Vehicle dynamic coordination method and device for signal light area

Abstract

The present disclosure provides a vehicle dynamic coordination method and device for a signal light area. By comprehensively evaluating the perception contribution between the candidate coordination vehicles and the host vehicle, the field of view complementary relationship between the candidate coordination vehicles, and the current network environment, the coordination vehicle set is dynamically determined. In combination with the task characteristics, the node resource state, and the signal light area vehicle queue prediction result, the task scheduling and strategy adjustment are performed, so as to reduce the communication redundancy in the cooperative perception process, improve the adaptability and reliability of the task offloading decision, and improve the passing prediction accuracy and driving decision stability of the host vehicle in the signal light area.

Description

Technical Field

[0001] This disclosure relates to the field of intelligent driving technology, and more specifically, to a method and apparatus for dynamic vehicle coordination in traffic light areas. Background Technology

[0002] With the development of vehicle-to-everything (V2X) communication, edge computing, and intelligent driving technologies, vehicles can acquire multi-source information such as surrounding vehicles, roadside facilities, and traffic lights during operation through vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) communication. Based on this information, they can perform collaborative perception, task offloading, and driving decisions. Especially in traffic light intersections in urban areas, where vehicle density is high, traffic participants are complex, and signal phase changes are frequent, it is often difficult for the vehicle to perceive the queue status ahead, blind spot vehicles, and intersection traffic trends in a timely and comprehensive manner by relying solely on its own sensors and onboard computing platform. Therefore, it is necessary to achieve dynamic collaboration with surrounding vehicles and roadside units.

[0003] Existing collaborative perception schemes typically employ broadcast requests or select cooperative vehicles based on single factors such as distance and signal strength. While these methods are simple to implement, they tend to select vehicles with highly overlapping perception fields, leading to data redundancy, high communication bandwidth consumption, and difficulty in adaptively adjusting based on vehicle perception contributions and link status when network bandwidth fluctuates. Furthermore, existing task offloading schemes often determine task execution nodes based on fixed rules or single resource indicators, failing to comprehensively consider task latency, node computing power, link stability, and dynamic changes in vehicles, resulting in delayed or unstable task scheduling results. Summary of the Invention

[0004] This disclosure provides at least one method and apparatus for dynamic vehicle coordination in traffic light areas. By comprehensively evaluating the perception contribution between candidate cooperating vehicles and the master vehicle, the visual complementarity between candidate cooperating vehicles, and the current network environment, a set of cooperating vehicles is dynamically determined. Task scheduling and strategy adjustment are performed by combining task characteristics, node resource status, and traffic light area vehicle queue prediction results. This reduces communication redundancy in the cooperative perception process, improves the adaptability and reliability of task offloading decisions, and enhances the accuracy of the master vehicle's passage prediction and the stability of its driving decisions in traffic light areas.

[0005] This disclosure provides a vehicle dynamic coordination method for traffic light areas, including: Obtain the status information of candidate cooperative vehicles around the master vehicle, the current network environment information, and the driving task information corresponding to the master vehicle; Based on the candidate cooperative vehicle status information and the current network environment information, the cooperative value of the candidate cooperative vehicles is evaluated, and a set of target cooperative vehicles for participating in the master vehicle cooperative perception is determined from the candidate cooperative vehicles. Based on the driving task information, the resource status information of each target cooperative vehicle in the target cooperative vehicle set, and the resource status information of available edge nodes, the task execution node corresponding to the driving task is determined. When the main vehicle approaches the traffic light area, the traffic light status information of the traffic light area and the real-time motion status information of the vehicles in the traffic light area are acquired. Based on the traffic light status information and the real-time motion status information, the vehicle queue status of the traffic light area is predicted to obtain the intersection traffic prediction result. Based on the intersection traffic prediction results, the target cooperative vehicle set and the task execution node are adjusted to obtain the cooperative perception and task scheduling results for the traffic light area.

[0006] In one optional implementation, the candidate cooperative vehicle status information includes at least one of the following: the candidate cooperative vehicle's position, speed, acceleration, heading angle, sensor configuration information, available computing resources, remaining battery power, available communication bandwidth, and link quality information with the host vehicle. The current network environment information includes at least one of the following: currently available bandwidth, link stability, communication latency, data transmission mode, and communication coverage. The driving task information includes at least one of the following: task type, task data volume, required computing resources, maximum tolerable latency, and task priority.

[0007] In one optional implementation, based on the candidate cooperative vehicle status information and the current network environment information, the cooperative value of the candidate cooperative vehicles is evaluated, and a set of target cooperative vehicles for participating in master vehicle cooperative perception is determined from the candidate cooperative vehicles, specifically including: Based on the positional relationship, perception capability, and link stability of the candidate cooperative vehicle relative to the master vehicle, the first relationship evaluation result corresponding to the candidate cooperative vehicle is determined. Based on the complementary perception field of vision between candidate cooperative vehicles and the set of selected cooperative vehicles, the second relationship evaluation result corresponding to the candidate cooperative vehicle is determined. Based on the first relationship evaluation results and the second relationship evaluation results, the comprehensive collaborative value of the candidate collaborative vehicles is determined; Based on the comprehensive collaborative value and the current network environment information, the target collaborative vehicle set is determined from the candidate collaborative vehicles.

[0008] In one optional implementation, the first relationship evaluation result and the second relationship evaluation result are determined based on the following steps: The perception quality of the candidate cooperative vehicles is determined based on their sensor configuration information and current perception mode. Based on the spatial distance, relative speed, relative position, and direction of motion between the candidate cooperative vehicle and the host vehicle, the link stability between the candidate cooperative vehicle and the host vehicle is determined, and based on the perceived quality, the spatial distance, and the link stability, the first relationship evaluation result is determined. Based on the location and sensor configuration information of the candidate cooperative vehicles, the perception field of view corresponding to the candidate cooperative vehicles is determined; The overlap of the perception field of view corresponding to the candidate cooperative vehicle with the perception field of view corresponding to each of the selected cooperative vehicles in the set of selected cooperative vehicles is analyzed to obtain the field of view complementarity information, and the second relationship evaluation result is determined based on the field of view complementarity information.

[0009] In one optional implementation, based on the comprehensive collaborative value and the current network environment information, the target collaborative vehicle set is determined from the candidate collaborative vehicles, specifically including: Based on the currently available bandwidth, determine the upper limit of the number of vehicles allowed to participate in collaborative sensing and the upper limit of the amount of collaborative sensing data transmitted. Initialize the selected set of cooperative vehicles; Determine the marginal collaboration gain of each unselected candidate collaborative vehicle after it has been added to the set of selected collaborative vehicles; Candidate cooperative vehicles whose marginal cooperative gains meet preset screening conditions are selected and added to the set of selected cooperative vehicles. When the data transmission volume corresponding to the selected set of cooperative vehicles reaches the upper limit of the cooperative sensing data transmission volume, or when the marginal cooperative gain corresponding to the unselected candidate cooperative vehicles is lower than the preset gain threshold, the filtering stops, and the current set of selected cooperative vehicles is determined as the target set of cooperative vehicles.

[0010] In one optional implementation, the task execution node corresponding to the driving task is determined based on the driving task information, the resource status information of each target cooperative vehicle in the target cooperative vehicle set, and the resource status information of available edge nodes, specifically including: The driving task information is subjected to feature processing to obtain task feature information; Obtain the resource status information corresponding to the master vehicle, each target collaborative vehicle in the target collaborative vehicle set, and the available edge nodes; Based on the task feature information and the resource status information, construct task unloading decision input information; The task unloading decision input information is input into a pre-trained intelligent decision model to obtain a selection score corresponding to at least one candidate execution node; Based on the selection score, the task execution node is determined from the master vehicle, the target collaborative vehicles in the target collaborative vehicle set, and the available edge nodes.

[0011] In one optional implementation, the intelligent decision-making model is used to output a selection score for the corresponding candidate execution node based on task characteristics, node computing resources, node communication resources, node remaining power, link stability, and task latency constraints. When the task execution node is the master vehicle, it controls the master vehicle to execute the driving task locally. When the task execution node is the target cooperative vehicle, the data corresponding to the driving task is sent to the target cooperative vehicle for execution via the vehicle-to-vehicle communication link. When the task execution node is an available edge node, the data corresponding to the driving task is sent to the available edge node for execution via the vehicle-to-infrastructure communication link.

[0012] In one optional implementation, based on the traffic light status information and the real-time motion status information, the vehicle queue status of the traffic light area is predicted to obtain the intersection traffic prediction result, specifically including: The real-time motion status information of vehicles within the traffic light area is used as the initial state; The following behavior of the vehicle is corrected based on the phase of the traffic light and the corresponding remaining time. Based on the corrected car-following behavior, the speed and position of vehicles in the traffic light area are extrapolated step-by-step over the future prediction period. Based on the vehicle speed and vehicle position obtained from the simulation, the maximum queue length and queue dissipation time corresponding to the traffic light area are determined as the intersection traffic prediction result. Specifically, when the traffic light phase is red, the expected following distance of the target vehicle is increased based on the remaining red light time; when the traffic light phase is green and the remaining green light time is less than a preset time threshold, the expected speed of the target vehicle is reduced.

[0013] In one optional implementation, the target cooperative vehicle set and the task execution node are adjusted based on the intersection traffic prediction results, specifically including: When the intersection traffic prediction result indicates that a vehicle queue will form in the traffic light area ahead, the weight corresponding to the complementary contribution of perception vision in the cooperative value assessment is increased, and the target cooperative vehicle set is re-determined based on the adjusted weight. Candidate cooperative vehicles that can observe the tail of the queue, the opposite area of ​​the intersection, the adjacent lane area, or the blind spot of the main vehicle are preferentially selected to be added to the target cooperative vehicle set. When the intersection traffic prediction result indicates that the expected waiting time of the main vehicle in the traffic light area is greater than the preset waiting time threshold, driving tasks that are not real-time and have a computational load greater than the preset computational load threshold will be prioritized for scheduling to available edge nodes for execution.

[0014] This disclosure also provides a vehicle dynamic coordination device for traffic light areas, including: The data acquisition module is used to acquire the status information of candidate cooperative vehicles around the master vehicle, the current network environment information, and the driving task information corresponding to the master vehicle. The collaborative vehicle determination module is used to evaluate the collaborative value of the candidate collaborative vehicles based on the candidate collaborative vehicle status information and the current network environment information, and to determine the set of target collaborative vehicles for participating in the master vehicle collaborative perception from the candidate collaborative vehicles. The task node determination module is used to determine the task execution node corresponding to the driving task based on the driving task information, the resource status information of each target cooperative vehicle in the target cooperative vehicle set, and the resource status information of available edge nodes. The status information acquisition module is used to acquire the status information of the traffic lights in the traffic light area and the real-time motion status information of the vehicles in the traffic light area when the main vehicle approaches the traffic light area. The traffic prediction module is used to predict the vehicle queue status of the traffic light area based on the traffic light status information and the real-time motion status information, and obtain the intersection traffic prediction result. The vehicle coordination and task scheduling module is used to adjust the target coordination vehicle set and the task execution node according to the intersection traffic prediction results, so as to obtain the coordination perception and task scheduling results for the traffic light area.

[0015] This disclosure also provides an electronic device, including: a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory via the bus. When the machine-readable instructions are executed by the processor, the steps of the above-described vehicle dynamic coordination method for traffic light areas, or any possible implementation of the above-described vehicle dynamic coordination method for traffic light areas, are performed.

