A Collaborative Lane Changing Decision-Making Method and System Based on Digital Twin and Dynamic Safety Boundaries

Abstract

This invention provides a lane-changing cooperative decision-making method and system based on digital twins and dynamic safety boundaries, relating to the fields of intelligent driving and vehicle-road cooperative technology. The invention constructs a vehicle-level digital twin mirror system to synchronize the status of the vehicle with surrounding vehicles in real time; after detecting lane-changing intentions, it hierarchically filters candidate solutions in the twin environment, accelerates the generation of a comprehensive disturbance score, and selects the optimal baseline solution; it calculates a dynamic safety boundary including hard and soft boundaries, and sends virtual trajectory commitment data packets to surrounding vehicles through a request-confirmation handshake mechanism, forming bidirectional implicit cooperation; during lane-changing execution, it dynamically compares the actual trajectory with the committed trajectory, corrects the safety boundary based on deviations, and dynamically terminates the lane-changing operation based on the completion rate; after lane-changing is completed, it de-identifies and stores event records, and optimizes weight coefficients through federated learning. This invention achieves multi-vehicle cooperation without real-time game theory, and has the advantages of low latency, high security, traceability, and privacy protection.

Description

Technical Field

[0001] This invention relates to the field of intelligent driving and vehicle-road cooperation technology, specifically to a lane-changing cooperative decision-making method and system based on digital twins and dynamic safety boundaries. Background Technology

[0002] With the rapid development of intelligent connected vehicle technology, lane-changing decision-making has evolved from single-vehicle perception to multi-vehicle collaboration. As a high-frequency operation during driving, the quality of lane-changing decisions directly impacts traffic safety and efficiency. Current technologies primarily rely on real-time information exchange and online game theory calculations. This involves vehicles acquiring the status of surrounding vehicles through vehicle-to-everything (V2X) communication and then calculating the optimal strategy using game theory or reinforcement learning. While these methods can achieve collaboration to some extent, they still suffer from the following technical limitations: First, real-time game theory computation places extremely high demands on communication latency and computing resources. In high-speed moving scenarios, vehicle states change rapidly, requiring online solutions to Nash equilibria or multi-round iterative optimization, resulting in high computational complexity and making it difficult to guarantee millisecond-level real-time decision-making. When packet loss or delay occurs in communication, the state information of the game model becomes distorted, significantly reducing the reliability of decisions. Especially in scenarios such as urban expressways or highways, where vehicle spacing is small and relative speeds are high, decision delays can directly lead to collision risks.

[0003] Secondly, existing collaborative methods are mostly two-sided interactions, making it difficult to handle complex multi-vehicle scenarios. When a lane-changing vehicle simultaneously affects multiple surrounding vehicles, two-sided game models cannot effectively aggregate the interests of all parties, easily leading to neglecting some aspects for others. For example, a lane-changing vehicle may simultaneously affect vehicles behind in the target lane, vehicles in front in the original lane, and vehicles in adjacent lanes. Existing methods often only consider a single interaction object or perform simple weighted processing, resulting in insufficient group adaptability in decision-making. This locally optimal decision-making approach may trigger cascading braking, traffic flow oscillations, and even induce traffic accidents.

[0004] Secondly, the collaborative decision-making results lack verifiability and accountability mechanisms. Existing methods only exchange intentions and states between vehicles, without creating a traceable record of commitments. When lane-changing behavior leads to risks or accidents, it is difficult to clearly define the boundaries of responsibility for each party, hindering the construction of a long-term, reliable vehicle collaboration environment. Furthermore, the lack of recording and verification mechanisms for the decision-making process means that algorithm optimization lacks effective data support, preventing the formation of a closed loop for continuous improvement.

[0005] Furthermore, existing methods tend to have a simplistic definition of safety boundaries, often employing fixed thresholds or judgments based on a single physical quantity. For example, relying solely on collision time or minimum safe distance for safety assessment ignores the differences in safety requirements across various driving scenarios and lacks mechanisms to address the unpredictable behavior of surrounding vehicles. When surrounding vehicles deviate from expected behavior, existing methods lack effective early warning and emergency response mechanisms.

[0006] Therefore, there is an urgent need to design a lane-changing collaborative decision-making method that can achieve effective collaboration in multi-vehicle scenarios without relying on real-time game theory calculations, while also possessing traceability and verifiability. Summary of the Invention

[0007] The purpose of this invention is to provide a lane-changing collaborative decision-making method and system based on digital twins and dynamic security boundaries, so as to solve at least one technical problem existing in the prior art.

[0008] On one hand, this invention provides a lane-changing collaborative decision-making method based on digital twins and dynamic safety boundaries, including the following steps: Step S1: Construct a vehicle-level digital twin mirror system, synchronize the dynamic status data of the vehicle and surrounding vehicles in real time through vehicle-to-everything (V2X) communication, and establish a simulation environment in virtual space that is consistent with the physical environment to form a digital twin environment; Step S2: When the vehicle's lane-changing intention is detected, generate a hierarchical selection of candidate lane-changing schemes in the digital twin environment, including pre-screening the candidate start time based on collision time and minimum predicted distance, eliminating schemes that do not meet the safety threshold, and performing accelerated simulation on the remaining schemes to form a candidate scheme set; Step S3: Perform accelerated simulation on each scheme in the candidate scheme set in the digital twin environment, predict the degree of disturbance to the vehicle, surrounding vehicles, and local traffic flow after the scheme is executed, and generate a comprehensive disturbance score corresponding to each scheme; Step S4: Based on the comprehensive disturbance score, select schemes that have an impact on traffic flow. The candidate scheme with the smallest impact is used as the benchmark execution scheme, and the dynamic safety boundary formed by the benchmark execution scheme during the deduction process is calculated. The dynamic safety boundary includes two levels: hard boundary and soft boundary. The hard boundary is calculated based on the minimum safe following distance standard, and the soft boundary is generated based on the hard boundary superimposed with a safety margin. Step S5: The dynamic safety boundary is used as a cooperative constraint parameter and broadcast to surrounding vehicles within the communication range through the vehicle network. A two-way commitment is established using a request-confirmation handshake mechanism, and a virtual trajectory commitment data packet containing lane-changing time window, expected trajectory characteristics, and liability clause fields is sent to surrounding vehicles. Step S6: The vehicle performs a lane-changing operation under the hard constraint of the broadcast dynamic safety boundary. During the execution, the deviation value between the actual trajectory and the virtual trajectory commitment signal is continuously compared. The safety boundary parameters are corrected in real time according to the deviation value, and the abort threshold is dynamically adjusted according to the lane-changing completion degree. Step S7: After the vehicle completes the lane-changing, the lane-changing event record is anonymized and stored as a training sample for subsequent decision optimization.

