the following embodiments is not intended to limit the preferred order of the embodiments.

[0025] Please see Figure 1 , Figure 1 This is a flowchart illustrating a method for controlling vehicles within a construction risk area, as provided in an embodiment of this application. Figure 1 As shown, the method in this application embodiment may include the following steps S101-S105: S101, when the target control vehicle is in the construction risk area, obtain the temporal feature matrix of the target monitoring vehicle in the construction risk area.

[0026] In this embodiment, the target control vehicle refers to the vehicle itself that needs to make driving decisions and route planning and is driving in a construction risk area. The construction risk area refers to an area with potential collision risk due to lane encroachment, road narrowing, or changes in traffic flow caused by road construction. This area is dynamically delineated by a region of interest based on a high-precision map or visual recognition, including the lane where the target control vehicle is located, adjacent lanes, and the construction encroachment area. The target monitoring vehicle refers to surrounding vehicles located within the construction risk area that are identified, associated, and tracked by the sensors of the target control vehicle. The temporal feature matrix is ​​a two-dimensional matrix composed of normalized, fixed-length multi-dimensional feature vectors arranged in chronological order, used to represent a snapshot of the target monitoring vehicle's dynamic behavior history within a recent time window.

[0027] In one feasible implementation, millimeter-wave radar and vehicle-mounted cameras are pre-deployed on the construction site and the controlled vehicles, with the vehicles also equipped with CAN bus communication modules. Millimeter-wave radar collects radar point cloud data within the construction area, vehicle-mounted cameras capture continuous road image frames, and the CAN bus reads the real-time operating status data of the controlled vehicles. After fusing this data, target identification, data association, and continuous tracking of monitoring vehicles within a defined area of ​​interest are performed, assigning a unique tracking identifier to each identified target vehicle.

[0028] Parallel feature extraction is performed on target monitoring vehicles in the target tracking list. The features are divided into three complementary dimensions: first, intent micro-features, which utilize the micro-Doppler effect of millimeter-wave radar to calculate the instantaneous speed difference between the left and right wheel sets of the target monitoring vehicle and visually identify the turn signal status; second, kinematic features, which acquire relative longitudinal distance, relative lateral distance, relative velocity, and relative acceleration; and third, interaction scene features, which calculate time margin, determine lane line positions, and estimate the remaining passable road width in construction risk areas. These three types of features are then integrated to output structured feature vector data.

[0029] Furthermore, a fixed-length First-In-First-Out (FIFO) queue buffer is maintained in memory for each stably tracked target vehicle using a unique tracking identifier. Whenever a new feature vector data packet is output, it is added to the end of the corresponding target queue buffer in chronological order. If the data exceeds a preset window size T, the oldest data is automatically removed. Before storage, the values ​​of each dimension of the feature vector data are normalized (e.g., using x_norm=(x-x_min) / (x_max-x_min)) to eliminate dimensional differences. During intent prediction, the normalized data in the buffer is extracted in chronological order and arranged into a temporal feature matrix of dimensions [T,D], where T is the time window size and D is the total number of dimensions of the feature vectors.

[0030] S102, determine the driving prediction data of the target monitored vehicle based on the time series feature matrix. The driving prediction data includes at least the lane change probability and the prediction confidence level corresponding to the lane change probability.

[0031] In this embodiment of the application, driving prediction data refers to structured data representing the prediction result, which includes at least the probability of the target monitored vehicle changing lanes to the left, the probability of changing lanes to the right, and the overall confidence level of the prediction.

[0032] In one feasible implementation, a temporal feature matrix of dimension [T,D] is input into a pre-trained driving intention prediction model (such as a Long Short-Term Memory network, LSTM) deployed on a cloud or vehicle-side computing platform. As the data flows through the LSTM layer, the network selectively remembers long-term important information (such as turning on the turn signal 1.5 seconds ago) and forgets short-term irrelevant fluctuations, thus learning typical lane-changing behavior patterns such as "a continuously increasing left wheel speed difference, accompanied by the left turn signal illuminating, followed by a stable leftward lateral speed." After temporal information encoding and intention decoding, a driving prediction data vector representing the prediction result is output. For example, the vector form is V=[PL, PR, C], where PL and PR are the probabilities of changing lanes to the left and right within a preset time (such as 2 seconds), respectively, and C is the prediction confidence level determined based on feature data quality, sequence integrity, or the model's internal state entropy value.

[0033] S103 generates the planned driving trajectory of the target control vehicle based on the real-time vehicle operation data of the target control vehicle, the target driving position and the current road data, and generates the predicted driving trajectory of the target monitoring vehicle based on the driving prediction data.

[0034] In this embodiment, real-time vehicle operation data includes the target vehicle's current speed, acceleration, heading angle, etc.; the target driving position includes the downstream waypoint specified by the global navigation path or the expected lane center; road data includes the lane topology, cone positions, and construction encroachment area boundaries extracted from high-precision maps or visual perception. Determining whether there is a conflict risk between two vehicles requires simultaneously satisfying the conditions of overlapping driving trajectories (i.e., spatial overlap) and overlapping driving times (i.e., temporal overlap). For the determination of temporal overlap, no safety threshold is set; as long as the two vehicles have an intersection of any duration on the time axis, it is determined that there is a conflict risk.

[0035] In one feasible implementation, real-time vehicle operation data, target driving position, and road data of the target controlled vehicle are acquired. The real-time vehicle operation data includes the target controlled vehicle's current speed, acceleration, heading angle, etc. The target driving position includes the downstream waypoint specified by the global navigation path or the lane center to be maintained. The road data includes the lane topology of the construction area, the location of cones, and the boundary of the construction encroachment area extracted by high-precision map or visual perception.

[0036] The trajectory planner based on the target-controlled vehicle integrates the above real-time vehicle operation data, target driving position, and road data. Under the premise of satisfying vehicle dynamics constraints and road boundary constraints, it generates a smooth planned driving trajectory using quadratic programming or spline curve algorithms. Simultaneously, considering potential path conflict risks, it extrapolates the target vehicle's driving corridor within a preset future time window based on the target vehicle's current kinematic state (position, speed) and the driving prediction data (lane change probability and intended direction) determined by S102. For example, if PL > 0.7 and the confidence level C is high, the predicted driving trajectory is generated according to the left lane change curve; if there is no clear lane change intention, the predicted driving trajectory is generated by extrapolating the longitudinal straight line of the current lane.