[0016] This disclosure also provides a computer-readable storage medium storing a computer program that, when executed by a processor, performs the steps of the above-described vehicle dynamic coordination method for traffic light areas, or any possible implementation of the above-described vehicle dynamic coordination method for traffic light areas.

[0017] This disclosure also provides a computer program product, including a computer program / instructions, which, when executed by a processor, implements the above-described vehicle dynamic coordination method for traffic light areas, or the steps in any possible implementation of the above-described vehicle dynamic coordination method for traffic light areas.

[0018] This disclosure provides a method and apparatus for dynamic vehicle coordination in traffic light areas. By comprehensively evaluating the perception contribution between candidate cooperating vehicles and the master vehicle, the visual complementarity between candidate cooperating vehicles, and the current network environment, the method dynamically determines the set of cooperating vehicles. It also combines task characteristics, node resource status, and traffic light area vehicle queue prediction results to perform task scheduling and strategy adjustment, thereby reducing communication redundancy in the cooperative perception process, improving the adaptability and reliability of task offloading decisions, and enhancing the accuracy of the master vehicle's passage prediction and the stability of its driving decisions in traffic light areas.

[0019] To make the above-mentioned objects, features and advantages of this disclosure more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0020] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings used in the embodiments will be briefly described below. These drawings are incorporated in and constitute a part of this specification. They illustrate embodiments conforming to this disclosure and, together with the specification, serve to explain the technical solutions of this disclosure. It should be understood that the following drawings only show some embodiments of this disclosure and should not be considered as limiting the scope. Those skilled in the art can obtain other related drawings based on these drawings without creative effort.

[0021] Figure 1 A flowchart of a vehicle dynamic coordination method for traffic light areas provided by an embodiment of this disclosure is shown; Figure 2 A flowchart is shown below for another vehicle dynamic coordination method for traffic light areas provided by an embodiment of this disclosure; Figure 3 A schematic diagram of a vehicle dynamic coordination device for traffic light areas provided in an embodiment of this disclosure is shown; Figure 4 A schematic diagram of an electronic device provided in an embodiment of the present disclosure is shown. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of this disclosure clearer, the technical solutions of the embodiments of this disclosure 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 this disclosure, and not all of them. The components of the embodiments of this disclosure described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this disclosure provided in the accompanying drawings is not intended to limit the scope of the claimed disclosure, but merely represents selected embodiments of this disclosure. All other embodiments obtained by those skilled in the art based on the embodiments of this disclosure without inventive effort are within the scope of protection of this disclosure.

[0023] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0024] In this document, the term "and / or" merely describes a relationship, indicating that three relationships can exist. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. Furthermore, the term "at least one" in this document means any combination of at least two of any one or more elements. For example, including at least one of A, B, and C can mean including any one or more elements selected from the set consisting of A, B, and C.

[0025] Research has revealed that existing collaborative sensing schemes typically employ broadcast requests or select collaborative vehicles based on single factors such as distance and signal strength. While these methods are simple to implement, they tend to select vehicles with highly overlapping perception fields, leading to data redundancy, high communication bandwidth consumption, and difficulty in adaptively adjusting based on vehicle perception contributions and link status when network bandwidth fluctuates. Furthermore, existing task offloading schemes often determine task execution nodes based on fixed rules or single resource indicators, failing to comprehensively consider task latency, node computing power, link stability, and dynamic changes in vehicles, resulting in delayed or unstable task scheduling results.

[0026] Based on the above research, this disclosure provides a method and apparatus for dynamic vehicle coordination in traffic light areas. By comprehensively evaluating the perception contribution between candidate cooperating vehicles and the master vehicle, the visual complementarity between candidate cooperating vehicles, and the current network environment, the set of cooperating vehicles is dynamically determined. Task scheduling and strategy adjustment are performed by combining task characteristics, node resource status, and traffic light area vehicle queue prediction results. This reduces communication redundancy in the cooperative perception process, improves the adaptability and reliability of task offloading decisions, and enhances the accuracy of the master vehicle's passage prediction and the stability of its driving decisions in traffic light areas.

[0027] To facilitate understanding of this embodiment, a detailed description of the vehicle dynamic coordination method for traffic light areas disclosed in this disclosure is provided first. The executing entity of the vehicle dynamic coordination method for traffic light areas provided in this disclosure is generally a computer device with certain computing capabilities. This computer device may include, for example, a terminal device, a server, or other processing devices. The terminal device may be a user equipment (UE), mobile device, user terminal, terminal, cellular phone, cordless phone, personal digital assistant (PDA), handheld device, computing device, in-vehicle device, wearable device, etc. In some possible implementations, this vehicle dynamic coordination method for traffic light areas can be implemented by a processor calling computer-readable instructions stored in memory.

[0028] See Figure 1 The diagram shows a flowchart of a vehicle dynamic coordination method for traffic light areas provided in an embodiment of this disclosure. The method includes steps S101 to S106, wherein: S101. Obtain the status information of candidate cooperative vehicles around the main vehicle, the current network environment information, and the driving task information corresponding to the main vehicle.

[0029] In practical implementation, when intelligent driving vehicles travel to urban roads, intersections, traffic light control areas, or other road areas that require collaborative perception and task scheduling, they can first obtain the status information of candidate collaborative vehicles around the main vehicle, the current network environment information, and the driving task information corresponding to the main vehicle, so as to provide basic data for subsequent collaborative vehicle screening, task execution node determination, and traffic status prediction in traffic light areas.

[0030] Specifically, the master vehicle can be a vehicle equipped with intelligent driving and vehicle-to-everything (V2X) communication capabilities. The master vehicle can acquire information about surrounding vehicles and the road environment through vehicle-to-vehicle communication, vehicle-to-infrastructure communication, or onboard sensing devices. When the master vehicle detects insufficient sensing range, is about to enter a traffic light intersection, needs to perform a computationally intensive driving task, or receives a collaborative sensing trigger command from the vehicle control system, it can generate a collaborative sensing request and a task scheduling request. The collaborative sensing request can be sent to surrounding vehicles within the master vehicle's communication range via vehicle-to-vehicle communication, or to roadside units via vehicle-to-infrastructure communication. The roadside units then collect and summarize the vehicle status information within their coverage area.

[0031] In this embodiment, the candidate cooperative vehicle status information can be understood as the status information of vehicles around the main vehicle that can participate in cooperative perception or task execution. Candidate cooperative vehicles may include vehicles located in front of, behind, to the front side, to the rear side, in adjacent lanes, in the opposite area of ​​an intersection, or near a queue of vehicles. For each candidate cooperative vehicle, the acquired status information may include basic kinematic information such as vehicle identification, vehicle position, vehicle speed, vehicle acceleration, vehicle heading angle, lane, direction of travel, relative distance to the main vehicle, and relative speed to the main vehicle.

[0032] Furthermore, the candidate cooperative vehicle status information may also include the candidate cooperative vehicle's perception capability information. This perception capability information may include the type of sensors configured on the candidate cooperative vehicle, the number of sensors, sensor installation orientation, maximum detection distance, horizontal field of view, vertical field of view, current sensor operating status, and the types of data that can be output.

[0033] For example, candidate cooperative vehicles can be equipped with cameras, millimeter-wave radar, lidar, ultrasonic radar, or other onboard sensing devices. By acquiring the sensing capability information of candidate cooperative vehicles, their current field of view can be estimated, and it can be further determined whether this field of view can cover the blind spot of the main vehicle, traffic light intersections, queue ends, or other areas that the main vehicle needs to focus on.

[0034] Furthermore, the candidate cooperative vehicle status information may also include the candidate cooperative vehicle's computing resource information and communication resource information. The computing resource information may include the candidate cooperative vehicle's available processor resources, available storage resources, current task load, remaining battery power or power status, etc.; the communication resource information may include the candidate cooperative vehicle's currently available communication bandwidth, communication interface type, communication latency, packet loss rate, signal reception strength, and link quality with the host vehicle or roadside unit, etc.

[0035] In some implementations, candidate cooperative vehicle status information can be obtained from vehicle status messages periodically broadcast by the candidate cooperative vehicles. These vehicle status messages may include basic vehicle safety information, cooperative perception status messages, or resource status messages. After receiving status messages from multiple candidate cooperative vehicles, the master vehicle or roadside unit can parse, deduplicate, and align the status messages according to the message timestamp, vehicle identification, and communication source to obtain the candidate cooperative vehicle status information for the same moment or within the same time window.

[0036] In some embodiments, the current network environment information may include the communication environment status between the host vehicle and candidate cooperative vehicles, between the host vehicle and roadside units, and between candidate cooperative vehicles and roadside units. The current network environment information may include currently available bandwidth, communication latency, link stability, packet loss rate, channel occupancy rate, communication coverage, signal strength, communication standard, and currently available data transmission modes.

[0037] Among them, the currently available bandwidth can be used to constrain the number of vehicles that can participate in collaborative sensing at the same time and the amount of collaborative sensing data transmission; communication latency and packet loss rate can be used to determine whether collaborative sensing data or task data can be reliably transmitted within the time limit; link stability can be used to determine whether candidate collaborative vehicles can still maintain reliable communication with the master vehicle or roadside unit in the subsequent time.

[0038] Specifically, the current network environment information can be estimated by the main vehicle based on the quality of the received communication signals, historical link changes, and relative motion status, or it can be uniformly generated by the roadside unit based on the communication resource occupancy and link measurement results within its coverage area.

[0039] For example, when the relative speed between the master vehicle and the candidate cooperating vehicle is small, their directions of motion are the same, and the communication signals are continuous and stable, the link stability between them can be considered high; when the relative speed between the candidate cooperating vehicle and the master vehicle is large, the vehicle is about to leave the communication coverage area, or the link packet loss rate increases, the link stability can be considered low.

[0040] In some embodiments, the driving task information corresponding to the master vehicle may include descriptions of driving-related tasks currently to be executed or scheduled by the master vehicle. Driving tasks may include cooperative perception tasks, point cloud fusion tasks, object detection tasks, lane line recognition tasks, local path planning tasks, vehicle speed planning tasks, intersection traffic decision-making tasks, forward vehicle queue prediction tasks, or other computational tasks related to intelligent driving control.

[0041] Here, driving task information may include task type, task data volume, required computing resources, maximum tolerable latency, task priority, source of task input data, type of task output result, and when to use the task execution result.

[0042] For example, when the driver is about to enter a traffic light intersection and needs to obtain the status of the queue ahead, the driving task information may include the task type used to predict the vehicle queue at the intersection, the required input data on the motion status of the vehicles ahead, the prediction duration, and the maximum tolerable delay.

[0043] For example, when the main vehicle needs to perform a high-precision multi-sensor fusion positioning task, the driving task information may include the amount of data to be fused, the required computing power, the result return time limit, and the relationship between the task and the subsequent driving control strategy.

[0044] In some implementations, the primary vehicle can generate driving task information based on the onboard domain controller, autonomous driving controller, or task management module. The task management module can identify the driving task to be executed based on the primary vehicle's current operating status, environmental perception results, navigation route, distance to traffic light areas, and vehicle control requirements, and then characterize the driving task. This characterization can be used to subsequently construct task offloading decision input information, enabling the task offloading decision process to simultaneously consider the task's complexity, latency constraints, and impact on driving safety and traffic efficiency.