[0009] Furthermore, step S1 includes: Step S101, deploying a digital twin engine on an edge computing node or in the cloud, adopting a layered computing architecture, and allocating an independent virtual mirror space for each vehicle connected to the system; Step S102, acquiring dynamic status data of the vehicle and surrounding vehicles in real time through vehicle-to-everything (V2X) communication, and writing the acquired dynamic status data into the corresponding virtual mirror space, so that the synchronization error between the virtual mirror and the physical entity is controlled within a preset accuracy range; Step S103, constructing a static background layer containing road geometry and traffic sign information in the virtual space as the basic environment for simulation and deduction; Step S104, setting a communication timeout handling mechanism for communication anomalies, including: when the status data of a vehicle is continuously missing for more than a first time limit, reducing the dependence weight of the vehicle in subsequent decisions; when the missing data exceeds a second time limit, terminating the collaborative dependence on the vehicle and switching to an independent decision-making mode based on onboard sensor information.

[0010] Further, step S2 includes: step S201, detecting the driver's lane-changing intention through a steering wheel angle sensor or turn signal; if the turn signal is activated or the steering wheel angle exceeds a preset threshold, triggering a candidate lane-changing scheme generation process; step S202, performing pre-screening, which includes first-level screening and second-level screening: the first-level screening is: based on the current time, generating multiple candidate sets of lane-changing start times with different delay times; for each candidate start time, estimating the collision time between the vehicle and the vehicle behind in the target lane, as well as the minimum predicted distance between the vehicle and surrounding vehicles; classifying the start times according to a preset collision time threshold, marking start times with collision times greater than or equal to the first time threshold as safe, marking start times with collision times between the second time threshold and the first time threshold as requiring caution, and marking start times with collision times less than the second time threshold as requiring caution. The starting time of the threshold is marked as dangerous; at the same time, if the minimum prediction distance is less than the hard boundary, it is directly marked as dangerous; all starting times marked as dangerous are eliminated, and the two types of starting times, safe and requiring caution, are retained to enter the second level of screening; wherein, the second time threshold is less than the first time threshold; the second level of screening is as follows: for each starting time that passes the first level of screening, a variety of lateral trajectory shapes with different degrees of aggression and a variety of longitudinal acceleration curves are generated, and the two are combined to form a candidate scheme set; wherein, the lateral trajectory shapes with different degrees of aggression correspond to different lane change durations, and the longitudinal acceleration curves include at least one of acceleration, constant speed, and deceleration; step S203, each candidate scheme in the candidate scheme set is sent to the digital twin engine, and multiple schemes are simultaneously accelerated for simulation and deduction using a parallel computing architecture, and the overall decision delay is controlled within a preset time range.

[0011] Furthermore, the generation of the comprehensive disturbance score for each scheme in step S3 includes: Step S301: In the digital twin environment, starting from the current moment, import each candidate scheme in the candidate scheme set into the simulation engine and perform accelerated simulation at a rate higher than the real-time speed for a predetermined time window; Step S302: During the simulation, record the disturbance indices, which include: the minimum collision time between the vehicle and the vehicle in front, the minimum collision time between the vehicle and the vehicle behind in the target lane, the maximum deceleration of the vehicle behind in the target lane to avoid the vehicle, the speed gain of the vehicle in front in the original lane due to the vehicle leaving, and the standard deviation of the acceleration change rate of surrounding vehicles; Step S303: Normalize each disturbance index and linearly map it to the [0,1] interval; Step S304: Dynamically determine the weight coefficient of each index according to the traffic flow density, and sum the normalized indices by weight to obtain the comprehensive disturbance score for each candidate scheme; wherein, the lower the score, the smaller the impact of the scheme on the traffic flow.

[0012] Furthermore, the formula for calculating the comprehensive disturbance score is as follows: ; In the formula, , , , and These are the weighting coefficients. ; This is the normalized value of the minimum collision time between the vehicle and the vehicle in front. This is the normalized value of the minimum collision time between the vehicle and the vehicle following in the target lane; The normalized value of the maximum deceleration of the following vehicle in the target lane to avoid the following vehicle; This is the normalized value of the speed gain gained by the vehicle in front in the original lane due to leaving the lane. This is the normalized value of the standard deviation of the rate of change of acceleration of surrounding vehicles; It is a constant; The weighting coefficients are dynamically adjusted based on traffic flow density, including: reducing the weight of the maximum deceleration term of the vehicle following in the target lane when the traffic flow density is lower than a first density threshold; and increasing the weight of the collision time term and the maximum deceleration term of the vehicle following in the target lane when the traffic flow density is higher than a second density threshold.

[0013] Further, step S4 includes: step S401, selecting the scheme with the lowest comprehensive disturbance score from the candidate scheme set as the benchmark execution scheme; step S402, calculating the insurmountable physical limit safety distance as a hard boundary based on the minimum safe following distance standard and combined with the vehicle's maximum braking capacity; step S403, taking a preset multiple of the hard boundary as the longitudinal warning distance to generate a longitudinal soft boundary, and combining it with the lane width parameter to generate a lateral safety envelope as a lateral soft boundary; step S404, combining the hard boundary and the soft boundary to form a dynamic safety boundary function that changes with time; wherein, the soft boundary is located outside the hard boundary in space.

[0014] Further, step S5 includes: step S501, obtaining the expected trajectory data contained in the benchmark execution scheme, wherein the expected trajectory data is determined by the lateral trajectory shape and longitudinal acceleration curve of the candidate scheme generated in step S2, including the expected lane change time window and the expected trajectory point sequence; step S502, encapsulating the dynamic safety boundary and the expected trajectory data into a virtual trajectory commitment data packet, wherein the virtual trajectory commitment data packet includes a contract unique identifier, an initiator vehicle desensitization identifier, the expected lane change time window, the expected trajectory point sequence, longitudinal safety distance hard boundary and soft boundary parameters, lateral safety envelope soft boundary parameters, and liability clause fields; step S503, through vehicle-to-everything (V2X) communication... The network communication channel sends the virtual trajectory commitment data packet to the cooperating vehicle and broadcasts a notification message to surrounding vehicles; Step S504: Receive confirmation information from the cooperating vehicle within a predetermined time, indicating acceptance of the commitment; Step S505: If confirmation information is received, the cooperating vehicle uses the virtual trajectory commitment data packet as an input constraint for its own decision-making module and adjusts its driving strategy outside the safety boundary, forming implicit cooperation with two-way confirmation; If no confirmation information is received, the initiator chooses to resend or abandon the scheme; If no confirmation information is received after a preset number of attempts, the scheme is automatically abandoned, and the system switches to a pre-calculated, pending-activation suboptimal candidate scheme; For partial confirmation, the safety boundary is adjusted before continuing execution.