[0037] Within a preset time window, the planned and predicted driving trajectories are discretized and sampled at fixed time steps. If a spatial-first, temporal order is adopted, the geometrical envelopes of the target control vehicle and the target monitoring vehicle at the same time are first calculated to see if there is spatial overlap. If so, the arrival times of the two vehicles in the spatially overlapping area are further checked. If a temporal-first, spatial order is adopted, overlapping segments of the two vehicles' driving trajectories on the time axis are first found, and then the spatial positions of the two vehicles within that time segment are checked to see if there is overlap. If both spatial and temporal overlap exists, the conflict condition is met, and the target monitoring vehicle is added to the list of vehicles with path conflicts.

[0038] S105. Based on the collision time margin, lane change probability, prediction confidence, and relative positional relationship between each vehicle and the construction obstacle in the list of vehicles in path conflict, calculate the conflict priority of each vehicle.

[0039] In this embodiment, the determination of conflict priority does not rely on a single static indicator, but rather comprehensively assesses the urgency of the collision, the clarity of intent, and the environmental pressure, employing a three-level progressive priority framework for dynamic arbitration. The collision time margin represents the estimated time required for the target vehicle and the target monitoring vehicle to reach the overlapping area of ​​their trajectories under the current motion state. A smaller collision time margin indicates a shorter time for the two vehicles to approach the conflict point, leaving less time for the target vehicle to take evasive or deceleration responses, meaning a more imminent collision risk. The relative positional relationship with construction obstacles reflects the retreat space of the target monitoring vehicle; the closer it is to the construction obstacle, the lower the likelihood of it maintaining its original lane, and the higher the mandatory and irreversible nature of its lane change into the target vehicle's lane.

[0040] In one feasible implementation, the list of vehicles with path conflicts is traversed, and the conflict priority of each vehicle in the list is determined according to the following three-level progressive logic: The first priority (avoidance principle) targets target vehicles whose signals are undisputed and whose actions are decisive. The criteria for judgment are a lane-change probability greater than a preset value and a high prediction confidence level, along with a consistent increase in the vehicle's intent micro-features and lateral acceleration. Furthermore, if the lateral distance between the target vehicle and the construction obstacle is extremely small, indicating an irreversible intrusion tendency due to environmental pressure, it is also included in this category. When multiple vehicles meet the first priority criteria, their collision time margins are further compared, and the vehicle with the smaller collision time margin is given priority for avoidance. For conflict parties determined to be of first priority, the target vehicle must actively avoid them.

[0041] The second priority (following principle) involves vehicles that meet the first priority criteria if none in the conflict list do. These vehicles exhibit "clear intent but hesitant behavior." They have a lane-changing probability between 0.5 and 0.8, moderate prediction confidence, intermittent turn signal activation, and oscillating lateral movement. They also maintain a safe distance from construction obstacles and are not yet under significant spatial pressure. For vehicles classified as second priority, the target vehicle adopts a following strategy.

[0042] The third priority (proactive principle) occurs when the conflicting vehicle is primarily a "hesitant vehicle" from the second priority category, and the system assesses the relative position of the target vehicle and the construction obstacle, determining that the target vehicle possesses a lane-changing advantage (i.e., its lane-changing conditions are superior); or when the target vehicle has a strong need for continuous lane changes and its signals are clear, the system determines that the target vehicle should adopt a proactive strategy. In this case, the conflict priority of the other vehicle is relatively reduced, and the target vehicle prioritizes executing its own driving intentions.

[0043] S105, the vehicle with the highest conflict priority is selected as the primary avoidance target, and vehicle control instructions are generated for the target vehicle by combining the driving prediction data of the primary avoidance target.

[0044] In this embodiment, the vehicle control command refers to an executable command that specifies the combined behavioral target of the target vehicle, namely its longitudinal target acceleration and lateral desired position or curvature. The primary avoidance target refers to the vehicle that poses the greatest threat to the target vehicle and requires primary coordinated response.

[0045] In one feasible implementation, the arbitrator selects the vehicle with the highest conflict priority from the list of conflicting vehicles as the primary avoidance target, and directly maps and generates the corresponding vehicle control command according to the priority category of the primary avoidance target determined in S104: If the object to be avoided belongs to the first priority (avoidance principle), it means that its actions are decisive or that it is under pressure from the construction obstacle and has an irreversible tendency to intrude. The arbitrator generates a vehicle control command of "actively slowing down and giving way, and slightly shifting laterally to indicate". If the primary obstacle to avoid is of the second priority (following principle), it indicates that its behavior is hesitant and there is a potential tendency to intrude. The arbitrator generates a vehicle control command to "maintain the current lane, moderately reduce speed, place behind the primary obstacle to avoid, and prepare to continue its driving trajectory". If the primary obstacle to avoid is of the third priority (active principle), it means that the target vehicle has an advantage in lane changing space or a strong need for continuous lane changing. The arbitrator generates a vehicle control command to "execute the original driving plan and apply a slight but clear lateral movement trend to convey the intention".

[0046] In summary, this application overcomes the limitation of existing technologies that rely solely on current-time data by acquiring the feature vector data of the target monitoring vehicle within the construction risk area when the target controlled vehicle is in that area. By performing temporal arrangement and normalization on the feature vector data, a temporal feature matrix of the target monitoring vehicle is obtained, effectively capturing the temporal change patterns of its state. The temporal feature matrix is ​​input into a preset driving intention prediction model to obtain the driving prediction data of the target monitoring vehicle. This driving prediction data includes at least the lane change probability and the corresponding prediction confidence level, thereby achieving accurate prediction of dynamic intentions. Based on the real-time vehicle operation data and driving prediction data of the target controlled vehicle, a list of vehicles with path conflicts is generated. Based on the conflict priority of each vehicle in the list, vehicle control commands are generated for the target controlled vehicle, avoiding the risk of vehicle collisions due to misjudgment of intentions and improving the driving safety of drivers.

[0047] In one feasible implementation, after generating vehicle control commands based on the driving prediction data of the primary avoidance object, if a change in conflict priority is detected, the vehicle control commands for the target control vehicle are immediately regenerated using the prediction data of the avoidance object with the highest conflict priority in the latest data.