[0045] Furthermore, after acquiring the status information of candidate cooperative vehicles, current network environment information, and driving task information, the above information can be preprocessed uniformly. Preprocessing can include data format conversion, time synchronization, outlier removal, missing value completion, coordinate system transformation, and standardization. For example, the positions of each candidate cooperative vehicle can be uniformly transformed to a local coordinate system with the master vehicle as the reference, speed, heading angle, and link quality information from different sources can be represented according to a unified data format, and vehicle status information and network environment information can be aligned according to timestamps.

[0046] In a preferred embodiment, candidate cooperative vehicle status information, current network environment information, and driving task information can be aggregated at a decision center. The decision center can be located locally on the primary vehicle or in a roadside unit covering the current road area. When the decision center is located locally on the primary vehicle, the primary vehicle can directly complete subsequent cooperative vehicle selection and task scheduling based on locally collected and received information. When the decision center is located in a roadside unit, the roadside unit can aggregate vehicle status, network status, and task requests within its communication coverage area and utilize its superior computing resources to execute subsequent decisions.

[0047] S102. Based on the candidate cooperative vehicle status information and the current network environment information, evaluate the cooperative value of the candidate cooperative vehicles, and determine the target cooperative vehicle set from the candidate cooperative vehicles to participate in the master vehicle cooperative perception.

[0048] In practice, after obtaining the status information of candidate cooperative vehicles around the master vehicle and the current network environment information, the cooperative value of the candidate cooperative vehicles can be evaluated based on the status information of the candidate cooperative vehicles and the current network environment information, and the set of target cooperative vehicles for participating in the master vehicle's cooperative perception can be determined from the candidate cooperative vehicles.

[0049] The candidate cooperative vehicle status information may include at least one of the following: the candidate cooperative vehicle's position, speed, acceleration, heading angle, sensor configuration information, available computing resources, remaining battery power, available communication bandwidth, and link quality information with the host vehicle; the current network environment information may include at least one of the following: current available bandwidth, communication latency, link stability, packet loss rate, channel occupancy rate, and available data transmission modes.

[0050] Specifically, when assessing the collaborative value, the first evaluation result of the candidate collaborative vehicle can be determined based on its positional relationship, perception capability, and link stability relative to the host vehicle. The first relationship evaluation result characterizes the direct perception contribution of the candidate collaborative vehicle relative to the host vehicle. In other words, this evaluation result reflects the perception gain that the candidate collaborative vehicle can provide to the host vehicle when participating in collaborative perception alone. If the distance between the candidate collaborative vehicle and the host vehicle is moderate, its relative motion is stable, its sensor field of view can cover the host vehicle's blind spot or key areas ahead, and the link quality between the candidate collaborative vehicle and the host vehicle or roadside unit is high, then the candidate collaborative vehicle can be determined to have a high first relationship evaluation result. If the candidate collaborative vehicle is too far away, about to leave the communication coverage area, has poor link quality, or its perception field of view highly overlaps with the host vehicle's own perception range, then the first relationship evaluation result of the candidate collaborative vehicle can be reduced.

[0051] Here, when determining the first relationship evaluation result, the perception quality of the candidate cooperative vehicles can be determined based on their sensor configuration information and current perception operating mode. Sensor configuration information may include the type, quantity, installation location, maximum detection range, horizontal field of view, vertical field of view, and current operating status of sensors such as cameras, millimeter-wave radar, lidar, and ultrasonic radar. The current perception operating mode may include raw data transmission mode, target-level data transmission mode, feature-level data transmission mode, or low-bandwidth summary data transmission mode.

[0052] In different perception modes, the accuracy and amount of perception information provided by candidate cooperative vehicles vary, thus corresponding to different perception quality evaluation results. For example, when network bandwidth is sufficient, candidate cooperative vehicles can be allowed to provide point clouds, images, or high-resolution target information, in which case their perception quality evaluation can be relatively high; when network bandwidth is limited, target-level or feature-level data transmission can be switched, and in this case, the perception quality can be reassessed based on the reduction in data accuracy.

[0053] Furthermore, the link stability between the candidate cooperative vehicle and the master vehicle can be determined based on the spatial distance, relative speed, relative position, and direction of motion between the candidate cooperative vehicle and the master vehicle, and the first relationship evaluation result can be determined based on the perceived quality, spatial distance, and link stability.

[0054] Specifically, when a candidate cooperating vehicle is close to the host vehicle but not outside the host vehicle's perception blind spot, its perception contribution may not be the highest. When a candidate cooperating vehicle is slightly farther from the host vehicle but is positioned to cover the host vehicle's blind spot, adjacent lanes, or opposite areas at intersections, its first relationship evaluation result can be improved accordingly. Link stability can be determined based on communication signal strength, communication latency, packet loss rate, relative vehicle speed, consistency of vehicle movement direction, and the expected duration of being within the communication coverage area.

[0055] Furthermore, the second relationship evaluation result for a candidate cooperative vehicle can be determined based on the complementary perception field of view between the candidate cooperative vehicle and the selected cooperative vehicle set. The second relationship evaluation result characterizes the incremental perception contribution of the candidate cooperative vehicle relative to the selected cooperative vehicle set. In other words, this evaluation result reflects whether the candidate cooperative vehicle, after joining the current selected cooperative vehicle set, can provide new perception areas, new observation angles, or supplementary coverage of the main vehicle's blind spots. If the perception field of view of the candidate cooperative vehicle highly overlaps with that of the selected cooperative vehicle set, the information provided by the candidate cooperative vehicle may have significant redundancy, and its second relationship evaluation result can be low; if the candidate cooperative vehicle can cover areas not yet covered by the selected cooperative vehicles, its second relationship evaluation result can be high.

[0056] Here, when determining the evaluation results of the second relationship, the perception field of view corresponding to the candidate cooperative vehicle can be determined based on the position and sensor configuration information of the candidate cooperative vehicle; at the same time, the perception field of view corresponding to each selected cooperative vehicle can be determined based on the position and sensor configuration information of each selected cooperative vehicle in the set of selected cooperative vehicles.

[0057] The perception field of view (PDV) can be determined by the vehicle's location, heading angle, sensor installation orientation, sensor detection distance, and field of view. It can be mapped to a local coordinate system referenced to the main vehicle or a road coordinate system referenced to a traffic light intersection. Subsequently, overlap analysis is performed between the PPV of the candidate cooperating vehicle and the PPV of each selected cooperating vehicle in the selected vehicle set to obtain PPV complementarity information. Based on this PPV complementarity information, the second relationship evaluation result is determined.

[0058] For example, multiple vehicles traveling consecutively in the same lane typically have relatively close forward perception fields, resulting in a high degree of overlap. In this situation, if one of these vehicles is already included in the selected set of cooperating vehicles, another vehicle traveling in the same direction, although close to the lead vehicle, has limited additional perception coverage, thus leading to a lower second relationship evaluation result. Conversely, vehicles located in the adjacent lane to the lead vehicle, in the opposite direction at an intersection, near the end of a queue, or on the other side of an area obscured by a large vehicle may have significantly different perception fields from the selected set of cooperating vehicles and can cover areas that are difficult for the lead vehicle or the selected vehicles to observe, thus leading to a higher second relationship evaluation result. By introducing the second relationship evaluation result, redundant data transmission during the cooperative perception process can be reduced, improving the overall perception coverage efficiency of the target cooperating vehicle set.

[0059] Furthermore, after obtaining the first and second relationship evaluation results, the comprehensive collaborative value of the candidate collaborative vehicles can be determined based on these results. The comprehensive collaborative value can be used to represent the overall priority of the candidate collaborative vehicles in joining the target collaborative vehicle set in the current scenario.

[0060] In one implementation, the evaluation results of the first relationship and the second relationship can be weighted to obtain a comprehensive collaborative value. The weight corresponding to the first relationship evaluation result reflects the direct value of the candidate collaborative vehicle relative to the host vehicle, while the weight corresponding to the second relationship evaluation result reflects the complementary value of the candidate collaborative vehicle relative to the set of selected collaborative vehicles.

[0061] It should be noted that the above weights can be preset weights or dynamically adjusted according to the current road scenario, network environment, driver's task, and traffic light area status.

[0062] For example, when the primary vehicle initiates cooperative perception and the selected cooperative vehicle set is empty, the weight corresponding to the first relationship evaluation result can be appropriately increased to prioritize vehicles that make a higher direct contribution to the primary vehicle's perception. When there are many vehicles around the primary vehicle and the perception fields of candidate cooperative vehicles overlap significantly, the weight corresponding to the second relationship evaluation result can be increased to prioritize vehicles that can provide complementary perspectives. When the available bandwidth is low or the communication link is congested, the comprehensive cooperative value can be further adjusted by considering the expected data transmission volume of candidate cooperative vehicles, so that candidate cooperative vehicles with higher perception gain per unit communication cost have a higher selection priority.

[0063] Furthermore, based on the comprehensive collaborative value and current network environment information, a target set of collaborative vehicles can be determined from the candidate collaborative vehicles. Specifically, based on the currently available bandwidth, the upper limit on the number of vehicles allowed to participate in collaborative sensing and the upper limit on the amount of collaborative sensing data transmission can be determined. The upper limit on the number of vehicles can be used to limit the number of vehicles participating in collaborative sensing simultaneously, and the upper limit on the amount of collaborative sensing data transmission can be used to limit the total amount of sensing data transmission generated by the target set of collaborative vehicles per unit time. Different candidate collaborative vehicles occupy different communication bandwidths due to differences in sensor configuration, data type, and transmission mode. Therefore, when selecting the target set of collaborative vehicles, it is necessary to comprehensively consider the comprehensive collaborative value of the vehicles and their corresponding data transmission costs.

[0064] In one implementation, the selected cooperative vehicle set can be initialized to empty, and then the marginal cooperative gain of each unselected candidate cooperative vehicle after being added to the selected cooperative vehicle set can be determined. The marginal cooperative gain can be determined based on the candidate cooperative vehicle's overall cooperative value, its newly added sensing coverage area, its coverage capability in the main vehicle's blind spot or key traffic light areas, and the communication resources it requires. Subsequently, candidate cooperative vehicles whose marginal cooperative gains meet preset screening criteria can be added to the selected cooperative vehicle set. The preset screening criteria may include at least one of the following: maximum marginal cooperative gain, marginal cooperative gain greater than a preset gain threshold, link stability greater than a preset link threshold, and data transmission volume not exceeding the upper limit after addition.

[0065] After each addition of a candidate cooperative vehicle to the selected cooperative vehicle set, the complementary perception field of view between the remaining candidate cooperative vehicles and the updated selected cooperative vehicle set can be redefined, and the second relationship evaluation result, comprehensive cooperative value, or marginal cooperative gain of the remaining candidate cooperative vehicles can be recalculated. Since the selected cooperative vehicle set changes, the additional perception coverage area that the remaining candidate cooperative vehicles can provide also changes. Therefore, iterative updates can avoid excessive overlap of perception fields of view among multiple selected vehicles.

[0066] Furthermore, during the iterative screening process, screening stop conditions can be set based on the current network environment information. When the data transmission volume corresponding to the selected set of collaborative vehicles reaches the upper limit of collaborative sensing data transmission volume, or when the number of selected collaborative vehicles reaches the upper limit of the number of vehicles, the addition of candidate collaborative vehicles can be stopped. When the marginal collaborative gain corresponding to an unselected candidate collaborative vehicle is lower than a preset gain threshold, it indicates that the perception improvement brought by continuing to add collaborative vehicles is limited, and screening can also be stopped. When the link stability of a candidate collaborative vehicle is lower than a preset link threshold, the expected communication latency is greater than a preset latency threshold, or the current network environment switches to a low-bandwidth state, the candidate collaborative vehicle can be temporarily not selected, or its data transmission mode can be adjusted to a low-bandwidth mode and then re-evaluated.