[0015] Further, step S6 includes: step S601, during lane change, collecting actual motion state data of the vehicle at a preset frequency, including actual position, speed, and acceleration; step S602, comparing the actual trajectory with the expected trajectory points in the virtual trajectory commitment data packet point by point, and calculating position deviation and speed deviation; step S603, if the position deviation does not exceed a predetermined deviation threshold, maintaining the current dynamic safety boundary; if the position deviation exceeds the predetermined deviation threshold, triggering safety boundary correction: using the actual position as a new starting point, recalculating the dynamic safety boundary within the remaining time window, smoothly transitioning to the originally committed endpoint state, and rebroadcasting the corrected dynamic safety boundary to the surrounding area. Vehicle; Step S604: Dynamically adjust the abort threshold based on lane change completion, including: when the lane change completion is less than a first predetermined completion threshold and the position deviation continues to exceed the predetermined deviation threshold for a predetermined number of times, trigger the lane change abort process; when the lane change completion is greater than a second predetermined completion threshold, even if the position deviation exists, the lane change is forcibly completed, and only the dynamic safety boundary is adjusted; when the lane change completion is between the first predetermined completion threshold and the second predetermined completion threshold, determine whether to abort based on the degree of deviation and the remaining space; the lane change completion is defined as 1 - (distance between the current lateral position and the center line of the target lane / distance between the original lane center line and the center line of the target lane).

[0016] Further, step S7 includes: step S701, after lane changing is completed, generating a lane changing event record, the record including at least the selected candidate scheme identifier, initial dynamic safety boundary parameters, and actual execution trajectory point sequence; step S702, collecting response data of surrounding vehicles during the lane changing process, the response data including at least the acceleration change record of each vehicle, whether emergency braking occurred, and whether evasive action was performed; step S703, desensitizing vehicle identifiers to remove personally identifiable information, and storing the desensitized event record in a local encrypted storage area; step S704, using federated learning technology, optimizing the weight coefficients used to generate the comprehensive disturbance score based on the stored event record, and improving the prediction accuracy of the digital twin environment; wherein, only the weight coefficients are uploaded to the cloud for aggregation, and the original trajectory data is not stored locally.

[0017] Secondly, embodiments of the present invention provide a lane-changing collaborative decision-making system based on digital twins and dynamic safety boundaries, implemented using the aforementioned lane-changing collaborative decision-making method based on digital twins and dynamic safety boundaries. The system includes: a vehicle-level digital twin mirror system construction module, suitable for constructing a vehicle-level digital twin mirror system, which synchronizes the dynamic state data of the vehicle and surrounding vehicles in real time through vehicle-to-everything (V2X) communication, establishing a simulation environment in virtual space that maintains consistency with the physical environment, thus forming a digital twin environment; a candidate scheme set generation module, suitable for generating graded and filtered candidate lane-changing schemes in the digital twin environment when a lane-changing intention of the vehicle is detected, including pre-screening the candidate start time based on collision time and minimum prediction distance, eliminating schemes that do not meet safety thresholds, and performing accelerated simulation on the remaining schemes to form a candidate scheme set; a comprehensive disturbance score generation module, suitable for performing accelerated simulation on each scheme in the candidate scheme set in the digital twin environment, predicting the disturbance degree to the vehicle, surrounding vehicles, and local traffic flow after the scheme is executed, and generating a comprehensive disturbance score corresponding to each scheme; and safety boundary generation. The system includes a module for selecting candidate schemes with the least impact on traffic flow based on the comprehensive disturbance score, and calculating the dynamic safety boundary formed during the simulation process of the benchmark execution scheme. The dynamic safety boundary includes two levels: hard boundary and soft boundary. The hard boundary is calculated based on the minimum safe following distance standard, and the soft boundary is generated based on the hard boundary plus a safety margin. The cooperative communication module is suitable for broadcasting the dynamic safety boundary as a cooperative constraint parameter to surrounding vehicles within the communication range via the vehicle network, and establishing a two-way commitment using a request-confirmation handshake mechanism, sending a virtual trajectory commitment data packet containing lane-changing time window, expected trajectory characteristics, and liability clause fields to surrounding vehicles. The execution and correction module is suitable for the vehicle to perform lane-changing operations under the hard constraints of the broadcast dynamic safety boundary. During the execution process, the deviation value between the actual trajectory and the virtual trajectory commitment signal is continuously compared, and the safety boundary parameters are corrected in real time according to the deviation value. The termination threshold is also dynamically adjusted according to the lane-changing completion degree. The data storage module is suitable for anonymizing and storing the lane-changing event record after the vehicle completes the lane-changing operation, as a training sample for subsequent decision optimization.

[0018] Thirdly, embodiments of the present invention also provide an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the above-described lane-changing collaborative decision-making method based on digital twins and dynamic security boundaries.

[0019] Fourthly, embodiments of the present invention also provide a readable storage medium, which, when the instructions in the storage medium are executed by the processor of an electronic device, enables the electronic device to execute the above-described lane-changing collaborative decision-making method based on digital twins and dynamic security boundaries.

[0020] The beneficial effects of this invention are as follows: 1. This invention constructs a digital twin mirror system, establishing a simulation environment in virtual space that is consistent with the physical environment. It migrates the lane-changing decision-making process from the physical world to the virtual space for accelerated simulation, allowing vehicles to fully anticipate the impact of multiple candidate scenarios on traffic flow before making a decision. Through a hierarchical screening strategy, obviously infeasible scenarios are first filtered out based on rapidly calculable indicators such as collision time. Then, the remaining scenarios are subjected to high-precision parallel simulation, avoiding the high latency and complexity of real-time game theory calculations. This achieves a decision delay of less than 1.5 seconds, significantly improving the reliability and real-time performance of the decision-making process.

[0021] 2. This invention achieves a paradigm shift from active game theory to trusted implicit collaboration by introducing dual security constraints of hard and soft boundaries. Employing a request-confirmation two-way handshake mechanism, surrounding vehicles do not need to participate in real-time negotiation; they only need to adhere to the security boundaries defined by the commitment to form collaboration. This significantly reduces communication overhead and interaction complexity. Simultaneously, digital signature technology records trajectory deviations and boundary intrusion events, providing objective data support for liability determination.

[0022] 3. This invention uses a dynamic abort threshold and emergency handling mechanism based on lane change completion. It defines lane change completion as the complement of the ratio of the distance between the current lateral position and the center line of the target lane to the distance between the original lane center line and the center line of the target lane. Based on the completion, it sets tiered handling rules: the lane change can be aborted when the completion is less than 30%, and is forced to complete when it is greater than 70%. The intermediate range is comprehensively judged to effectively deal with various edge cases during the execution process, which significantly improves the robustness and safety of the system.