[0048] Specifically, when a change in conflict priority is detected (e.g., the appearance of a new conflict vehicle with a more decisive intent or a smaller collision time margin, or the original vehicle being avoided being downgraded in priority due to hesitation), the instruction replanning logic will be immediately triggered. The newly determined highest priority vehicle to avoid and its driving prediction data will be used as input, and the decision strategy will be remapped in combination with its corresponding priority category. This will overwrite and replace the currently executing vehicle control instructions, ensuring that the target vehicle always responds to the most threatening conflict vehicle in the most timely manner.

[0049] Please see Figure 2 , Figure 2 For the present invention Figure 1 A schematic diagram of an embodiment of S102. (See attached diagram.) Figure 2 As shown, the method in this application embodiment may include the following steps S201-S204: S201, acquire the intent micro-features, kinematic features, and interaction scene features of the target monitoring vehicles in the construction risk area.

[0050] In this embodiment, the intent micro-features refer to feature data reflecting the subtle operational tendencies of the target monitoring vehicle, including the instantaneous speed difference between the left and right wheel sets and the turn signal status. Kinematic features refer to feature data describing the spatial and motional state of the target monitoring vehicle relative to the target control vehicle, including relative longitudinal distance, relative lateral distance, relative speed, and relative acceleration. Interaction scenario features refer to feature data characterizing the interactive constraints between the target monitoring vehicle, the target control vehicle, and the construction environment, including time margin, lane line position, and remaining passable road width.

[0051] In one feasible implementation, meaningful features are matched to the target monitoring vehicle within the construction risk area from sensor information. For intent micro-features, the instantaneous speed difference between the left and right wheel sets of the target monitoring vehicle is calculated using the micro-Doppler effect of millimeter-wave radar, while the turn signal status of the target monitoring vehicle is identified by visual sub-analysis of camera image frames.

[0052] For kinematic characteristics, the relative longitudinal distance, relative lateral distance, relative speed, and relative acceleration of the target monitoring vehicle relative to the target control vehicle are obtained from radar tracking information.

[0053] For interactive scenario characteristics, the time margin between the target control vehicle and the target monitoring vehicle is calculated, the lane position of the target monitoring vehicle is determined, and the remaining passable road width in the construction risk area is estimated. The time margin refers to the remaining time margin for the target monitoring vehicle and the target control vehicle to interact or collide at the current relative speed and distance, which is used to measure the urgency of the interaction.

[0054] S202 generates feature vector data of the target monitoring vehicle based on intent micro-feature extraction, kinematic feature extraction, and interaction scene features.

[0055] In this embodiment of the application, feature vector data refers to structured vector data that combines intentional micro-features, kinematic features, and interaction scene features in a preset dimensional order.

[0056] In one feasible implementation, for the target monitoring vehicle in the target tracking list, the intent micro-features (including the instantaneous speed difference between the left and right wheel sets and the turn signal status), kinematic features (including relative longitudinal distance, relative lateral distance, relative speed and relative acceleration), and interaction scene features (including time margin, lane line position and remaining passable road width) extracted at the same time are spliced ​​and aligned, and a fixed vector position is assigned to each feature dimension, thereby generating time-synchronized structured feature vector data corresponding to the unique tracking identifier of the target monitoring vehicle.

[0057] S203, the feature vector data is stored in the target queue buffer corresponding to the target monitoring vehicle in chronological order, and the feature vector data is normalized.

[0058] In this embodiment of the application, the target queue buffer refers to a fixed-length first-in-first-out (FIFO) queue buffer maintained in memory for the target monitoring vehicle, used to store discrete feature vector data arranged in chronological order.

[0059] In one feasible implementation, a target queue buffer is maintained in memory for each stably tracked target vehicle using a unique tracking identifier. Whenever a new feature vector data packet is output for a target vehicle, it is added to the end of the target queue buffer corresponding to that vehicle in chronological order. If the length of the target queue buffer exceeds a preset window size T (T is 20, corresponding to the past 2 seconds of data, with a processing cycle of 100 milliseconds), the oldest feature vector data packet is automatically removed, ensuring that the target queue buffer always retains the complete behavioral history of the target vehicle for the most recent T cycles. Before being stored in the target queue buffer, the values ​​of each dimension of the feature vector data are normalized to eliminate dimensional differences between different features and accelerate model convergence.

[0060] S204. Extract and arrange the normalized feature vector data in the target queue buffer in chronological order to obtain the temporal feature matrix of the target monitoring vehicle.

[0061] In the embodiments of this application, the temporal feature matrix refers to a two-dimensional matrix data structure obtained by arranging the normalized feature vector data in the target queue buffer according to time steps.

[0062] In one feasible implementation, all normalized feature vector data in the target queue buffer are extracted in chronological order and arranged into a temporal feature matrix of dimension [T, D], where T is the size of the time window and D is the total dimension of the feature vectors. This temporal feature matrix is ​​a snapshot of the target monitored vehicle's behavior within the most recent T time units, depicting dynamic processes such as instantaneous wheel speed difference, turn signal status, and relative lateral speed.

[0063] In summary, this application obtains multi-dimensional complementary features and uses a target queue buffer to perform temporal storage and normalization processing on the feature vector data, thereby extracting a temporal feature matrix. This enables an accurate characterization of the historical trend of the target monitored vehicle from its static instantaneous state to its dynamic behavior, eliminating the influence of different feature dimensions on the model and providing a standardized and structured input guarantee for the subsequent extraction of time dependencies in the temporal model.

[0064] Please see Figure 3 , Figure 3 For the present invention Figure 1 A schematic diagram of an embodiment of S104. (See attached diagram.) Figure 3 As shown, the method in this application embodiment may include the following steps S301-S302: S301, based on the lane change probability and prediction confidence in the driving prediction data, identify target monitoring vehicles with valid lane change intentions.

[0065] In this embodiment, lane change probability refers to the numerical value of the likelihood that the target monitored vehicle will change lanes to the left or right within a preset time interval in the future. Prediction confidence refers to an indicator reflecting the degree of certainty that the driving intention prediction model is about the current lane change probability prediction result. Valid lane change intention refers to a deterministic lane change tendency where the lane change probability exceeds a probability threshold and the prediction confidence exceeds a confidence threshold.