[0067] In some embodiments, after determining the target cooperative vehicle set, the host vehicle or roadside unit can send a cooperative perception confirmation message to each target cooperative vehicle in the target cooperative vehicle set. The cooperative perception confirmation message may include at least one of the following: target perception area, cooperative perception data type, data upload period, data resolution, data compression method, communication link parameters, and continuous cooperative time. After receiving the cooperative perception confirmation message, the target cooperative vehicle can collect or organize the corresponding perception data according to the confirmation message requirements and send it to the host vehicle or roadside unit via vehicle-to-vehicle communication link or vehicle-to-infrastructure communication link. For candidate cooperative vehicles that are not identified as target cooperative vehicles, only periodic status broadcasts can be maintained without uploading high-volume perception information, thereby reducing invalid communication load.

[0068] In some preferred embodiments, the target cooperative vehicle set can be periodically updated or event-triggered updates based on changes in the location of the primary vehicle, the vehicle queue status, the traffic light status, and the network environment. For example, when the primary vehicle enters the preset range of the traffic light intersection, the overall cooperative value of candidate cooperative vehicles that can cover the stop line, the end of the queue, the opposite area of ​​the intersection, or adjacent lanes can be increased; when the available bandwidth decreases, the number of target cooperative vehicles can be reduced, or target cooperative vehicles with higher overall cooperative value and lower data transmission volume can be prioritized; when a selected target cooperative vehicle leaves the communication coverage area or the link quality deteriorates, a replacement vehicle can be reselected from the remaining candidate cooperative vehicles. Thus, the target cooperative vehicle set can adapt to the actual scenarios of dense traffic, rapid status changes, and fluctuating network bandwidth within the traffic light area.

[0069] The above scheme will now be described in conjunction with specific implementation methods.

[0070] Vehicle-to-Vehicle Relationship Score: This component quantifies the direct perceived value of a candidate vehicle relative to the primary vehicle. The specific calculation formula can be expressed as:

[0071] in, Indicates vehicle Main vehicle-vehicle relationship; It is an adjustment coefficient used to control the overall weight of this component; Indicates vehicle The perceived quality coefficient is a comprehensive indicator determined based on the vehicle's sensor configuration and current operating mode. Its calculation method can be dynamically adjusted under different communication modes to reflect adaptability to network conditions. Indicates vehicle The spatial distance between the vehicle and the main vehicle can be calculated based on the coordinates of the two vehicles; This represents the link stability factor, with a value between 0 and 1, used to characterize vehicles. The estimated stability of the communication link between the two vehicles. This factor is dynamically estimated based on the relative speed and relative position of the two vehicles, and takes a larger value when the relative speed of the two vehicles is small and their directions of motion are the same.

[0072] Inter-vehicle relationship score: This quantifies the perceptual complementarity of a candidate vehicle relative to the currently selected vehicle set. The specific calculation formula can be expressed as:

[0073] in, Indicates vehicle The vehicle relationships are relative to the selected set S; It is an adjustment coefficient used to control the overall weight of this component; This represents the currently selected set of vehicles; and Representing vehicles and vehicles The perceived quality coefficient; Indicates vehicle and vehicles Spatial distance between them; Indicates vehicle With vehicles The visual field complementarity coefficient between the two vehicles ranges from 0 to 1. This coefficient is calculated based on the vehicle's position and sensor configuration, projecting the perceived field of view onto a plane centered on the main vehicle. When the visual field overlap between the two vehicles is high, The value is close to 0; when the fields of vision of the two vehicles do not overlap, Its value is close to 1. Specifically, it can be represented as... , This represents the overlap of the field of view.

[0074] Combining the above two components, candidate vehicles Its comprehensive value is divided into:

[0075] in, and These are weighting coefficients, satisfying... These coefficients are used to balance the importance of the master vehicle-vehicle relationship and the inter-vehicle relationship. The system can dynamically adjust these two coefficients according to the current scenario, for example, by appropriately increasing them when a rapid expansion of the perception range is needed. The value of .

[0076] S103. Based on the driving task information, the resource status information of each target cooperative vehicle in the target cooperative vehicle set, and the resource status information of available edge nodes, determine the task execution node corresponding to the driving task.

[0077] In practice, after determining the target collaborative vehicle set, the task execution node corresponding to the driving task can be determined based on the driving task information, the resource status information of each target collaborative vehicle in the target collaborative vehicle set, and the resource status information of available edge nodes.

[0078] The task execution node can be the main vehicle's local computing node, any target cooperative vehicle in the target cooperative vehicle set, or an available edge node deployed in a traffic light area, on the roadside, or in an edge computing network.

[0079] Specifically, the driving task information corresponding to the main vehicle can first be characterized to obtain task feature information. Driving tasks may include collaborative perception fusion tasks, target detection tasks, point cloud processing tasks, vehicle platoon prediction tasks, path planning tasks, vehicle speed planning tasks, intersection traffic decision-making tasks, or other computational tasks related to intelligent driving control. Task feature information may include at least one of the following: task type, task data volume, required computing resources, maximum tolerable latency, task priority, source of task input data, type of task output result, and timing of use of task execution results.

[0080] For example, for point cloud fusion tasks, the task characteristic information may include the amount of point cloud data to be fused, the fusion accuracy requirements, the required computing resources, and the result return time limit; for traffic light area vehicle queue prediction tasks, the task characteristic information may include the prediction area range, the number of input vehicles, the prediction duration, the prediction result type, and the maximum tolerable delay; for path planning or vehicle speed planning tasks, the task characteristic information may include the planning interval length, the number of constraints, the expected output frequency, and the degree of impact of the task result on vehicle control safety.

[0081] Furthermore, resource status information can be obtained for the master vehicle, each target collaborative vehicle in the target collaborative vehicle set, and available edge nodes. The master vehicle's resource status information may include its currently available computing resources, current processor load, available storage resources, number of currently executing tasks, power status, and local task queue status. The target collaborative vehicle's resource status information may include its available computing resources, remaining battery power, current task load, available communication bandwidth, link stability with the master vehicle, link stability with roadside units, and estimated sustainable collaborative time. The available edge node's resource status information may include its available processor resources, available storage resources, communication interface status, current task queue length, communication latency with the master vehicle, communication latency with target collaborative vehicles, and edge node coverage area.

[0082] In some implementations, available edge nodes may include roadside units, roadside edge servers, computing nodes associated with traffic light control facilities, or cloud-edge collaborative nodes communicatively connected to the current traffic light area. Available edge nodes may periodically broadcast their resource status to the host vehicle or roadside unit, or return their current resource status upon receiving a task scheduling request. The resource status information of the target collaborative vehicle may be obtained along with the status information of the candidate collaborative vehicles, or, after determining the target collaborative vehicle set, the target collaborative vehicles may further report more detailed information on available computing resources and task execution capabilities.

[0083] After obtaining the task characteristic information and the resource status information of each candidate execution node, task offloading decision input information can be constructed based on the task characteristic information and resource status information. The task offloading decision input information can include task-side characteristics and node-side characteristics. Among them, task-side characteristics can include task data volume, required computing power, maximum tolerable latency, and task priority; node-side characteristics can include the available computing resources, available bandwidth, communication latency, link stability, remaining power, and current task load of each candidate execution node.

[0084] For example, when a driving task has high real-time requirements, the task offloading decision input information can highlight the maximum tolerable latency and communication link stability; when a driving task is a computationally intensive task with a large amount of data but relatively low real-time requirements, the task offloading decision input information can highlight the available computing resources of the node, the task queue length, and the available bandwidth; when a driving task needs to obtain the result before the main vehicle passes through the traffic light area, the task offloading decision input information can further include the latest return time of the task execution result.

[0085] In some embodiments, task unloading decision input information can be fed into a pre-trained intelligent decision-making model to obtain selection scores corresponding to at least one candidate execution node. The intelligent decision-making model can be deployed on a roadside unit, edge node, or locally on the vehicle. The intelligent decision-making model can be a lightweight intelligent decision-making model, a large language model compression model, a neural network model, a graph neural network model, an attention mechanism model, or other models capable of processing multi-dimensional state inputs and outputting node selection results. This model can be pre-trained based on task unloading samples, vehicle resource state samples, network state samples, and task execution effect samples, enabling it to output corresponding selection scores based on the matching relationship between the current task and node state.

[0086] Specifically, the intelligent decision-making model can output a selection score for each candidate execution node based on task characteristics, node computing resources, node communication resources, remaining node power, link stability, and task latency constraints. The selection score characterizes the overall suitability of assigning the current driving task to the corresponding candidate execution node in terms of meeting task latency, computing resource, communication stability, and energy consumption constraints. If a candidate execution node has high available computing power, low communication latency, high link stability, and a low current task load, its selection score can be high; conversely, if a candidate execution node has sufficient computing power but poor link stability, or its communication link latency cannot meet the maximum tolerable latency of the driving task, its selection score can be lowered.

[0087] In one specific implementation, candidate execution nodes may include the master vehicle's local node, each target collaborative vehicle in the target collaborative vehicle set, and available edge nodes. For each candidate execution node, a corresponding node feature vector can be constructed, and the task feature vector is combined with the node feature vectors and then input into the intelligent decision-making model. The intelligent decision-making model can output a selection score for each candidate execution node. Subsequently, based on the selection score, the task execution node can be determined from the master vehicle, the target collaborative vehicles in the target collaborative vehicle set, and the available edge nodes.

[0088] Typically, the candidate execution node with the highest score can be selected as the task execution node. In other implementations, the final task execution node can be further determined from multiple candidate execution nodes whose scores meet a preset score threshold, taking into account load balancing strategies, task priorities, or communication link stability.

[0089] In some implementations, when the task execution node is the master vehicle, the driving task can be executed locally by the master vehicle. This method is suitable for situations where the task data volume is small, the latency requirements are extremely high, or the current network environment is not suitable for task offloading. For example, for tasks with high safety and real-time requirements, such as emergency braking judgment and short-term obstacle avoidance control, the master vehicle can be given priority in executing these tasks locally to reduce the uncertainty caused by the communication transmission process.

[0090] Here, when the task execution node is the target cooperative vehicle, the data corresponding to the driving task can be sent to the target cooperative vehicle for execution via the vehicle-to-vehicle communication link. This method is suitable when the target cooperative vehicle has high available computing resources, a stable link with the host vehicle, and the target cooperative vehicle itself can provide task-related perception data.

[0091] For example, when a target cooperative vehicle is located in front of the master vehicle and can obtain the status of vehicles queuing at an intersection, and its available computing power is high, some perception data processing or local target detection tasks can be scheduled to be executed by the target cooperative vehicle, thereby reducing the local computing pressure on the master vehicle.

[0092] Here, when the task execution node is an available edge node, the data corresponding to the driving task can be sent to the available edge node for execution via the vehicle-to-infrastructure communication link. This method is suitable for situations where the driving task has a large computational load, requires the fusion of information from multiple vehicles, or where the available edge node has stronger computing power and a more stable communication link.

[0093] For example, for tasks such as multi-vehicle queue prediction in traffic light areas, global situational awareness fusion at intersections, and high-precision path planning, the task data can be offloaded to roadside units or edge computing nodes, which can then perform centralized calculations based on the perception data uploaded by multiple vehicles and traffic light status information, and then return the calculation results to the main vehicle.