[0023] 4. This invention uses data anonymization and federated learning techniques to anonymize lane change event records and store them locally. It then uses federated learning techniques for model training, uploading only model parameters or gradient information to the cloud for aggregation. The original trajectory data does not leave the local machine, achieving a balance between algorithm optimization needs and user privacy protection, and complies with relevant data security regulations.

[0024] 5. This invention provides valuable training samples for the continuous optimization of subsequent decision-making models by fully recording the entire process data of lane-changing events, including candidate scheme deduction data, dynamic safety boundary evolution curves, actual execution trajectories, and response data of surrounding vehicles. This enables the system to learn from historical experience, continuously improve decision quality, and achieve the effect of mutual promotion and co-evolution between the digital twin environment and the physical world. Attached Figure Description

[0025] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0026] Figure 1 This is a flowchart of a lane-changing collaborative decision-making method based on digital twins and dynamic security boundaries provided in Embodiment 1 of the present invention.

[0027] Figure 2 This is a schematic diagram of a hierarchical screening and deduction process provided in Embodiment 1 of the present invention.

[0028] Figure 3 This is a schematic diagram of the interaction between dynamic security boundary and virtual trajectory commitment provided in Embodiment 1 of the present invention.

[0029] Figure 4 This is a schematic diagram of a lane-changing collaborative decision-making system based on digital twins and dynamic security boundaries provided in Embodiment 2 of the present invention.

[0030] Figure 5 This is a partial block diagram of the electronic device provided in Embodiment 3 of the present invention. Detailed Implementation

[0031] Before discussing the exemplary embodiments in more detail, it should be noted that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although the flowcharts describe the operations as sequential processes, many of these operations can be performed in parallel, concurrently, or simultaneously. Furthermore, the order of the operations can be rearranged. The process can be terminated when its operation is completed, but may also have additional steps not included in the figures. The process can correspond to a method, function, procedure, subroutine, subroutine, etc.

[0032] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0033] The present invention will now be described in detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0034] Example 1 The specific implementation method is as follows: like Figure 1 The diagram shown is a flowchart of a lane-changing collaborative decision-making method based on digital twins and dynamic security boundaries provided by the present invention.

[0035] As an example, the process includes the following steps: Step S1: Construct a vehicle-level digital twin mirror system, synchronizing the dynamic status data of the vehicle and surrounding vehicles in real time through vehicle-to-everything (V2X) communication, and establishing a simulation environment in virtual space that is consistent with the physical environment, thus forming a digital twin environment; Step S2: When the vehicle's lane-changing intention is detected, generate a tiered selection of candidate lane-changing schemes in the digital twin environment, including pre-screening the candidate start time based on collision time and minimum predicted distance, eliminating schemes that do not meet safety thresholds, and performing accelerated simulation on the remaining schemes to form a candidate scheme set; Step S3: Perform accelerated simulation on each scheme in the candidate scheme set in the digital twin environment, predicting the degree of disturbance to the vehicle, surrounding vehicles, and local traffic flow after the scheme is executed, and generating a comprehensive disturbance score for each scheme; Step S4: Based on the comprehensive disturbance score, select the candidate scheme with the least impact on traffic flow as the benchmark execution scheme. The dynamic safety boundary formed during the simulation of the benchmark execution scheme is calculated. The dynamic safety boundary includes two levels: hard boundary and soft boundary. The hard boundary is calculated based on the minimum safe following distance standard, and the soft boundary is generated based on the hard boundary plus a safety margin. Step S5: The dynamic safety boundary is used as a cooperative constraint parameter and broadcast to surrounding vehicles within the communication range through the vehicle network. A two-way commitment is established using a request-confirmation handshake mechanism. A virtual trajectory commitment data packet containing lane-changing time window, expected trajectory characteristics, and liability clause fields is sent to surrounding vehicles. Step S6: The vehicle performs a lane-changing operation under the hard constraint of the broadcast dynamic safety boundary. During the execution, the deviation value between the actual trajectory and the virtual trajectory commitment signal is continuously compared. The safety boundary parameter is corrected in real time according to the deviation value, and the abort threshold is dynamically adjusted according to the lane-changing completion degree. Step S7: After the vehicle completes the lane-changing, the lane-changing event record is anonymized and stored as a training sample for subsequent decision optimization.

[0036] In some feasible implementations, step S1 includes: Step S101, deploying a digital twin engine on an edge computing node or in the cloud, adopting a layered computing architecture, and allocating an independent virtual mirror space for each vehicle connected to the system; Step S102, acquiring dynamic status data of the vehicle and surrounding vehicles in real time through vehicle-to-everything (V2X) communication, and writing the acquired dynamic status data into the corresponding virtual mirror space, so that the synchronization error between the virtual mirror and the physical entity is controlled within a preset accuracy range; Step S103, constructing a static background layer containing road geometry and traffic sign information in the virtual space as the basic environment for simulation and deduction; Step S104, setting a communication timeout handling mechanism for communication anomalies, including: when the status data of a vehicle is continuously missing for more than a first time limit, reducing the dependence weight of the vehicle in subsequent decisions; when the missing data exceeds a second time limit, terminating the collaborative dependence on the vehicle and switching to an independent decision-making mode based on onboard sensor information.

[0037] Preferably, step S1 includes: deploying a digital twin engine on an edge computing node or in the cloud, adopting a layered computing architecture: simple scenarios are directly processed by vehicle-side chips (such as Orin, Xavier), while complex scenarios are supported by parallel computing of edge / cloud GPUs to simulate multiple scenarios simultaneously; allocating an independent virtual mirror space for each vehicle connected to the system, which includes a vehicle motion state model, a sensor data cache, and a simulation simulation interface. Real-time synchronization of dynamic state data; receiving real-time motion state data of the vehicle and surrounding vehicles at a frequency of 10 Hz via dedicated short-range communication or cellular vehicle-to-everything (V2X) channels, the state data includes latitude and longitude coordinates, speed, acceleration, heading angle, and yaw rate; the surrounding vehicles are networked vehicles within the communication range that have potential interaction relationships with the vehicle, typically including vehicles in front in the same lane, vehicles behind in the target lane, and vehicles in adjacent lanes. Kalman filtering is used to fuse and smooth the received multi-source data, eliminating sensor noise and communication jitter. The fused state data is written into the virtual mirror space of the corresponding vehicle, ensuring that the state synchronization error between the virtual mirror and the physical entity is controlled within a preset threshold range, with a position error not exceeding 0.2 meters and a speed error not exceeding 0.5 km / h. A high-precision road map model is constructed in the virtual space as a static background layer for simulation. The high-precision road map model includes lane line geometric parameters, speed limit information, and traffic sign locations. In case of communication interruption or state synchronization failure, the system sets a communication timeout counter: if no state update is received from a vehicle for three consecutive cycles (first timeout), it is assumed that the vehicle is maintaining its current constant speed, and the weight of the vehicle in subsequent disturbance scoring is reduced; if the timeout lasts for more than 2 seconds (second timeout), the cooperative dependency on the vehicle is terminated, and lane changing is performed solely based on the vehicle's onboard sensor perception information.