[0066] In one feasible implementation, the intent of each target monitored vehicle in the driving prediction data is binarized for determination. A probability threshold P_th, preferably 0.7, and a confidence threshold C_th, preferably 0.6, are set.

[0067] If the probability of a target vehicle changing lanes to the left (PL) or to the right (PR) exceeds the probability threshold P_th, and the prediction confidence C is greater than the confidence threshold C_th, then the target vehicle is determined to have a clear and valid lane-changing intention, and the intended direction (left or right) and the probability of lane change are recorded. Conversely, if the lane-changing probability does not exceed the probability threshold P_th or the prediction confidence does not exceed the confidence threshold C_th, then the target vehicle is determined not to have a valid lane-changing intention and is filtered out.

[0068] For example, vehicle A is predicted to have a left lane change probability PL=0.76 and a prediction confidence level C=0.68. Since 0.76>0.7 and 0.68>0.6, both threshold conditions are met, so vehicle A is determined to have a valid left lane change intention, and the intended direction is recorded as left with a lane change probability of 0.76. Vehicle B has a right lane change probability PR=0.72 and a prediction confidence level C=0.55. Although the lane change probability exceeds the threshold, the confidence level 0.55 is less than 0.6, failing to meet all the judgment conditions, and is directly filtered out. Vehicle C has a left lane change probability PL=0.63 and a confidence level C=0.71. Since the lane change probability does not reach 0.7, it is also determined to have no valid lane change intention and is removed.

[0069] In one feasible implementation, the real-time vehicle operation data of the target control vehicle (current position, current lane, and the desired path planned by the decision-maker) is compared with the status of all target monitoring vehicles marked as having valid lane-changing intentions to extrapolate and compare their future spatiotemporal trajectories.

[0070] Based on the current motion state (speed, position) of the target control vehicle and the target monitoring vehicle with valid lane change intention, and combined with the intention direction of the target monitoring vehicle with valid lane change intention, the trajectory corridor within a preset time interval is simply predicted, thereby obtaining the trajectory deduction result containing the spatial range of their respective trajectories and corresponding time information.

[0071] S302, if the trajectory projection result indicates that the target control vehicle and the target monitoring vehicle with a valid lane change intention have spatial and temporal overlap in their future spatiotemporal trajectories, then a list of vehicles with path conflicts is generated based on the target monitoring vehicle with a valid lane change intention.

[0072] In this embodiment, the future spatiotemporal trajectory refers to the predicted path corridor that the vehicle will traverse within a preset time interval in the future, along with the corresponding time information. Spatial overlap refers to the overlap between the planned trajectory corridor of the target controlled vehicle and the predicted trajectory corridor of the target monitored vehicle on the road plane. Temporal overlap refers to the intersection of the time windows in which the target controlled vehicle and the target monitored vehicle are expected to arrive at the spatially overlapping area. Determining whether there is a conflict risk between the two vehicles requires both spatial and temporal overlap conditions to be met simultaneously; neither can be omitted. For the determination of temporal overlap, no safety threshold is set; as long as the two vehicles have an intersection of any duration on the time axis, it is determined that there is a conflict risk.

[0073] In one feasible implementation, it is determined whether the trajectory extrapolation results simultaneously satisfy the two core conditions of spatial overlap and temporal overlap. The determination order can be either to first determine spatial overlap and then check for temporal overlap, or to first determine temporal misalignment and then check for spatial overlap; both methods are equivalent. Specifically, if the spatial-to-temporal order is adopted, it is first calculated whether there is an overlapping area between the trajectory corridors of the two vehicles on the road. If so, it is further checked whether the time windows of the two vehicles reaching the overlapping area intersect. If the temporal-to-spatial order is adopted, it is first found that the two vehicles overlap in time segments, and then it is checked whether there is an intersection in the spatial positions of the two vehicles within those time segments. If both spatial and temporal overlap conditions are satisfied, it is determined that there is a lane-change intention conflict between the target controlled vehicle and the target monitoring vehicle with a valid lane-change intention, a conflict status flag is generated as true, and the target monitoring vehicle with a valid lane-change intention is added to the path conflict vehicle list. This path conflict vehicle list records the unique tracking identifier of all target monitoring vehicles that conflict with the target controlled vehicle, as well as detailed information of each conflicting target, including intention direction, prediction probability, prediction confidence, relative position, and expected spatiotemporal point of conflict.

[0074] In summary, this application uses a dual threshold discrimination method based on lane change probability and prediction confidence to screen out vehicles with genuine lane change intentions. It also performs a dual overlap test based on trajectory extrapolation in both space and time, which can accurately identify vehicles that may conflict in the future path contention. This eliminates the interference of false detections and non-threatening vehicles, providing an accurate conflict list for subsequent dynamic priority arbitration and improving the safety and effectiveness of decision-making.

[0075] In one feasible implementation, when generating vehicle control instructions for the target control vehicle based on the conflict priority of each vehicle in the path conflict vehicle list, the following is also specifically executed: Traverse the list of vehicles with path conflicts. If the main avoidance target meets the preset avoidance conditions, generate an active deceleration and yield command as a vehicle control command. The avoidance conditions are that the lane change probability is greater than a preset threshold, the turn signal is continuously illuminated, and the intention micro-features and lateral acceleration are enhanced in a consistent manner. In this application embodiment, the avoidance condition refers to the first priority judgment criterion used to determine whether the conflicting vehicle, i.e., the primary avoidance target, has an indisputable lane-change signal and acts decisively. The active deceleration and yielding instruction refers to a coordinated decision-making instruction that instructs the target vehicle to actively reduce its speed and make room for the conflicting vehicle, while simultaneously generating a slight lateral offset to indicate its yielding intention. The preset threshold is a critical value used to measure whether the probability of the conflicting vehicle changing lanes has reached an extremely high degree of certainty; a preferred value is 0.9. The turn signal refers to the indicator light on the target monitoring vehicle used to convey its turning intention to other traffic participants. Lateral acceleration refers to the acceleration component of the target monitoring vehicle perpendicular to its longitudinal direction of travel, reflecting the dynamic execution force of its lateral lane-change.