[0094] Furthermore, after determining the task execution node, a corresponding task scheduling instruction can be generated. The task scheduling instruction may include at least one of the following: task identifier, task input data, task execution requirements, result return address, result return time limit, task priority, and exception handling strategy. Upon receiving the task scheduling instruction, the master vehicle, target cooperative vehicle, or available edge node can execute the corresponding computational task according to the instruction and return the task execution result to the master vehicle or roadside unit after completion. If a link quality degradation, task execution timeout, insufficient node resources, or the target cooperative vehicle leaving the communication coverage area is detected during task execution, the task execution node selection process can be re-triggered, or the task can be switched to a backup execution node.

[0095] In some preferred embodiments, the intelligent decision-making model can also be dynamically updated by combining current network environment information and changes in the target cooperative vehicle set. Since vehicles may experience state changes such as deceleration, queuing, stopping, and starting within the traffic light area, the link stability and sustainable cooperation time between the target cooperative vehicle and the master vehicle may also change. Therefore, the determination of the task execution node does not need to be a one-time event, but can be re-executed according to a preset period or when a change in event is detected.

[0096] For example, when the link stability of the original task execution node is lower than a preset threshold, a new task execution node can be determined based on the current driving task information, the resource status of the target collaborative vehicle, and the resource status of available edge nodes. When the master vehicle is expected to wait for a long time at a traffic light intersection, edge nodes can be prioritized to execute non-real-time but computationally intensive tasks, so as to use the waiting time to complete the computational tasks required for subsequent driving in advance.

[0097] The above scheme will now be described in conjunction with specific implementation methods.

[0098] The reasoning process of the model can be represented as follows:

[0099] in, Indicates the input state Select Node The probability, This represents the mapping function of the trained intelligent decision-making model.

[0100] The model's internal mechanism uses an attention mechanism to capture the complex relationship between task features and the states of each candidate node. The trained weights within the model dynamically and implicitly balance multiple conflicting objectives. For example, when the task is extremely sensitive to latency and a V2V candidate node has high link stability, the model will give it a higher score; conversely, if the link stability is too low, even if the vehicle has high computing power, the model will lower its score due to the high risk of task failure.

[0101] The model's output is a selection score for each candidate node, satisfying... This score reflects the probability that the model believes selecting this node to perform the task best satisfies the multi-objective optimization requirements.

[0102] The main vehicle or roadside unit receives the model's output and selects the node with the highest score as the final task unloading target.

[0103] Then, the system performs the corresponding task transmission and scheduling based on the type of the selected node: if For the main vehicle itself, the operation is performed locally.

[0104] if If it is a roadside unit, then the task data will be offloaded to the roadside unit.

[0105] if To coordinate vehicles, task data is sent to the selected vehicles via a V2V link.

[0106] S104. When the main vehicle approaches the traffic light area, acquire the traffic light status information of the traffic light area and the real-time motion status information of the vehicles in the traffic light area.

[0107] In practical implementation, when a vehicle approaches the traffic light area, the traffic light status information and the real-time movement status information of vehicles within the traffic light area can be obtained. This allows for subsequent prediction of the intersection's vehicle queue status based on traffic light phase changes and vehicle movement. The traffic light area can be understood as the intersection area controlled by traffic lights and the road area within a preset upstream distance. This includes, for example, the lane area within a certain distance in front of the stop line, adjacent lane areas, the area where queued vehicles are located, and areas where oncoming or lateral traffic participants may affect the passage of the main vehicle.

[0108] Specifically, the vehicle can determine whether it is approaching a traffic light area based on its current location, navigation map information, high-precision map information, or road facility information broadcast by roadside units. When the distance between the vehicle's current location and the stop line of the traffic light intersection ahead is less than a preset distance threshold, or when the vehicle determines that it is about to enter a traffic light-controlled intersection based on the navigation route, the traffic light area information acquisition process can be triggered.

[0109] The preset distance threshold can be set according to road grade, vehicle speed, communication coverage, and forecasting needs. For example, it can be set to a range of several hundred meters in front of the main vehicle, or it can be dynamically adjusted according to the current vehicle speed. The higher the current vehicle speed, the larger the preset distance threshold can be to ensure that the system has sufficient time for prediction and scheduling.

[0110] Here, after triggering the traffic light area information acquisition process, the main vehicle can obtain traffic light status information of the traffic light area from roadside units, traffic light control equipment, traffic infrastructure platforms, or other roadside equipment capable of providing traffic light status information via the vehicle-to-infrastructure communication link. The traffic light status information may include at least one of the following: current signal phase, remaining time of the current signal phase, signal cycle, phase switching sequence, lane clearance relationships for each direction, remaining time of the red light, remaining time of the green light, remaining time of the yellow light, and traffic light control information corresponding to the main vehicle's current lane.

[0111] For example, when the vehicle is in the straight lane, the current signal phase and its remaining time can be obtained; when the vehicle is in the left-turn lane or needs to turn left according to the navigation route, the current phase and remaining time of the left-turn signal can be obtained; when there are separate phase controls for straight, left, and right turns at the traffic light intersection, the target signal phase related to the vehicle's traffic behavior can be determined based on the vehicle's lane, navigation intent, and lane guidance information, and the corresponding remaining time of the phase can be extracted.

[0112] In some implementations, traffic light status information can be obtained from the traffic signal controller by the roadside unit and then sent to the vehicle, or the vehicle can directly receive the signal phase and timing information broadcast by the traffic signal control system. If multiple traffic light control phases exist, the roadside unit or the vehicle can match the traffic light status information based on lane markings, road topology, and the vehicle's navigation route to determine the valid traffic light status information corresponding to the vehicle's current target lane. If the received traffic light status information includes a timestamp, the remaining time of the current signal phase can be corrected based on the difference between the time the vehicle receives the information and the timestamp, thereby improving the timeliness of the traffic light status information.

[0113] Furthermore, while acquiring traffic light status information, the system can also acquire real-time motion status information of vehicles within the traffic light area. This real-time motion status information can include at least one of the following: the position, speed, acceleration, heading angle, lane, vehicle length, vehicle type, distance to the vehicle in front, relative distance to the main vehicle, distance to the stop line, and whether the vehicle is in a stopped or low-speed queue.

[0114] The aforementioned real-time motion status information can be acquired by the vehicle's own sensors or provided by the target cooperative vehicle, roadside unit, or other traffic infrastructure equipment.

[0115] In one implementation, the host vehicle can acquire real-time motion status information of vehicles in front of and around it through its own cameras, millimeter-wave radar, lidar, or other onboard sensing devices. The host vehicle can perform target detection, target tracking, and lane association processing on the data collected by the onboard sensors to obtain the position, velocity, acceleration, and lane information of each vehicle in the host vehicle's local coordinate system. For vehicles that can be stably observed within the host vehicle's sensing range, the host vehicle's sensing results can be directly used as real-time motion status information.

[0116] In another implementation, the main vehicle can acquire real-time motion status information of vehicles within the traffic light area through the aforementioned defined set of target cooperative vehicles. The target cooperative vehicles can perceive surrounding vehicles based on their own sensors and transmit the target vehicle's position, speed, acceleration, lane location, and relative relationship with the stop line at the intersection to the main vehicle or roadside unit. Since the target cooperative vehicles may be located in front of, to the side of, in adjacent lanes, or near the end of a queue, their field of vision can cover areas that are difficult for the main vehicle to observe, thus supplementing the vehicle motion status information in the main vehicle's blind spots.

[0117] In another implementation, real-time motion status information of vehicles within the traffic light area can be acquired by roadside units or intersection sensing devices. Roadside units can detect and track vehicles within the traffic light intersection and its upstream lanes using roadside cameras, millimeter-wave radar, lidar, or other roadside sensing devices, obtaining the position, speed, lane, and queuing status of each vehicle in the intersection coordinate system. Roadside units can also aggregate vehicle status information uploaded by the host vehicle and target cooperating vehicles, fusing multi-source data to form unified real-time motion status information of vehicles within the traffic light area, which is then sent to the host vehicle or used for subsequent centralized prediction calculations.

[0118] In some embodiments, to ensure the accuracy of subsequent vehicle queue prediction, traffic light status information and real-time motion status information can be spatiotemporally aligned. Specifically, data from different sources can be time-synchronized based on the timestamps carried by the traffic light status information, vehicle motion status information, and steerable vehicle status information, so that they correspond to the same sampling time or the same time window. For location data from different sources, they can be uniformly transformed to the same coordinate system, such as a local coordinate system with the steerable vehicle as the origin, a road coordinate system with the intersection stop line as the reference, or an intersection coordinate system with the roadside unit as the reference.

[0119] Furthermore, lane matching and target association processing can be performed on the real-time motion status information of vehicles within the traffic light area. Specifically, the lane of each vehicle can be determined based on vehicle position, heading angle, road lane line information, and high-precision map information; the same vehicle from different times or different sensing sources can be associated based on vehicle historical trajectory, vehicle identification, or target tracking results; for the same vehicle repeatedly observed by the master vehicle, target cooperating vehicles, and roadside units, fusion or deduplication processing can be performed based on observation accuracy, timestamp freshness, and sensor reliability.

[0120] In some implementations, the presence of vehicles in a queue can be preliminarily determined based on vehicle speed, acceleration, distance to the vehicle in front, and distance to the stop line. For example, if a vehicle is within a preset range in front of the stop line, its speed is below a preset low-speed threshold, and its distance to the vehicle in front is less than a preset queue spacing threshold, the vehicle can be marked as a suspected queue vehicle. When multiple vehicles meet the low-speed or stopping conditions and are located in the same lane, it can be preliminarily determined that a vehicle queue exists in that lane.

[0121] In some embodiments, if the main vehicle cannot directly obtain complete traffic light status information or real-time vehicle motion status information, compensation methods can be used to improve the input data. For example, when the remaining time of the signal phase cannot be obtained temporarily, it can be estimated based on the most recently received traffic light status information, signal cycle, and current time; when some vehicles are obscured or temporarily lost, short-term prediction can be performed based on the vehicle's historical speed, acceleration, and trajectory; when there are conflicts between data from different sources, data with updated timestamps, higher confidence, or consistent with the roadside unit's perception results can be given priority.

[0122] In a preferred embodiment, the acquired traffic light status information and real-time motion status information can be organized into traffic light area status data. This data may include the target signal phase, remaining time of the target phase, number of vehicles in each lane, vehicle position, vehicle speed, vehicle acceleration, lane location of each vehicle, distance between each vehicle and the stop line, distance between each vehicle and the vehicle in front, and preliminary queue status indicators. This traffic light area status data can be input into a subsequent vehicle queue status prediction module to predict the maximum queue length, queue dissipation time, and the estimated waiting time for the main vehicle in the traffic light area.

[0123] S105. Based on the traffic light status information and the real-time motion status information, predict the vehicle queue status of the traffic light area to obtain the intersection traffic prediction result.

[0124] In practice, after acquiring the traffic light status information and the real-time movement status information of vehicles within the traffic light area, the vehicle queue status of the traffic light area can be predicted based on the traffic light status information and real-time movement status information, thus obtaining the intersection traffic prediction result. The intersection traffic prediction result may include at least one of the following: the maximum queue length for each lane, the number of vehicles in the queue, the queue dissipation time, the estimated waiting time for the main vehicle, the estimated arrival time of the main vehicle at the end of the queue, and the recommended passage time for the main vehicle through the traffic light area.