[0038] In some feasible implementations, combined with Figure 2 As shown, step S2 includes: Step S201, detecting the driver's lane-changing intention through a steering wheel angle sensor or turn signal; if the turn signal is activated or the steering wheel angle exceeds a preset threshold, triggering a candidate lane-changing scheme generation process; Step S202, performing pre-screening, which includes first-level screening and second-level screening: The first-level screening is as follows: based on the current time, generating multiple candidate sets of lane-changing start times with different delay times; for each candidate start time, estimating the collision time between the vehicle and the vehicle behind in the target lane, as well as the minimum predicted distance between the vehicle and surrounding vehicles; classifying the start times according to a preset collision time threshold, marking start times with collision times greater than or equal to the first time threshold as safe, marking start times with collision times between the second time threshold and the first time threshold as requiring caution, and marking start times with collision times less than the second time threshold as requiring caution. The starting time of the value is marked as dangerous; at the same time, if the minimum prediction distance is less than the hard boundary, it is directly marked as dangerous; all starting times marked as dangerous are removed, and the two types of starting times, safe and requiring caution, are retained to enter the second level of screening; wherein, the second time threshold is less than the first time threshold; the second level of screening is as follows: for each starting time that passes the first level of screening, a variety of lateral trajectory shapes with different degrees of aggression and a variety of longitudinal acceleration curves are generated, and the two are combined to form a candidate scheme set; wherein, the lateral trajectory shapes with different degrees of aggression correspond to different lane change durations, and the longitudinal acceleration curves include at least one of acceleration, constant speed, and deceleration; step S203, each candidate scheme in the candidate scheme set is sent to the digital twin engine, and multiple schemes are simultaneously accelerated for simulation and deduction using a parallel computing architecture, and the overall decision delay is controlled within a preset time range.

[0039] Preferably, step S2 includes: detecting the driver's lane-changing intention through a steering wheel angle sensor or turn signal; if the turn signal is activated or the steering wheel angle exceeds a preset threshold, a candidate solution generation process is triggered. The first-level screening involves a safety assessment from the initial moment, using the current moment as a baseline to generate a candidate set of lane-changing start times (immediate lane change, 0.5-second delay, 1.0-second delay); for each candidate start time, the collision time-to-collision (TTC) between the vehicle and the vehicle behind in the target lane, as well as the minimum predicted distance between the vehicle and surrounding vehicles, are quickly estimated; the collision time threshold is dynamically set based on vehicle speed and traffic density, classifying candidate start times into three categories: safe, requiring caution, and dangerous; solutions with a minimum predicted distance less than a hard boundary are eliminated, and the safe and requiring caution categories are retained for the second-level screening. Generally, in typical highway scenarios, the second time threshold is set to 1.5s, the first time threshold is set to 2.5s, TTC ≥ 2.5 seconds is considered safe, 1.5 seconds ≤ TTC < 2.5 seconds is considered cautious, and TTC < 1.5 seconds is considered dangerous. If the minimum predicted distance is less than the hard boundary calculated based on ISO 15622 (e.g., the hard boundary is about 45 meters at a vehicle speed of 95 km / h), it is directly marked as dangerous and eliminated. The safe and cautious categories are retained. The second-level screening generates various lateral trajectory shapes and longitudinal acceleration curves with different degrees of aggression for the starting time that passes the first-level screening, combining them to form 2-6 candidate schemes. The lateral trajectory shapes include conservative, general, and aggressive types, corresponding to different lane-change durations. In this example, the conservative lane-change duration is 3.5 seconds, the general lane-change duration is 2.5 seconds, and the aggressive lane-change duration is 1.8 seconds. The longitudinal acceleration curves include acceleration curves, constant speed curves, and deceleration curves. The selected candidate solutions are sent to the digital twin engine, and multiple solutions are simultaneously accelerated through simulation and deduction using a parallel computing architecture. The overall decision delay is controlled within the preset time of 1.5 seconds.

[0040] In some feasible implementations, generating the comprehensive disturbance score for each scheme in step S3 includes: Step S301: In the digital twin environment, starting from the current moment, import each candidate scheme in the candidate scheme set into the simulation engine and perform accelerated simulation at a rate higher than the real-time speed for a predetermined time window; Step S302: During the simulation, record disturbance indices, including: the minimum collision time between the vehicle and the vehicle in front, the minimum collision time between the vehicle and the vehicle behind in the target lane, the maximum deceleration of the vehicle behind in the target lane to avoid the vehicle, the speed gain of the vehicle in front in the original lane due to the vehicle leaving, and the standard deviation of the acceleration change rate of surrounding vehicles; Step S303: Normalize each disturbance index and linearly map it to the [0,1] interval; Step S304: Dynamically determine the weight coefficient of each index according to the traffic flow density, and sum the normalized indices by weight to obtain the comprehensive disturbance score for each candidate scheme; wherein, the lower the score, the smaller the impact of the scheme on the traffic flow.

[0041] Preferably, the formula for calculating the comprehensive disturbance score is: ; In the formula, , , , and These are the weighting coefficients. ; This is the normalized value of the minimum collision time between the vehicle and the vehicle in front. This is the normalized value of the minimum collision time between the vehicle and the vehicle following in the target lane; The normalized value of the maximum deceleration of the following vehicle in the target lane to avoid the following vehicle; This is the normalized value of the speed gain gained by the vehicle in front in the original lane due to leaving the lane. This is the normalized value of the standard deviation of the rate of change of acceleration of surrounding vehicles; It is a constant; The weighting coefficients are dynamically adjusted based on traffic flow density, including: reducing the weight of the maximum deceleration term of the vehicle following in the target lane when the traffic flow density is lower than a first density threshold; and increasing the weight of the collision time term and the maximum deceleration term of the vehicle following in the target lane when the traffic flow density is higher than a second density threshold.