[0076] In one feasible implementation, the arbitrator traverses the list of conflicting vehicles, searching for vehicles whose signals are undisputed and whose actions are decisive. The arbitrator's specific judgment criteria are: the probability of the conflicting vehicle changing lanes is greater than a preset threshold, the turn signal of the conflicting vehicle is continuously illuminated, and the intentional micro-features of the conflicting vehicle (i.e., the instantaneous speed difference between the left and right wheel sets is continuously increasing) and the lateral acceleration show a consistent increasing trend.

[0077] If multiple conflicting vehicles in the path conflict vehicle list simultaneously meet the above avoidance conditions, further evaluation will be conducted, prioritizing avoidance of the conflicting vehicle with a higher relative speed or less time margin. For the conflicting party determined to meet the avoidance conditions, i.e., the first priority, the arbitrator generates an active deceleration and yield instruction for the target vehicle, controls the target vehicle to actively decelerate and yield, and generates a slight lateral offset to convey the yielding indication to the conflicting vehicle.

[0078] If the main obstacle to avoid does not meet the avoidance conditions but meets the preset following conditions, a following deceleration command is generated as a vehicle control command. The following conditions are that the lane change probability is within a preset range and the lateral movement is in an oscillating mode.

[0079] In this embodiment, the following condition refers to a second priority criterion used to determine whether a conflicting vehicle has a clear intention to change lanes but exhibits hesitant behavior. The following deceleration instruction refers to a collaborative decision-making instruction that instructs the target vehicle to maintain its current lane, moderately and gradually reduce its speed to position itself to the side and rear of the conflicting vehicle, and prepare to continue its driving trajectory. The preset interval refers to a numerical range used to measure the probability of the conflicting vehicle changing lanes with moderate certainty and exhibiting hesitant behavior; a preferred value is between 0.5 and 0.8. The oscillation mode refers to the cyclical fluctuation state of the conflicting vehicle's lateral movement, exhibiting "advance-retreat-advance," reflecting its hesitation in its lane-changing intention.

[0080] In one feasible implementation, if no conflicting vehicle meeting the avoidance criteria is found in the list of conflicting vehicles, the arbitration process proceeds to the second priority: the following principle. This involves judging conflicting vehicles with clear intent but hesitant behavior. These vehicles have a lane-changing probability within a preset range (0.5 to 0.8), their turn signals are intermittent, and their lateral movement exhibits an oscillating "forward-backward-forward" pattern. For conflicting vehicles meeting the following criteria, the arbitrator generates a following deceleration command for the target vehicle, controls it to maintain its current lane, moderately and gently decelerates, positions the target vehicle to the side and rear of the conflicting vehicle, and prepares to continue the conflicting vehicle's trajectory, avoiding the danger of a forced lane-changing attempt during its hesitation period.

[0081] In one feasible implementation, the third priority proactive principle is activated in two situations: first, the main conflicting party is a hesitant vehicle that meets the following conditions, and the target vehicle's lane-changing conditions meet preset superior conditions (e.g., no vehicles to the side and rear of the target vehicle, and sufficient remaining passable road width in adjacent lanes, i.e., superior lane-changing conditions); second, the target vehicle itself has a very strong need for continuous lane changes and its steering signals are clear. When the above proactive conditions are met, the arbitrator determines that the target vehicle has adopted a proactive strategy, generates an proactive execution instruction as a vehicle control instruction, controls the target vehicle to firmly execute the original plan, and applies a slight, clear lateral movement trend to convey the proactive intention to the conflicting vehicles. Throughout the arbitration process, the vehicle feature flow is continuously monitored, the priority is refreshed in milliseconds, and finally, a vehicle control instruction is output that clearly defines the joint behavioral target of the target vehicle's longitudinal target acceleration and lateral desired position or curvature.

[0082] In summary, this application traverses the list of vehicles with path conflicts and performs dynamic arbitration based on three progressively higher priorities according to avoidance, following, and proactive conditions. This simulates the deep game and multi-factor weighted evaluation process of human drivers, generating corresponding proactive deceleration and yielding instructions, following deceleration instructions, or proactive execution instructions for different conflict scenarios. This enables the target vehicle to flexibly and reasonably adopt avoidance, following, or proactive strategies when facing path contention, avoiding the rigidity or danger of single decision-making and improving the safety and traffic efficiency of multi-vehicle interaction in construction risk areas.

[0083] Please see Figure 4 , Figure 4 For the present invention Figure 1 A schematic diagram of an embodiment of S105. (See attached diagram.) Figure 4 As shown, the method in this application embodiment may include the following steps S401-S403: S401 parses the vehicle control commands to obtain the decision objective.

[0084] In the embodiments of this application, the decision objective refers to the specific and quantifiable control task indicators that are transformed from the high-level collaborative strategy.

[0085] In one feasible implementation, the vehicle behavior decision-maker receives and parses vehicle control commands (such as proactive deceleration and yield commands, follow-deceleration commands, or proactive execution commands), converting them into specific, quantifiable decision objectives. The vehicle behavior decision-maker can be deployed on a cloud server or a local vehicle terminal.

[0086] For example, when the vehicle control command is a follow deceleration command, the parsed decision objective is "to increase the following distance between the target controlled vehicle and the vehicle in front from 1.2 seconds to 1.8 seconds within 3 seconds, while maintaining the vehicle's center line in the current lane"; when the vehicle control command is an active execution command, the parsed decision objective is "to move laterally to the center line of the left lane within 2 seconds, while maintaining the current driving speed".

[0087] S402, based on the decision objective, the dynamic constraints of the target control vehicle, and the road boundary constraints, trajectory planning is performed to obtain a reference driving trajectory.

[0088] In this embodiment, dynamic constraints refer to the kinematic boundaries of the target controlled vehicle determined by its physical structure and tire grip, such as maximum acceleration, maximum braking deceleration, and maximum steering wheel angle. Road boundary constraints refer to the physical boundaries of the drivable area constituted by lane lines, construction encroachment areas, and road edges within the construction risk area. The reference driving trajectory refers to the optimal motion trajectory that meets the decision-making objective and various constraints, consisting of a series of discrete path points and their corresponding time, speed, and acceleration information.