[0125] Specifically, the real-time motion status information of vehicles within the traffic light area can be used as the initial state for prediction. The initial state for prediction can include information such as the current position, speed, acceleration, lane, distance to the vehicle in front, distance to the stop line, and whether the vehicle is at low speed or stopped. For target lanes related to the current direction of travel of the main vehicle, vehicles can be sorted according to their distance to the stop line to determine the following relationship between vehicles; for adjacent lanes or vehicles that may change lanes, the potential lane-changing relationship can also be determined based on vehicle position, heading angle, and lane boundary information.

[0126] Here, after determining the initial predicted state, the following behavior of the vehicle can be corrected based on the traffic light phase and corresponding remaining time in the traffic light status information. The traffic light phase can include the red light phase, green light phase, and yellow light phase, and the remaining time can include the remaining time of the red light, green light, or yellow light.

[0127] In one implementation, when the traffic light is red, the expected following distance for the target vehicle can be increased based on the remaining red light time. This simulates the behavior of vehicles releasing the accelerator early, coasting to decelerate, or gradually approaching the end of the queue when the remaining red light time is long. Specifically, when the remaining red light time is long and there is still a certain distance between the target vehicle and the vehicles in front of it, the vehicle usually does not continue to approach the stop line at a high speed, but decelerates in advance, resulting in a relatively smooth deceleration process. Therefore, a correction parameter related to the remaining red light time can be introduced into the following behavior model of the target vehicle, so that the expected following distance of the target vehicle increases with the increase of the remaining red light time, thereby reducing the predicted vehicle acceleration or increasing the predicted deceleration.

[0128] In another implementation, when the traffic light phase is green and the remaining green time is less than a preset time threshold, the expected speed of the target vehicle can be reduced to simulate the vehicle's behavior of abandoning acceleration, preparing to decelerate and stop, or avoiding overtaking as the green light is about to end. Specifically, if the remaining green time is short, vehicles located far from the stop line will find it difficult to pass through the intersection before the green light ends, even if they continue to accelerate, and their driving behavior tends to decelerate and wait for the next cycle. Therefore, the expected speed of the target vehicle can be reduced based on the remaining green time, making the speed changes of the target vehicle in subsequent simulations more consistent with actual driving behavior within the traffic light area.

[0129] In some implementations, when the traffic light is in the yellow phase, the travel tendency of the target vehicle can be determined based on the distance between the target vehicle and the stop line, its current speed, and its safe braking distance. If the target vehicle can safely cross the stop line, it can be predicted that the vehicle will continue to travel; if the target vehicle has difficulty safely crossing the stop line during the yellow light, it can be predicted that the vehicle will decelerate and stop.

[0130] Furthermore, after correcting the car-following behavior, the speed and position of vehicles within the traffic light area can be predicted step-by-step over a future forecast period based on the corrected car-following behavior. The future forecast period can be set according to the current signal cycle, the speed of the lead vehicle, the distance between the lead vehicle and the stop line, and task scheduling requirements. For example, the future forecast period can cover the remaining time of the current signal phase and the initial stage of the next signal phase, in order to predict whether a queue can form within the current cycle and the time required for the queue to dissipate after the green light turns on.

[0131] In the time-step simulation, the traffic light phase and remaining time of the phase can be updated at each predicted time step. For each vehicle, its acceleration in the next time step can be determined based on its distance from the vehicle in front, relative speed, current speed, lane location, and current traffic light status. Subsequently, the vehicle's speed and position in the next time step can be updated based on this acceleration. For vehicles behind in a queue, if the vehicle in front is moving slowly or stopped, and the current traffic light phase does not permit passage, the vehicle's predicted speed can gradually decrease, eventually stopping near the end of the queue. For vehicles in the green light lane, if the vehicle in front has already started and is gradually crossing the stop line, subsequent vehicles can start sequentially according to their start delay and following distance.

[0132] In some implementations, lane-changing behavior can also be used to predict vehicle queuing status. Specifically, it can be determined whether a vehicle needs to change lanes based on the vehicle's current lane, target direction of travel, traffic conditions in adjacent lanes, distance between the vehicle and the vehicle in front, and safe clearance between adjacent lanes. When a vehicle meets the lane-changing conditions, its current lane can be updated during the prediction process, and it can be included in the new lane's vehicle queuing relationship.

[0133] In some embodiments, the maximum queue length and queue dissipation time corresponding to the traffic light area can be determined based on the estimated vehicle speed and vehicle position. Specifically, for each lane, vehicles with speeds below a preset low-speed threshold, located in front of the stop line, and whose distance from the vehicles in front and behind them meets the queue spacing condition can be identified as queuing vehicles within the predicted time range. For vehicles that continuously meet the queuing conditions, the queue length of the corresponding lane at that predicted time can be determined.

[0134] The maximum queue length can be represented by the number of vehicles or by the spatial distance from the stop line to the end of the queue. For example, the number of vehicles continuously queuing in the same lane can be used as the queue length, and the spatial queue length can be determined based on the distance between the last vehicle in the queue and the stop line. For multi-lane traffic light areas, the maximum queue length for each lane can be determined separately, or the target queue length related to the passage of the main vehicle can be determined based on the lane where the main vehicle is located or the lane corresponding to the main vehicle's navigation route.

[0135] Furthermore, the queue dissipation time can be determined based on the starting process and the process of queuing vehicles crossing the stop line after the green light begins. In one implementation, the starting status and stop-line crossing status of queuing vehicles can be statistically analyzed step by step from the start time of the green light phase corresponding to the target lane until the last queuing vehicle starts and crosses the stop line; this time interval is then determined as the queue dissipation time. In another implementation, it can also be estimated based on the number of queuing vehicles, the starting delay of each vehicle, the saturation headway, and the time it takes for vehicles to cross the stop line.

[0136] In some implementations, the estimated waiting time for the lead vehicle can also be determined based on the current speed of the lead vehicle, the distance between the lead vehicle and the end of the queue, the maximum queue length of the target lane, and the queue dissipation time. Specifically, if the prediction indicates that the queue ahead has not yet dissipated when the lead vehicle arrives at the traffic light area, the estimated waiting time can be determined based on the time difference between the estimated arrival time of the lead vehicle at the end of the queue and the time when the queue has completely dissipated. If the prediction indicates that the queue has dissipated before the lead vehicle reaches the stop line and the traffic light is still in the permitted state, then the estimated waiting time for the lead vehicle can be determined to be short or that no stop is required.

[0137] In some embodiments, to improve the reliability of intersection traffic prediction results, the real-time vehicle motion state information continuously updated by the target cooperating vehicle and roadside units can be integrated during the time-step extrapolation process. That is, vehicle queue state prediction does not need to be based solely on the initial state at a single moment, but can be continuously updated as new data is acquired during the prediction process. When the deviation between the actual vehicle motion state and the predicted state is detected to be greater than a preset deviation threshold, the latest vehicle motion state and the latest traffic light state can be used as the initial state again to re-predict the vehicle queue state.

[0138] In some preferred embodiments, intersection traffic prediction results can be organized into structured prediction data. This structured prediction data may include target lane identification, current signal phase, remaining phase time, predicted maximum number of vehicles in the queue, predicted queue tail position, predicted queue dissipation time, estimated waiting time for the lead vehicle, estimated stopping position for the lead vehicle, and prediction confidence level. The prediction confidence level can be determined based on the completeness of the input data, vehicle status update frequency, traffic light status reliability, and the deviation between the actual and predicted states during the prediction process. When the prediction confidence level is low, measures such as adding cooperating vehicles, requesting supplementary data from roadside units, or shortening the prediction update cycle can be triggered.

[0139] The above scheme will now be described in conjunction with specific implementation methods.

[0140] The classic car-following model calculates the following car. At any moment acceleration This invention incorporates the effect of remaining traffic light time into the classic model to simulate the driver's anticipatory behavior. The improved model adds an adjustment term related to the remaining red light time when calculating the desired following distance.

[0141] in, Indicates the desired following distance. Indicates the minimum safe distance. Indicates the current speed of the following vehicle. Indicates the safe time interval. Indicates the speed difference between the front and rear vehicles. Indicates the maximum acceleration. Indicates comfortable deceleration. This represents the predictive braking coefficient, used to quantify a driver's tendency to adjust their following distance in advance based on the remaining time at a red light. Indicates the remaining time of the red light.

[0142] When the remaining time of the red light is long A larger value leads to a larger expected following distance, which in turn reduces the acceleration calculated by the model, simulating the driver's behavior of decelerating and coasting in advance.

[0143] Similarly, to simulate the behavior of some drivers who give up accelerating as the green light is about to end, the desired speed can be adjusted appropriately:

[0144] in, Indicates the original expected speed. This indicates an adjustment to the strength coefficient. Indicates the remaining time of the green light. This represents the time constant.

[0145] When the remaining time of the green light is very short Approaching 1 leads to a decrease in the expected speed, simulating the driver's behavior of abandoning acceleration and preparing to stop.

[0146] Using the improved car-following model described above, and combined with the lane-changing model, the vehicles ahead within the perception range of the roadside unit are analyzed step-by-step over time. Iterative deduction: Initialization, based on the current time The states of all vehicles are used as initial values. For each future time... Update traffic light phase and remaining time. For each vehicle, calculate its acceleration for the next moment using an improved car-following model, based on its distance from the vehicle in front, relative speed, and the current traffic light status. The system uses a lane-changing model to determine if a vehicle needs to change lanes; if so, it updates its current lane. The vehicle's speed and position are updated based on acceleration.

[0147]

[0148] Repeat the iteration until the preset prediction duration is reached. Based on the simulation results, key information was extracted: maximum queue length. For each lane, statistics are collected on speeds below a certain threshold during red light periods. And the maximum number of vehicles that stop consecutively at the intersection. Queue dissipation time. From the moment the green light turns on The time elapsed from the start of the queue until the last vehicle in the queue moves and crosses the stop line at the intersection. This time can be estimated based on the queue length.

[0149] in, It's the time to clear. It is the first Vehicle start-up delay time.

[0150] S106. Based on the intersection traffic prediction results, adjust the target cooperative vehicle set and the task execution node to obtain the cooperative perception and task scheduling results for the traffic light area.

[0151] In practice, after obtaining the intersection traffic prediction results, the target cooperative vehicle set and task execution nodes can be adjusted based on these results to obtain the cooperative perception and task scheduling results for the traffic light area. The cooperative perception and task scheduling results can include the adjusted target cooperative vehicle set, the adjusted task execution nodes, the cooperative perception data transmission strategy, the task offloading strategy, and control assistance information related to the primary vehicle's passage through the traffic light area.

[0152] Specifically, the intersection traffic prediction results can include information such as the maximum queue length in the traffic light area, the number of vehicles in the queue, the queue dissipation time, the estimated waiting time for the primary vehicle, the predicted position of the queue tail, the estimated stopping position of the primary vehicle, and whether the primary vehicle can pass through the intersection within the current signal cycle. Since these prediction results reflect the traffic conditions that the primary vehicle will face in the traffic light area, they can be used as environmental context information to feed back into the cooperative vehicle selection process and the task scheduling process, enabling the target cooperative vehicle set and task execution nodes to be dynamically adjusted according to the traffic conditions in the traffic light area.

[0153] In some implementations, the target cooperative vehicle set can be adjusted based on the intersection traffic prediction results. When the intersection traffic prediction results indicate that a vehicle queue will form in the area of ​​the traffic light ahead, or when the predicted tail of the queue is within a preset range in front of the main vehicle, the weight corresponding to the complementary contribution of perception vision in the cooperative value assessment can be increased.