[0042] Preferably, step S3 includes: in the digital twin environment, starting from the current moment, importing each candidate solution into the simulation engine and performing accelerated simulation at 20 times the real-time speed, with a simulation time window of 8 seconds in the future. During the simulation, the following disturbance indicators are recorded: the minimum collision time between the vehicle and the vehicle in front, the minimum collision time between the vehicle and the vehicle behind in the target lane, the maximum deceleration value generated by the vehicle behind in the target lane to avoid the vehicle, the speed gain value gained by the vehicle in front in the original lane due to the vehicle leaving, and the standard deviation of the acceleration change rate of surrounding vehicles. The indicators are as follows: minimum collision time between the vehicle and the vehicle in front (in seconds), representing the risk of collision; a smaller value indicates a higher risk. Minimum collision time between the vehicle and the vehicle behind in the target lane (in seconds), representing the risk of collision; a smaller value indicates a higher risk. Maximum deceleration of the vehicle behind in the target lane to avoid the vehicle (in meters per second²), representing the impact on the driving comfort of the vehicle behind; a larger value indicates a more severe disturbance. Speed ​​gain of the vehicle in the original lane due to the vehicle leaving (in kilometers per hour), representing the improvement in the efficiency of the vehicle in front; a larger value is more beneficial to traffic flow. Standard deviation of the rate of change of acceleration of surrounding vehicles (in meters per second³), representing the degree of disturbance to the stability of local traffic flow; a larger value indicates a more severe disturbance. Each indicator is normalized and linearly mapped to the [0,1] interval. A weighted summation is performed to obtain the comprehensive disturbance score for each candidate scheme. The weights are dynamically adjusted according to traffic flow density; a lower comprehensive disturbance score indicates a smaller impact of the scheme on traffic flow. The weighting coefficient adjustment rule is as follows: when traffic flow density is low, the weight of the following vehicle deceleration term is reduced; in dense traffic flow, the weights of the following vehicle deceleration and collision time terms are increased. Typical weighting coefficient values ​​are as follows: In sparse traffic flow ; In dense traffic flow .

[0043] In some feasible implementations, step S4 includes: step S401, selecting the scheme with the lowest comprehensive disturbance score from the candidate scheme set as the benchmark execution scheme; step S402, calculating the insurmountable physical limit safety distance as a hard boundary based on the minimum safe following distance standard and combined with the vehicle's maximum braking capacity; step S403, taking a preset multiple of the hard boundary as the longitudinal warning distance to generate a longitudinal soft boundary, and combining it with the lane width parameter to generate a lateral safety envelope as a lateral soft boundary; step S404, combining the hard boundary and the soft boundary to form a dynamic safety boundary function that changes with time; wherein, the soft boundary is located outside the hard boundary in space.

[0044] Preferably, step S4 includes: selecting the scheme with the lowest comprehensive disturbance score from the candidate scheme set as the benchmark execution scheme; extracting the minimum distance value between the vehicle and surrounding vehicles at each moment during the simulation process of the benchmark execution scheme; and calculating the insurmountable physical limit safe distance as a hard boundary based on the minimum safe following distance standard and the maximum braking capacity of the vehicle. In this embodiment, the insurmountable physical limit safe distance is calculated based on the minimum safe following distance standard in the ISO 15622 standard and the maximum braking capacity of the vehicle. At a vehicle speed of 95 km / h, ISO 15622 recommends a minimum safe following distance of approximately 45 meters. A soft boundary is generated by taking a preset multiple of the hard boundary as the longitudinal warning distance, and this is combined with lane width parameters to generate a lateral safety envelope; specifically, 1.2 times the hard boundary is taken as the longitudinal warning distance, i.e. Meters; Lateral safety envelope Take 80% of the lane width (approximately 2.9 meters). Combine the hard and soft boundaries to form a dynamic safety boundary function that varies over time. The soft boundary is spatially located outside the hard boundary. In the embodiment, the system issues a warning when the following vehicle approaches 54 meters and applies emergency braking when it approaches 45 meters. The calculated actual minimum distance is 50 meters, which is greater than the hard boundary of 45 meters, thus meeting the safety requirements.

[0045] In some feasible implementations, step S5 includes: step S501, obtaining the expected trajectory data contained in the baseline execution scheme, wherein the expected trajectory data is determined by the lateral trajectory shape and longitudinal acceleration curve of the candidate scheme generated in step S2, including the expected lane change time window and the expected trajectory point sequence; step S502, encapsulating the dynamic safety boundary and the expected trajectory data into a virtual trajectory commitment data packet, wherein the virtual trajectory commitment data packet includes a contract unique identifier, an initiator vehicle de-identification identifier, the expected lane change time window, the expected trajectory point sequence, longitudinal safety distance hard and soft boundary parameters, lateral safety envelope soft boundary parameters, and liability clause fields; step S503, ... The virtual trajectory commitment data packet is sent to the collaborating vehicle via the vehicle-to-everything (V2X) communication channel, and a notification message is broadcast to surrounding vehicles. Step S504: Receive confirmation information from the collaborating vehicle within a predetermined time, indicating acceptance of the commitment. Step S505: If confirmation information is received, the collaborating vehicle uses the virtual trajectory commitment data packet as an input constraint for its own decision-making module, and adjusts its driving strategy outside the safety boundary, forming implicit collaboration with two-way confirmation. If no confirmation information is received, the initiator chooses to resend or abandon the scheme. If no confirmation information is received after a preset number of attempts, the scheme is automatically abandoned, and the system switches to a pre-calculated, pending-activation suboptimal candidate scheme. For partial confirmations, the safety boundary is adjusted before execution continues.

[0046] Preferred, combined Figure 3 As shown, step S5 includes: encapsulating the expected trajectory data in the dynamic safety boundary and baseline execution scheme into a virtual trajectory commitment data packet. The virtual trajectory commitment data packet includes a unique contract identifier, an anonymized identifier for the initiating vehicle, an expected lane-change time window, an expected trajectory point sequence, longitudinal safety distance hard and soft boundary parameters, lateral safety envelope soft boundary parameters, and liability clause fields. The system records trajectory deviation data and safety boundary intrusion events in real time, using digital signature technology to ensure data immutability. Through the vehicle-to-everything (V2X) low-latency communication channel, the virtual trajectory commitment data packet is directionally sent to the cooperating vehicle located in the target lane and behind the predicted lane-change trajectory of the current vehicle. Simultaneously, a notification message including the contract identifier and the initiating vehicle identifier is broadcast to all surrounding connected vehicles. The cooperating vehicle should reply with a confirmation message within a time limit dynamically adjusted according to channel quality. If confirmation is received, the collaborative vehicle will use the virtual trajectory commitment as an input constraint for its own decision-making module and adjust its driving strategy outside the safety boundary, forming an implicit collaboration with two-way confirmation. If no confirmation is received, the initiator will choose to resend or abandon the plan. If no confirmation is received after a preset number of attempts, the plan will be automatically abandoned, and the system will switch to a pre-calculated and pending suboptimal candidate plan. In the case of partial confirmation, the system will adjust the safety boundary and continue execution.

[0047] Specifically, the virtual trajectory commitment data packet is sent to vehicle B, located 35 meters behind the target lane, and simultaneously broadcast to all connected vehicles within a 100-meter radius. Vehicle B replies with a confirmation message after 220 milliseconds, indicating acceptance of the commitment and agreement to adhere to the safety boundaries. Vehicle B marks A's commitment as a constraint and maintains a distance greater than 54 meters in subsequent planning. This establishes an implicit, two-way confirmation collaboration between the two vehicles.