[0089] In one feasible implementation, the trajectory planning module receives the decision objective output by the vehicle behavior decision-maker. This module can be deployed on a cloud server or the vehicle's local terminal. It comprehensively considers the dynamic constraints of the target vehicle (such as maximum longitudinal acceleration and maximum lateral acceleration), road boundary constraints (such as the remaining passable road width in construction areas and lane line positions), and obstacle constraints to search for or optimize an optimal trajectory in a spatiotemporal joint space. This optimal trajectory consists of a series of discrete path points and corresponding time, speed, and acceleration information, thus obtaining a smooth and safe reference driving trajectory that satisfies the decision objective.

[0090] S403 performs rolling optimization calculations on the reference driving trajectory and the current state of the target controlled vehicle to obtain the vehicle control quantity.

[0091] In this embodiment, the current state refers to the dynamic state information of the target controlled vehicle at the current moment, such as its actual position, actual speed, actual acceleration, and actual heading angle. Rolling optimization calculation refers to calculating the control sequence within the future finite time domain by solving the optimal control problem within each control cycle, based on the deviation between the current state and the reference driving trajectory, and executing the optimization method for the control quantity of the current cycle only. Vehicle control quantities refer to the physical quantities sent to the vehicle's underlying actuators for directly controlling the vehicle's motion, including the target drive torque, target braking pressure, and target steering wheel angle.

[0092] In one feasible implementation, the vehicle motion controller of the target controlled vehicle compares the reference driving trajectory output by the trajectory planning module with the current state of the target controlled vehicle, and calculates the position deviation, speed deviation, and heading deviation between the current state and the desired point on the reference driving trajectory. Through rolling optimization calculations, the optimal control quantity required to achieve reference driving trajectory tracking is determined. This vehicle control quantity specifically includes direct control commands sent to the vehicle's underlying actuators, such as target drive torque, target acceleration, target braking pressure, and target steering wheel angle.

[0093] In summary, this application obtains a quantitative decision target by parsing vehicle control commands, and obtains a reference driving trajectory by integrating dynamic constraints and road boundary constraints. Finally, the reference driving trajectory is transformed into a precise vehicle control quantity through rolling optimization calculation, realizing a smooth transition from high-level collaborative decision-making to low-level execution mechanism. This ensures that the movement of the target controlled vehicle satisfies both the safety strategy of collaborative arbitration and conforms to the vehicle's physical limits and road constraints, guaranteeing the accuracy and smoothness of trajectory execution.

[0094] Please see Figure 5 , Figure 5 For the present invention Figure 1 A schematic diagram of an embodiment of S105. (See attached diagram.) Figure 5 As shown, the method in this embodiment may include the following steps S501-S502: S501 responds to vehicle control variables to control the target vehicle's movement and obtains real-time status information after the target vehicle has started moving.

[0095] In this embodiment of the application, real-time state information refers to the set of data on the kinematic state of the target controlled vehicle and its relative relationship with the surrounding environment after the vehicle responds to the vehicle control quantity and performs actual motion.

[0096] In one feasible implementation, the actuator of the target controlled vehicle responds to the vehicle control quantities (such as target drive torque, target braking pressure, and target steering wheel angle) issued in step S403, generating actual vehicle motion so that the target controlled vehicle travels along the planned reference driving trajectory. This motion changes the relative state between the target controlled vehicle and the construction risk area, especially with surrounding target monitoring vehicles. Real-time status information such as the new relative distance and relative speed after the target controlled vehicle's movement is acquired, along with the latest environmental data collected by sensors, to form complete real-time status information.

[0097] S502, the step of feeding back real-time status information to the step of obtaining feature vector data of target monitoring vehicles in the construction risk area.

[0098] In this embodiment of the application, the step of feeding back real-time status information to the feature vector data of target monitoring vehicles in the construction risk area refers to the formation of a closed-loop control mechanism, which re-inputs the results of environmental and status changes after execution into the perception stage to update subsequent predictions and decisions.

[0099] In one feasible implementation, the acquired real-time status information is used as real-time feedback and sent back to the target recognition and tracking module in step S101 as the input benchmark for the spatiotemporal synchronization and fusion of sensor data in the next cycle. This feedback loop enables the entire "perception-prediction-coordination-planning-control" chain to operate continuously and in real time. It can dynamically adjust subsequent feature extraction, intent prediction, and collaborative decision-making based on the execution results of vehicle control quantities and the dynamically changing construction risk area environment, forming an autonomously evolving collaborative driving closed loop.

[0100] In summary, this application controls vehicle movement by responding to vehicle control variables and obtaining real-time status information after movement, which is then fed back to the perception step. This constructs a complete closed-loop control architecture, enabling real-time correction of perception and prediction inputs based on execution results and dynamic changes in the external environment. This ensures that the entire cooperative driving system can continuously adapt to complex and ever-changing interactive scenarios in construction risk areas, thereby improving its robustness and autonomous evolution capabilities.

[0101] In one feasible implementation, after the vehicle control command is issued to the target vehicle, the real-time driving behavior of the driver of the target vehicle is monitored in real time at the same time through the steering wheel angle sensor, accelerator pedal displacement sensor and brake pedal pressure sensor of the target vehicle. The actual steering wheel operation torque, accelerator pedal opening and brake pedal opening applied by the driver are extracted as actual operation features. At the same time, the currently issued vehicle control command is analyzed to obtain the target driving behavior corresponding to the vehicle control command, that is, the expected operation trend represented by the target steering wheel angle, target drive torque and target braking pressure.

[0102] Subsequently, the actual operational characteristics are compared with the expected operational trends to determine whether the direction of their actions are the same and whether the deviation in operational force is within the preset safety tolerance range. If the comparison result indicates that the real-time driving behavior is consistent with the target driving behavior, it is determined that the driver's current operational intention and decision overlap. To avoid control conflicts during human-machine co-driving, control is immediately allocated to the driver first, with the real-time driving behavior of the target vehicle driver as the active execution source. The automatic execution output of the current vehicle control command is suppressed or terminated, and only the underlying safety boundary monitoring is retained. If the comparison result indicates that they are inconsistent, the original vehicle control command continues to be executed, thereby achieving smooth coordination and conflict-free transition of human-machine control authority.