[0154] For example, when predictions indicate that a long queue will form in the lane where the main vehicle is located before the red light ends, candidate cooperating vehicles located near the tail of the queue, in adjacent lanes, or upstream of the intersection can be prioritized to provide information such as queue length changes, vehicle deceleration status, and the position of the tail of the queue. When predictions indicate that oncoming or lateral areas at the intersection may affect the passage of the main vehicle, candidate cooperating vehicles capable of observing oncoming or lateral areas can be prioritized. When a large vehicle in front of the main vehicle obstructs its view, preventing the main vehicle from directly observing the stop line or the queue ahead, candidate cooperating vehicles located in front of or to the side of the obstruction and capable of supplementing information about the obstructed area can be prioritized.

[0155] Furthermore, when adjusting the target set of cooperative vehicles, the perception quality evaluation results of candidate cooperative vehicles that are about to stop and have a fixed field of vision can be reduced. Specifically, when the intersection traffic prediction results indicate that a candidate cooperative vehicle is about to stop in the queue, and its stopping position will result in its field of vision being fixed for a long time or being blocked by vehicles in front, the priority of this candidate cooperative vehicle in the subsequent selection of cooperative vehicles can be reduced. Conversely, for candidate cooperative vehicles that are expected to maintain a good field of vision, be able to observe the queue evolution process, or be able to cover the key blind spots of the main vehicle, their overall cooperative value can be increased.

[0156] In some embodiments, adjusting the target cooperative vehicle set may include adding, replacing, deleting, or retaining target cooperative vehicles. Specifically, if prediction results indicate that the existing target cooperative vehicle set cannot cover the predicted platoon tail position, the area near the intersection stop line, or key areas of adjacent lanes, vehicles that meet the conditions of complementary vision and link stability can be added from the candidate cooperative vehicles. If the link stability of a vehicle in the existing target cooperative vehicles decreases, is about to leave the communication coverage area, or has a highly overlapping perception field with other target cooperative vehicles, it can be removed from the target cooperative vehicle set, and a new candidate cooperative vehicle can be selected as a replacement vehicle. If the current target cooperative vehicle set can already cover the key areas required for prediction, and the data transmission volume does not exceed the current network bandwidth constraint, the target cooperative vehicle set can be kept unchanged.

[0157] Furthermore, when adjusting the target cooperative vehicle set, the cooperative sensing data transmission strategy can be determined by combining the current network environment information. When a long vehicle queue is predicted in the traffic light area and the available bandwidth is sufficient, the target cooperative vehicles can be allowed to upload higher-precision sensing data, such as target-level data, local trajectory data, or some raw sensing data. When the available bandwidth is insufficient or the link is congested, the number of target cooperative vehicles can be reduced, or the data transmission mode can be adjusted to a low-bandwidth mode, such as uploading only key summary information such as vehicle target boxes, vehicle speed, queue tail position, and vehicle status near the stop line.

[0158] In some implementations, the task execution nodes can be adjusted based on intersection traffic prediction results. When the intersection traffic prediction results indicate that the estimated waiting time for the main vehicle in the traffic light area is greater than a preset waiting time threshold, driving tasks that are not real-time and have a computational load exceeding a preset computational load threshold can be prioritized for execution on available edge nodes. Non-real-time driving tasks with high computational loads may include tasks such as next-segment speed planning, global path local optimization, multi-vehicle cooperative perception fusion, intersection global situation analysis, or high-precision map local update.

[0159] For example, if predictions indicate that the main vehicle will be waiting in line for a considerable time during the current red light cycle, the speed planning task for the next road segment or the computationally intensive perception fusion task can be offloaded to roadside units or edge computing nodes in advance. Edge nodes can complete calculations while the main vehicle is waiting and return the results to the main vehicle when it starts moving or is about to cross the intersection. This way, the main vehicle can directly perform driving control based on the returned task execution results after the green light turns green, avoiding response delays caused by initiating calculations after crossing the intersection.

[0160] Conversely, when the intersection traffic prediction results indicate a short expected waiting time for the primary vehicle, or that the primary vehicle can successfully pass through the intersection within the current green light cycle, the task execution node can be prioritized based on the task's real-time requirements and link stability. For driving tasks with high latency requirements and where the results need to be immediately used for vehicle control, the primary vehicle can be prioritized for local execution, or the target collaborative vehicle with the most stable link can be selected for execution. For tasks with high computational demands but where the results are used later, edge nodes can continue to be selected for execution.

[0161] In some embodiments, when prediction results indicate that the master vehicle is about to enter the vehicle queue or stop, the priority of edge nodes in task scheduling can be appropriately increased. Since the relative positional change between the master vehicle and roadside units is small when the master vehicle is stopped or traveling at low speed, the vehicle-to-infrastructure communication link typically exhibits good stability. In this case, scheduling computationally intensive tasks to roadside units is more likely to meet latency and reliability requirements. When prediction results indicate that the master vehicle is about to leave the coverage area of ​​a roadside unit or quickly passes through an intersection after a green light, the task offloading ratio to that roadside unit can be reduced, or the task execution results can be sent back to the master vehicle in advance to avoid communication link interruptions affecting task result acquisition.

[0162] Furthermore, when adjusting task execution nodes, hierarchical scheduling can be implemented based on task priority and time limits. For tasks with high real-time safety requirements, such as emergency obstacle avoidance, collision risk assessment, and short-term trajectory control, priority can be given to executing them locally on the main vehicle. For tasks related to cooperative perception and dependent on perception data from target cooperative vehicles, priority can be given to scheduling them to target cooperative vehicles or roadside units with relevant field of view and available computing power. For non-real-time tasks with high computational demands, scheduling can be performed on available edge nodes based on the main vehicle's expected waiting time and the resource status of edge nodes.

[0163] In some implementations, adjusting the task execution node may also include reselecting a backup execution node. When the link stability corresponding to the original task execution node is lower than a preset link threshold, the expected task completion time exceeds the maximum tolerable latency, the node's current load increases, or the node is about to leave the communication range, the task execution node can be re-determined based on the latest intersection traffic prediction results, the target cooperative vehicle resource status, and the available edge node resource status.

[0164] For example, if the original target collaborative vehicle's perception field of view is limited and the link quality deteriorates due to queuing and parking, the task can be switched to the roadside unit for execution; if the roadside unit is overloaded and a certain target collaborative vehicle still has a stable link and sufficient computing power, some tasks can be switched to that target collaborative vehicle for execution.

[0165] In some preferred embodiments, the results of adjusting the target cooperative vehicle set and the results of adjusting the task execution nodes can be jointly optimized. Specifically, when a candidate cooperative vehicle has both high perceptual complementarity value and high available computing resources, it can be preferentially selected into the target cooperative vehicle set, and some tasks related to its perceptual field of view can be scheduled to be executed by that vehicle.

[0166] For example, a target cooperative vehicle located near the rear of the platoon can not only sense the status of the vehicles at the rear of the platoon, but also has sufficient computing resources. Therefore, the target cooperative vehicle can perform tasks such as extracting the status of the rear of the platoon or tracking local targets, and then send the processed results to the master vehicle or roadside unit.

[0167] Furthermore, when available edge nodes possess strong computing capabilities and can aggregate perception data uploaded by multiple target cooperative vehicles, tasks such as multi-vehicle perception fusion, vehicle queue prediction, and intersection global situational awareness analysis can be scheduled to be executed on available edge nodes. Simultaneously, vehicles in the target cooperative vehicle set can provide perception data according to the data format and upload cycle issued by the edge nodes. This allows for a collaborative working mode where target cooperative vehicles provide perception data, edge nodes perform centralized computation, and the primary vehicle receives the decision results.

[0168] See Figure 2 The diagram shows a flowchart of another vehicle dynamic coordination method for traffic light areas provided in this disclosure embodiment. The method includes steps S201-S203, wherein: S201. Control the target cooperative vehicles in the target cooperative vehicle set to send cooperative perception data to the host vehicle or roadside unit.

[0169] S202. Control the task execution node to execute the driving task and return the task execution result to the master vehicle.

[0170] S203. Based on the intersection traffic prediction results and the task execution results, determine the driving control strategy of the main vehicle in the traffic light area, wherein the driving control strategy includes at least one of coasting to the end of the predicted queue, smoothly stopping, following the vehicle in front to start, and passing through the intersection at the planned speed.

[0171] In practical implementation, after obtaining the collaborative perception and task scheduling results for the traffic light area, collaborative perception data interaction and task scheduling can be performed based on these results. Specifically, the target collaborative vehicles in the adjusted target collaborative vehicle set can be controlled to send collaborative perception data to the master vehicle or roadside unit; the adjusted task execution nodes can be controlled to execute the corresponding driving tasks and return the task execution results to the master vehicle; the master vehicle can determine its driving control strategy within the traffic light area based on the intersection traffic prediction results, collaborative perception data, and task execution results. The driving control strategy may include at least one of the following: coasting to the predicted end of the queue, smooth stopping, following the vehicle in front, passing through the intersection at the planned speed, or adjusting the lane in advance.

[0172] Alternatively, the roadside unit may send collaborative sensing data; the adjusted task execution node will execute the corresponding driving task and return the task execution result to the master vehicle; the master vehicle can determine its driving control strategy within the traffic light area based on the intersection traffic prediction result, collaborative sensing data, and task execution result. The driving control strategy may include at least one of the following: coasting to the predicted end of the queue, smoothly stopping, following the vehicle in front, passing through the intersection at the planned speed, or adjusting the lane in advance.

[0173] This disclosure provides a vehicle dynamic cooperation method for traffic light areas. By comprehensively evaluating the perception contribution between candidate cooperative vehicles and the master vehicle, the visual complementarity between candidate cooperative vehicles, and the current network environment, the method dynamically determines the cooperative vehicle set. It also combines task characteristics, node resource status, and traffic light area vehicle queue prediction results to perform task scheduling and strategy adjustment, thereby reducing communication redundancy in the cooperative perception process, improving the adaptability and reliability of task offloading decisions, and enhancing the accuracy of the master vehicle's passage prediction and the stability of its driving decisions in traffic light areas.

[0174] Those skilled in the art will understand that, in the above-described method of the specific implementation, the order in which each step is written does not imply a strict execution order and does not constitute any limitation on the implementation process. The specific execution order of each step should be determined by its function and possible internal logic.

[0175] Based on the same inventive concept, this disclosure also provides a vehicle dynamic coordination device for traffic light areas, corresponding to the vehicle dynamic coordination method for traffic light areas. Since the principle of the device in this disclosure for solving the problem is similar to the vehicle dynamic coordination method for traffic light areas described above, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be described again.

[0176] Please see Figure 3 , Figure 3 This is a schematic diagram of a vehicle dynamic coordination device for traffic light areas provided in an embodiment of this disclosure. Figure 3 As shown in the figure, the vehicle dynamic coordination device 300 for traffic light areas provided in this embodiment includes: The data acquisition module 310 is used to acquire the status information of candidate cooperative vehicles around the master vehicle, the current network environment information, and the driving task information corresponding to the master vehicle.

[0177] The collaborative vehicle determination module 320 is used to evaluate the collaborative value of the candidate collaborative vehicles based on the candidate collaborative vehicle status information and the current network environment information, and to determine the set of target collaborative vehicles for participating in the master vehicle collaborative perception from the candidate collaborative vehicles.

[0178] The task node determination module 330 is used to determine the task execution node corresponding to the driving task based on the driving task information, the resource status information of each target cooperative vehicle in the target cooperative vehicle set, and the resource status information of available edge nodes.