[0048] In some feasible implementations, step S6 includes: step S601, during lane change execution, collecting actual motion state data of the vehicle at a preset frequency, including actual position, speed, and acceleration; step S602, comparing the actual trajectory with the expected trajectory points in the virtual trajectory commitment data packet point by point, and calculating position deviation and speed deviation; step S603, if the position deviation does not exceed a predetermined deviation threshold, maintaining the current dynamic safety boundary; if the position deviation exceeds the predetermined deviation threshold, triggering safety boundary correction: using the actual position as a new starting point, recalculating the dynamic safety boundary within the remaining time window, smoothly transitioning to the originally committed endpoint state, and rebroadcasting the corrected dynamic safety boundary. To surrounding vehicles; Step S604: Dynamically adjust the abort threshold based on lane change completion, including: when the lane change completion is less than a first predetermined completion threshold, and the position deviation continues to exceed the predetermined deviation threshold for a predetermined number of times, trigger the lane change abort process; when the lane change completion is greater than a second predetermined completion threshold, even if the position deviation exists, the lane change is forcibly completed, and only the dynamic safety boundary is adjusted; when the lane change completion is between the first predetermined completion threshold and the second predetermined completion threshold, determine whether to abort based on the degree of deviation and the remaining space; the lane change completion is defined as 1 - (distance between the current lateral position and the center line of the target lane / distance between the original lane center line and the center line of the target lane).

[0049] Preferably, step S6 includes: during lane-changing execution, collecting the vehicle's actual position, speed, and acceleration data at a preset frequency. The actual trajectory is compared point-by-point with the expected trajectory points in the virtual trajectory commitment, and position and speed deviations are calculated. If the position deviation is within a preset threshold range, the original dynamic safety boundary is maintained; if the position deviation exceeds the threshold, safety boundary correction is triggered. Correcting the safety boundary involves recalculating the safety boundary within the remaining time window using the actual position as the new starting point, and smoothly transitioning to the originally committed endpoint state using linear interpolation. The corrected safety boundary is then broadcast to surrounding vehicles via the vehicle network and awaits confirmation. The abort threshold is dynamically adjusted based on the lane-changing completion rate: if the completion rate is less than a first preset threshold and the deviation continues to exceed the threshold for a predetermined number of times, the lane-changing abort process is triggered; if the completion rate is greater than a second preset threshold, the lane change is forcibly completed even if a deviation occurs; if the completion rate is between the first and second preset thresholds, a comprehensive judgment is made based on the degree of deviation and remaining space. If the abort procedure is triggered, first check whether returning to the original lane will cause greater risk and assess the collision time with the vehicle behind in the original lane; if the return path is safe, proceed with returning to the original lane; otherwise, continue to complete the lane change and strengthen the warning.

[0050] In some feasible implementations, step S7 includes: step S701, after lane changing is completed, generating a lane changing event record, the record including at least the selected candidate scheme identifier, initial dynamic safety boundary parameters, and the actual execution trajectory point sequence; step S702, collecting response data of surrounding vehicles during the lane changing process, the response data including at least the acceleration change record of each vehicle, whether emergency braking occurred, and whether evasive action was performed; step S703, desensitizing vehicle identifiers to remove personally identifiable information, and storing the desensitized event record in a local encrypted storage area; step S704, using federated learning technology, optimizing the weight coefficients used to generate the comprehensive disturbance score based on the stored event record, and improving the prediction accuracy of the digital twin environment; wherein, only the weight coefficients are uploaded to the cloud for aggregation, and the original trajectory data is not stored locally.

[0051] Preferably, step S7 includes: after the lane-changing action is completed, the system generates a lane-changing event summary record, including the lane-changing start time, end time, selected candidate scheme identifier, initial dynamic safety boundary parameters, actual execution trajectory point sequence, whether corrections were triggered during execution, and the number of corrections. Response data of surrounding vehicles during the lane-changing process is collected, including acceleration change records of each vehicle, whether emergency braking occurred, and whether evasive maneuvers were performed. Vehicle identifiers are anonymized to remove personally identifiable information, and the anonymized event records are stored in a local encrypted storage area. Federated learning technology is used for model training, uploading only model parameters or gradient information to the cloud for aggregation; the original trajectory data remains local. The event records are used to periodically analyze the effectiveness of lane-changing decisions, optimize the weight parameters in the candidate scheme generation strategy, and improve the prediction accuracy of the digital twin model.

[0052] To facilitate understanding of the above implementation methods, specific application examples are provided here: Step S1: In an urban expressway scenario, the system establishes virtual mirror images of vehicle A and 12 other connected vehicles within a 100-meter radius. Basic safety messages sent by each vehicle at 10 Hz are received via the PC5 interface. After data fusion, the positional error of each vehicle in the virtual space does not exceed 0.2 meters, and the speed error does not exceed 0.5 km / h. The digital twin environment can complete a full-scene state update within milliseconds, laying a data foundation for subsequent simulations.

[0053] In response to edge cases of communication interruption or loss of state synchronization, the system sets a communication timeout counter: if no state update is received from a vehicle for 3 consecutive cycles (0.3 seconds), the vehicle's motion state is switched to constant speed driving, and the weight of the vehicle in the disturbance score is reduced; if the timeout lasts for more than 2 seconds, the system terminates its dependence on the vehicle, and the system performs lane changing only based on the perception information of the vehicle's onboard sensors.

[0054] On the highway, the driver turns on the left turn signal, intending to change lanes, and the system immediately detects this intention. The current vehicle speed is 95 km / h, the vehicle in front is 85 km / h, and the vehicle behind in the target lane is traveling at 100 km / h, with a distance of 35 meters.

[0055] Step S2, Combining Figure 2 As shown, the first-level filtering uses the current time as a baseline to generate a candidate set of lane-change start times (immediate lane change, 0.5-second delay, 1.0-second delay). For each start time, the time-to-collision (TTC) between the vehicle and the following vehicle in the target lane is quickly estimated. Thresholds are dynamically set based on vehicle speed and traffic density: in typical highway scenarios, a TTC ≥ 2.5 seconds is considered safe, 1.5 seconds ≤ TTC < 2.5 seconds is considered cautious, and TTC < 1.5 seconds is considered dangerous. Simultaneously, the minimum predicted distance between the vehicle and surrounding vehicles at that start time is estimated. If this distance is less than the hard boundary calculated based on ISO 15622 (approximately 45 meters at a speed of 95 km / h), it is directly marked as dangerous and removed.

[0056] Specifically, the lane change intention is triggered by detecting the turn signal being activated via the steering wheel angle sensor, which then triggers the generation of candidate solutions.