[0103] The following will combine Figure 6 This application provides a detailed description of the control device for vehicles within construction risk areas provided in its embodiments. It should be noted that... Figure 6 The control device 600 for vehicles within the construction risk area is used to perform the functions described in this application. Figures 1-5 The methods shown in the embodiments are for illustrative purposes only, illustrating the parts relevant to the embodiments of this application. For specific technical details not disclosed, please refer to this application. Figures 1-5 In the illustrated embodiment, the control device 600 for vehicles within the construction risk area may include a data acquisition unit 601, a data processing unit 602, a vehicle list generation unit 603, and an instruction generation unit 604, as detailed below: The data acquisition unit 601 is used to acquire the time-series feature matrix of the target monitoring vehicle in the construction risk area when the target control vehicle is in the construction risk area; The data processing unit 602 is used to determine the driving prediction data of the target monitoring vehicle based on the time series feature matrix, and obtain the time series feature matrix of the target monitoring vehicle. The vehicle list generation unit 603 is used to generate the predicted driving trajectory of the target monitoring vehicle based on the driving prediction data. If the planned driving trajectory of the target control vehicle and the predicted driving trajectory meet the preset conflict conditions, the target monitoring vehicle is added to the path conflict vehicle list. The conflict conditions include overlapping driving trajectories and overlapping driving times. The instruction generation unit 604 is used to determine the conflict priority of each vehicle based on the collision time margin, lane change probability, prediction confidence and the relative positional relationship between each vehicle and the construction obstacle in the list of vehicles in path conflict; the vehicle with the highest conflict priority is taken as the primary avoidance target, and vehicle control instructions for the target control vehicle are generated based on the driving prediction data of the primary avoidance target.

[0104] The control device 600 for vehicles in construction risk areas provided in the above embodiments can realize the technical solutions described in the above embodiments of the control method for vehicles in construction risk areas. The specific implementation principles of each module or unit can be found in the corresponding content in the above embodiments of the control method for vehicles in construction risk areas, which will not be repeated here.

[0105] like Figure 7 As shown, the present invention also provides a vehicle 700. The vehicle 700 includes a processor 701, a memory 702, a display 703, and a data acquisition unit 704. Figure 7 Only some components of vehicle 700 are shown, but it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead.

[0106] In some embodiments, processor 701 may be a central processing unit (CPU), microprocessor, or other data processing chip, used to run program code stored in memory 702 or process data, such as the vehicle control method for construction risk areas in this invention.

[0107] In some embodiments, processor 701 may be a single server or a group of servers. The server group may be centralized or distributed. In some embodiments, processor 701 may be local or remote. In some embodiments, processor 701 may be implemented on a cloud platform. In one embodiment, the cloud platform may include a private cloud, public cloud, hybrid cloud, community cloud, distributed cloud, intranet, multi-cloud, etc., or any combination thereof.

[0108] In some embodiments, memory 702 may be an internal storage unit of vehicle 700, such as a hard disk or memory of vehicle 700. In other embodiments, memory 702 may also be an external storage device of vehicle 700, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on vehicle 700.

[0109] Furthermore, the memory 702 may include both internal storage units of the vehicle 700 and external storage devices. The memory 702 is used to store application software and various types of data installed on the vehicle 700.

[0110] In some embodiments, display 703 may be an LED display, a liquid crystal display, a touch-screen liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. Display 703 is used to display information from vehicle 700 and to display a user interface for visualization. Data acquisition unit 704 is used to acquire feature vector data of target monitoring vehicles in construction risk areas. Components 701-703 of vehicle 700 communicate with each other via a bus.

[0111] In one embodiment, when processor 701 executes a control program for vehicles within a construction risk area stored in memory 702, the following steps can be implemented: When the target controlled vehicle is in the construction risk area, obtain the temporal feature matrix of the target monitored vehicle in the construction risk area; The driving prediction data of the target monitoring vehicle is determined based on the time-series feature matrix. The driving prediction data includes at least the lane change probability and the prediction confidence level corresponding to the lane change probability. Based on the driving prediction data, the predicted driving trajectory of the target monitoring vehicle is generated. If the planned driving trajectory of the target control vehicle and the predicted driving trajectory meet the preset conflict conditions, the target monitoring vehicle is added to the path conflict vehicle list. The conflict conditions include overlapping driving trajectories and overlapping driving times. Based on the collision time margin, lane change probability, prediction confidence, and relative positional relationship between each vehicle and construction obstacle in the list of vehicles in path conflict, the conflict priority of each vehicle is determined. The vehicle with the highest conflict priority is selected as the primary avoidance target, and vehicle control commands are generated for the target vehicle based on the driving prediction data of the primary avoidance target.

[0112] It should be understood that when the processor 701 executes the control program for vehicles in the construction risk area stored in the memory 702, in addition to the functions mentioned above, it can also perform other functions, as detailed in the description of the corresponding method embodiments above.

[0113] Furthermore, this embodiment of the invention does not specifically limit the type of vehicle 700 mentioned. Vehicle 700 can be a portable vehicle such as a tablet computer, a personal digital assistant (PDA), a wearable device, or a laptop computer. Exemplary embodiments of portable vehicles include, but are not limited to, portable vehicles running iOS, Android, Microsoft, or other operating systems. The aforementioned portable vehicle can also be other portable vehicles, such as a laptop computer with a touch-sensitive surface (e.g., a touch panel). It should also be understood that in some other embodiments of the invention, vehicle 700 may not be a portable vehicle, but rather a desktop computer with a touch-sensitive surface (e.g., a touch panel).

[0114] Accordingly, this application also provides a computer-readable storage medium for storing computer-readable programs or instructions. When the programs or instructions are executed by a processor, they can implement the steps or functions of the control methods for vehicles in construction risk areas provided in the above-described method embodiments.

[0115] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware (such as a processor, controller, etc.), and the computer program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0116] The above provides a detailed description of the control method, device, and vehicle for vehicles in construction risk areas provided by the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for controlling vehicles within a construction risk area, characterized in that, include: When the target controlled vehicle is in a construction risk area, obtain the temporal feature matrix of the target monitored vehicle in the construction risk area; The driving prediction data of the target monitored vehicle is determined based on the time-series feature matrix, and the driving prediction data includes at least the lane change probability and the prediction confidence level corresponding to the lane change probability; Based on the driving prediction data, a predicted driving trajectory of the target monitoring vehicle is generated. If the planned driving trajectory of the target control vehicle and the predicted driving trajectory meet preset conflict conditions, the target monitoring vehicle is added to the path conflict vehicle list. The conflict conditions include overlapping driving trajectories and overlapping driving times. Based on the collision time margin, lane change probability, prediction confidence, and relative positional relationship between each vehicle and construction obstacle in the list of vehicles in the path conflict, the conflict priority of each vehicle is determined. The vehicle with the highest conflict priority is selected as the primary avoidance target, and vehicle control commands are generated for the target vehicle based on the driving prediction data of the primary avoidance target.