[0179] The status information acquisition module 340 is used to acquire the status information of the traffic lights in the traffic light area and the real-time motion status information of the vehicles in the traffic light area when the main vehicle approaches the traffic light area.

[0180] Traffic prediction module 350 is used to predict the vehicle queue status of the traffic light area based on the traffic light status information and the real-time motion status information, and obtain the intersection traffic prediction result. The vehicle coordination and task scheduling module 360 ​​is used to adjust the target coordination vehicle set and the task execution node according to the intersection traffic prediction results, so as to obtain the coordination perception and task scheduling results for the traffic light area.

[0181] The processing flow of each module in the device and the interaction flow between each module can be referred to the relevant descriptions in the above method embodiments, and will not be detailed here.

[0182] This disclosure provides a vehicle dynamic coordination device for traffic light areas. By comprehensively evaluating the perception contribution between candidate cooperating vehicles and the master vehicle, the visual complementarity between candidate cooperating vehicles, and the current network environment, it dynamically determines the set of cooperating vehicles. It also combines task characteristics, node resource status, and traffic light area vehicle queue prediction results to perform task scheduling and strategy adjustment, thereby reducing communication redundancy in the cooperative perception process, improving the adaptability and reliability of task offloading decisions, and enhancing the accuracy of the master vehicle's passage prediction and the stability of its driving decisions in traffic light areas.

[0183] Corresponding to Figure 1 and Figure 2 In the vehicle dynamic coordination method for traffic light areas, this disclosure also provides an electronic device 400, such as... Figure 4 The diagram shown is a structural schematic of an electronic device 400 provided in an embodiment of this disclosure, including: Processor 41, memory 42, and bus 43; memory 42 is used to store execution instructions, including main memory 421 and external memory 422; the main memory 421, also called internal memory, is used to temporarily store the computational data in processor 41, as well as the data exchanged with external memory 422 such as hard disk. Processor 41 exchanges data with external memory 422 through main memory 421. When the electronic device 400 is running, processor 41 and memory 42 communicate through bus 43, enabling processor 41 to execute... Figure 1 and Figure 2 The steps of the vehicle dynamic coordination method for traffic light areas.

[0184] This disclosure also provides a computer-readable storage medium storing a computer program that, when executed by a processor, performs the steps of the vehicle dynamic coordination method for traffic light areas described in the above-described method embodiments. The storage medium can be a volatile or non-volatile computer-readable storage medium.

[0185] This disclosure also provides a computer program product, which includes computer instructions. When the computer instructions are executed by a processor, they can perform the steps of the vehicle dynamic coordination method for traffic light areas described in the above method embodiments. For details, please refer to the above method embodiments, which will not be repeated here.

[0186] Finally, it should be noted that the above-described embodiments are merely specific implementations of this disclosure, used to illustrate the technical solutions of this disclosure, and not to limit it. The protection scope of this disclosure is not limited thereto. Although this disclosure has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features, within the scope of the technology disclosed in this disclosure. Such modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this disclosure, and should all be covered within the protection scope of this disclosure. Therefore, the protection scope of this disclosure should be determined by the protection scope of the claims.

Claims

1. A vehicle dynamic coordination method for a signal light area, characterized by, include: Obtain the status information of candidate cooperative vehicles around the master vehicle, the current network environment information, and the driving task information corresponding to the master vehicle; Based on the candidate cooperative vehicle status information and the current network environment information, the cooperative value of the candidate cooperative vehicles is evaluated, and a set of target cooperative vehicles for participating in the master vehicle cooperative perception is determined from the candidate cooperative vehicles. Based on the driving task information, the resource status information of each target cooperative vehicle in the target cooperative vehicle set, and the resource status information of available edge nodes, the task execution node corresponding to the driving task is determined. When the main vehicle approaches the traffic light area, the traffic light status information of the traffic light area and the real-time motion status information of the vehicles in the traffic light area are acquired. Based on the traffic light status information and the real-time motion status information, the vehicle queue status of the traffic light area is predicted to obtain the intersection traffic prediction result. Based on the intersection traffic prediction results, the target cooperative vehicle set and the task execution node are adjusted to obtain the cooperative perception and task scheduling results for the traffic light area.

2. The method according to claim 1, characterized in that: The candidate cooperative vehicle status information includes at least one of the following: the candidate cooperative vehicle's position, speed, acceleration, heading angle, sensor configuration information, available computing resources, remaining battery power, available communication bandwidth, and link quality information with the host vehicle. The current network environment information includes at least one of the following: currently available bandwidth, link stability, communication latency, data transmission mode, and communication coverage. The driving task information includes at least one of the following: task type, task data volume, required computing resources, maximum tolerable latency, and task priority.

3. The method of claim 1, wherein, Based on the candidate cooperative vehicle status information and the current network environment information, the cooperative value of the candidate cooperative vehicles is evaluated, and a set of target cooperative vehicles for participating in the master vehicle cooperative perception is determined from the candidate vehicles, specifically including: Based on the positional relationship, perception capability, and link stability of the candidate cooperative vehicle relative to the master vehicle, the first relationship evaluation result corresponding to the candidate cooperative vehicle is determined. Based on the complementary perception field of vision between candidate cooperative vehicles and the set of selected cooperative vehicles, the second relationship evaluation result corresponding to the candidate cooperative vehicle is determined. Based on the first relationship evaluation results and the second relationship evaluation results, the comprehensive collaborative value of the candidate collaborative vehicles is determined; Based on the comprehensive collaborative value and the current network environment information, the target collaborative vehicle set is determined from the candidate collaborative vehicles.

4. The method of claim 3, wherein, The first relationship evaluation result and the second relationship evaluation result are determined based on the following steps: The perception quality of the candidate cooperative vehicles is determined based on their sensor configuration information and current perception mode. Based on the spatial distance, relative speed, relative position, and direction of motion between the candidate cooperative vehicle and the host vehicle, the link stability between the candidate cooperative vehicle and the host vehicle is determined, and based on the perceived quality, the spatial distance, and the link stability, the first relationship evaluation result is determined. Based on the location and sensor configuration information of the candidate cooperative vehicles, the perception field of view corresponding to the candidate cooperative vehicles is determined; The overlap of the perception field of view corresponding to the candidate cooperative vehicle with the perception field of view corresponding to each of the selected cooperative vehicles in the set of selected cooperative vehicles is analyzed to obtain the field of view complementarity information, and the second relationship evaluation result is determined based on the field of view complementarity information.

5. The method of claim 3, wherein, Based on the comprehensive collaborative value and the current network environment information, the target collaborative vehicle set is determined from the candidate collaborative vehicles, specifically including: Based on the currently available bandwidth, determine the upper limit of the number of vehicles allowed to participate in collaborative sensing and the upper limit of the amount of collaborative sensing data transmitted. Initialize the selected set of cooperative vehicles; Determine the marginal collaboration gain of each unselected candidate collaborative vehicle after it has been added to the set of selected collaborative vehicles; Candidate cooperative vehicles whose marginal cooperative gains meet preset screening conditions are selected and added to the set of selected cooperative vehicles. When the data transmission volume corresponding to the selected set of cooperative vehicles reaches the upper limit of the cooperative sensing data transmission volume, or when the marginal cooperative gain corresponding to the unselected candidate cooperative vehicles is lower than the preset gain threshold, the filtering stops, and the current set of selected cooperative vehicles is determined as the target set of cooperative vehicles.

6. The method of claim 1, wherein, Based on the driving task information, the resource status information of each target cooperative vehicle in the target cooperative vehicle set, and the resource status information of available edge nodes, the task execution node corresponding to the driving task is determined, specifically including: The driving task information is processed to obtain task feature information; Obtain the resource status information corresponding to the master vehicle, each target collaborative vehicle in the target collaborative vehicle set, and the available edge nodes; Based on the task feature information and the resource status information, construct task unloading decision input information; The task unloading decision input information is input into a pre-trained intelligent decision model to obtain a selection score corresponding to at least one candidate execution node; Based on the selection score, the task execution node is determined from the master vehicle, the target collaborative vehicles in the target collaborative vehicle set, and the available edge nodes.

7. The method according to claim 6, characterized in that, The intelligent decision-making model is used to output a selection score for the corresponding candidate execution node based on task characteristics, node computing resources, node communication resources, node remaining power, link stability, and task latency constraints. When the task execution node is the master vehicle, it controls the master vehicle to execute the driving task locally. When the task execution node is the target cooperative vehicle, the data corresponding to the driving task is sent to the target cooperative vehicle for execution via the vehicle-to-vehicle communication link. When the task execution node is an available edge node, the data corresponding to the driving task is sent to the available edge node for execution via the vehicle-to-infrastructure communication link.

8. The method according to claim 1, characterized in that, Based on the traffic light status information and the real-time motion status information, the vehicle queue status in the traffic light area is predicted to obtain the intersection traffic prediction result, specifically including: The real-time motion status information of vehicles within the traffic light area is used as the initial state; The following behavior of the vehicle is corrected based on the phase of the traffic light and the corresponding remaining time. Based on the corrected car-following behavior, the speed and position of vehicles in the traffic light area are extrapolated step-by-step over the future prediction period. Based on the vehicle speed and vehicle position obtained from the simulation, the maximum queue length and queue dissipation time corresponding to the traffic light area are determined as the intersection traffic prediction result. Specifically, when the traffic light phase is red, the expected following distance of the target vehicle is increased based on the remaining red light time; when the traffic light phase is green and the remaining green light time is less than a preset time threshold, the expected speed of the target vehicle is reduced.

9. The method according to claim 1, characterized in that, Based on the intersection traffic prediction results, the target cooperative vehicle set and the task execution node are adjusted, specifically including: When the intersection traffic prediction result indicates that a vehicle queue will form in the traffic light area ahead, the weight corresponding to the complementary contribution of perception vision in the cooperative value assessment is increased, and the target cooperative vehicle set is re-determined based on the adjusted weight. Candidate cooperative vehicles that can observe the tail of the queue, the opposite area of ​​the intersection, the adjacent lane area, or the blind spot of the main vehicle are preferentially selected to be added to the target cooperative vehicle set. When the intersection traffic prediction result indicates that the expected waiting time of the main vehicle in the traffic light area is greater than the preset waiting time threshold, driving tasks that are not real-time and have a computational load greater than the preset computational load threshold will be prioritized for scheduling to available edge nodes for execution.

10. A vehicle dynamic coordination device for traffic light areas, characterized in that, include: The data acquisition module is used to acquire the status information of candidate cooperative vehicles around the master vehicle, the current network environment information, and the driving task information corresponding to the master vehicle. The collaborative vehicle determination module is used to evaluate the collaborative value of the candidate collaborative vehicles based on the candidate collaborative vehicle status information and the current network environment information, and to determine the set of target collaborative vehicles for participating in the master vehicle collaborative perception from the candidate collaborative vehicles. The task node determination module is used to determine the task execution node corresponding to the driving task based on the driving task information, the resource status information of each target cooperative vehicle in the target cooperative vehicle set, and the resource status information of available edge nodes. The status information acquisition module is used to acquire the status information of the traffic lights in the traffic light area and the real-time motion status information of the vehicles in the traffic light area when the main vehicle approaches the traffic light area. The traffic prediction module is used to predict the vehicle queue status of the traffic light area based on the traffic light status information and the real-time motion status information, and obtain the intersection traffic prediction result. The vehicle coordination and task scheduling module is used to adjust the target coordination vehicle set and the task execution node according to the intersection traffic prediction results, so as to obtain the coordination perception and task scheduling results for the traffic light area.

Images