2. The method according to claim 1, characterized in that, The step of obtaining the temporal feature matrix of the target monitoring vehicles in the construction risk area includes: Acquire feature vector data of the target monitored vehicle; The feature vector data is stored in the target queue buffer corresponding to the target monitoring vehicle in chronological order, and the feature vector data is normalized. The normalized feature vector data in the target queue buffer is extracted and arranged in chronological order to obtain the temporal feature matrix of the target monitoring vehicle.

3. The method according to claim 2, characterized in that, The acquisition of feature vector data of the target monitored vehicle includes: The intent micro-features, kinematic features, and interaction scene features of the target monitoring vehicle in the construction risk area are acquired. The intent micro-features include the instantaneous speed difference of the wheel set and the turn signal status. The kinematic features include relative longitudinal distance, relative lateral distance, relative speed, and relative acceleration. The interaction scene features include time margin, lane line position, and remaining passable road width. Based on the intent micro-features, the kinematic features, and the interaction scene features, feature vector data of the target monitoring vehicle is generated.

4. The method according to claim 1, characterized in that, The step of determining the driving prediction data of the target monitored vehicle based on the time-series feature matrix includes: The temporal feature matrix is ​​input into the driving intention prediction model for temporal information encoding and intention decoding to obtain the driving prediction data of the target monitored vehicle.

5. The method according to claim 1, characterized in that, If the planned driving trajectory of the target controlled vehicle and the predicted driving trajectory meet a preset conflict condition, then the target monitored vehicle is added to the path conflict vehicle list, including: Based on the lane change probability and prediction confidence in the driving prediction data, target monitoring vehicles with valid lane change intentions are identified. If the target monitoring vehicle with valid lane change intention and the target control vehicle have overlapping driving trajectories and overlapping driving times on the predicted driving trajectory and the planned driving trajectory, then the target monitoring vehicle with valid lane change intention is added to the list of vehicles with path conflict.

6. The method according to claim 1, characterized in that, The step of designating the vehicle with the highest conflict priority as the primary avoidance target and generating vehicle control commands for the target vehicle based on the driving prediction data of the primary avoidance target includes: Traverse the list of vehicles with path conflicts. If the main avoidance target meets the preset avoidance conditions, generate an active deceleration and yield command as a vehicle control command. The avoidance conditions are that the lane change probability is greater than a preset threshold, the turn signal is continuously illuminated, and the intention micro-features and lateral acceleration are enhanced in the same way. If the main obstacle to avoid does not meet the avoidance conditions but meets the preset following conditions, a following deceleration command is generated as a vehicle control command. The following conditions are that the lane change probability is within a preset range and the lateral movement is in an oscillating mode.

7. The method according to claim 1, characterized in that, After identifying the vehicle with the highest conflict priority as the primary avoidance target and generating vehicle control commands for the target vehicle based on the driving prediction data of the primary avoidance target, the method further includes: The decision objective is obtained by parsing the vehicle control commands; Based on the decision objective, the dynamic constraints of the target-controlled vehicle, and the road boundary constraints, trajectory planning is performed to obtain a reference driving trajectory; The vehicle control quantity is obtained by performing rolling optimization calculations on the reference driving trajectory and the current state of the target controlled vehicle.

8. The method according to claim 7, characterized in that, After obtaining the vehicle control quantity, the method further includes: The system responds to the vehicle control input to control the target vehicle to drive, and obtains the real-time status information of the target vehicle after it has started driving. The real-time status information is fed back to the step of obtaining the feature vector data of the target monitoring vehicles in the construction risk area.

9. A control device for vehicles in a construction risk area, characterized in that, The device includes: The data acquisition unit is used to acquire the time-series feature matrix of the target monitored vehicle in the construction risk area when the target controlled vehicle is in the construction risk area; A data processing unit is used to determine the driving prediction data of the target monitoring vehicle based on the time-series feature matrix, and obtain the time-series feature matrix of the target monitoring vehicle. The vehicle list generation unit generates a predicted driving trajectory of the target monitored vehicle based on the driving prediction data. If the planned driving trajectory of the target controlled vehicle and the predicted driving trajectory meet preset conflict conditions, the target monitored vehicle is added to the path conflict vehicle list. The conflict conditions include overlapping driving trajectories and overlapping driving times. The instruction generation unit is used to determine the conflict priority of each vehicle based on the collision time margin, lane change probability, prediction confidence, and relative positional relationship between each vehicle and the construction obstacle in the list of vehicles in the path conflict; to take the vehicle with the highest conflict priority as the primary avoidance target, and to generate vehicle control instructions for the target control vehicle based on the driving prediction data of the primary avoidance target.

10. A vehicle, characterized in that, The vehicle includes a data acquisition unit, a memory, and a processor, wherein, The data acquisition device is used to acquire the time-series feature matrix of the target monitored vehicle in the construction risk area when the target controlled vehicle is in the construction risk area; The memory is used to store executable program code; The processor is configured to: determine the driving prediction data of the target monitoring vehicle based on the time-series feature matrix; generate the planned driving trajectory of the target control vehicle based on the real-time vehicle operation data, target driving position, and current road data of the target control vehicle; generate the predicted driving trajectory of the target monitoring vehicle based on the driving prediction data; if the planned driving trajectory of the target control vehicle and the predicted driving trajectory meet preset conflict conditions, add the target monitoring vehicle to the path conflict vehicle list, the conflict conditions including overlapping driving trajectories and overlapping driving times; determine the conflict priority of each vehicle based on the collision time margin, lane change probability, prediction confidence, and relative positional relationship between each vehicle and construction obstacles in the path conflict vehicle list; designate the vehicle with the highest conflict priority as the primary avoidance target, and generate vehicle control commands for the target control vehicle based on the driving prediction data of the primary avoidance target.

Images