[0057] First-level screening (starting time safety assessment): Based on the current time, a candidate set of lane-change starting times is generated: immediate lane change, 0.5-second delay, and 1.0-second delay. For each starting time, the time to collision (TTC) between the vehicle and vehicle B in the target lane is quickly estimated, and a threshold is dynamically set based on vehicle speed and traffic density (in typical highway scenarios: TTC ≥ 2.5 seconds is safe, 1.5 seconds ≤ TTC < 2.5 seconds requires caution, and TTC < 1.5 seconds is dangerous). Simultaneously, the minimum predicted distance between the vehicle and surrounding vehicles at that starting time is estimated. If this distance is less than the hard boundary calculated based on ISO 15622 (approximately 45 meters at 95 km / h), it is directly marked as dangerous and removed.

[0058] Calculation and judgment results: Immediate lane change: TTC = 1.26 seconds (dangerous), minimum predicted distance 38 meters (<45 meters) → Dangerous, remove.

[0059] 0.5-second delay: TTC = 1.76 seconds (caution required), minimum prediction distance 42 meters (<45 meters) → dangerous, eliminated.

[0060] Delay 1.0 second: TTC = 2.26 seconds (caution required), minimum prediction distance 48 meters (>45 meters) → Keep.

[0061] Therefore, the second-level screening begins with a 1.0-second delay at the start time.

[0062] Second-level filtering (trajectory pattern combination): For the starting time (delayed by 1.0 second) that passes the first-level screening, three different lateral trajectory shapes and three longitudinal acceleration curves with varying degrees of aggression are generated, resulting in a maximum of nine candidate schemes: Lateral trajectory shape: Conservative (lane change duration 3.5 seconds), Normal (2.5 seconds), Aggressive (1.8 seconds).

[0063] Longitudinal acceleration curves: acceleration, constant speed, deceleration.

[0064] Based on the driver's current accelerator pedal position (neutral, preferring constant speed), the options are narrowed down to three, which are then fed into subsequent analysis: Option 1: 1.0 second delay + conservative lateral movement + uniform longitudinal movement; Option 2: 1.0 second delay + standard horizontal movement + accelerated vertical movement; Option 3: 1.0 second delay + aggressive lateral movement + decelerated longitudinal movement.

[0065] Step S3: Perform accelerated simulation on the above three candidate schemes to predict the degree of disturbance to the vehicle, surrounding vehicles and local traffic flow after each scheme is implemented, and generate a comprehensive disturbance score for each scheme.

[0066] In this embodiment, the traffic flow is of medium density, and the calculation results are as follows: Option 1 (Conservative lateral movement + uniform longitudinal movement): ; Option 2 (General lateral movement + accelerated longitudinal movement): ; Option 3 (Aggressive Lateral Movement + Decelerating Vertical Movement): .

[0067] Step S4: The system selects the lowest-scoring scheme 1 (0.2215) as the baseline execution scheme. Scheme 1 is extracted, and its dynamic safety boundary formed during the simulation process is calculated.

[0068] Dynamic security boundaries include: Hard boundary: Based on the ISO15622 standard, it is approximately 45 meters at a vehicle speed of 95 km / h (an insurmountable physical limit of safe distance).

[0069] Soft boundary: Take 1.2 times the hard boundary as the longitudinal warning distance, i.e., 54 meters; the lateral safety envelope is 80% of the lane width, about 2.9 meters.

[0070] Step S5, Combining Figure 3As shown, the calculated dynamic safety boundary and the expected trajectory data in the baseline execution plan are encapsulated into a virtual trajectory commitment data package, including a unique contract identifier, a pseudonymous identifier for the initiating vehicle, an expected lane-changing time window (3.2 seconds to 5.5 seconds), an expected trajectory point sequence (one point every 0.2 seconds, for a total of 12 points), a longitudinal safety distance hard boundary of 45 meters and a soft boundary of 54 meters, a lateral safety envelope soft boundary of 2.9 meters, and a liability clause field. This field records the intention of the parties to divide responsibilities. The system records trajectory deviation data and safety boundary intrusion events in real time, and uses digital signature technology to ensure that the data is tamper-proof.

[0071] The virtual trajectory commitment data packet is sent via the PC5 interface to vehicle B, located 35 meters behind the target lane, and simultaneously broadcast to all connected vehicles within a 100-meter radius. Vehicle B replies with an acknowledgment message after 220 milliseconds, indicating acceptance of the commitment and agreement to adhere to the safety boundaries. Vehicle B marks A's commitment as a constraint and maintains a distance greater than 54 meters in subsequent planning. This establishes an implicit, two-way, acknowledgment-based collaboration.

[0072] Step S6: Starting from vehicle A, lane change is executed, and actual position, speed, and acceleration data are collected at a frequency of 10 Hz and compared with the expected trajectory points point by point.

[0073] Lane change completion rate definition: Completion rate = 1 - (Distance between current lateral position and target lane centerline / Distance between original lane centerline and target lane centerline).

[0074] Dynamic abort threshold rules: If the completion rate is less than 30% and the deviation continues to exceed the threshold (3 consecutive samples, i.e. 0.3 seconds), the lane change abort process is triggered; if the completion rate is greater than 70%, the lane change should be forced to complete even if there is a deviation, and only the safety boundary is adjusted; if the completion rate is between 30% and 70%, a comprehensive judgment is made based on the degree of deviation and the remaining space.

[0075] Emergency: When the completion rate reaches 15%, the lateral deviation is 0.25 meters due to crosswind (the preset deviation threshold is 0.2 meters). The system determines that the process should be terminated after three consecutive samplings that exceed the threshold.

[0076] Abort safety check: Check if returning to the original lane would pose a greater risk: If the distance to the vehicle behind in the original lane is 50 meters and the collision time is 3.2 seconds (greater than the common safety threshold), return safely.

[0077] Execution aborted: The system performs a return operation and simultaneously broadcasts the cancellation of the original commitment. The vehicle smoothly returns to its original lane, and the lane change is aborted.

[0078] Step S7: After lane changing is completed, a lane changing event record is generated, including the selected scheme identifier, initial dynamic safety boundary parameters, actual trajectory point sequence, and whether a correction was triggered. Response data from surrounding vehicles (acceleration changes, braking status, etc. of vehicle B) is collected. Vehicle identifiers are anonymized to remove personally identifiable information, and the anonymized event record is stored in a local encrypted storage area. Federated learning technology is used to optimize the weight coefficients used to generate the comprehensive disturbance score based on the stored event record, improving the prediction accuracy of the digital twin environment; only the weight coefficients are uploaded to the cloud for aggregation, and the original trajectory data remains local.

[0079] Example 2 Please see Figure 4 This embodiment provides a schematic diagram of a lane-changing collaborative decision-making system based on digital twins and dynamic safety boundaries.

[0080] As an example, the system is implemented using the lane-changing collaborative decision-making method based on digital twins and dynamic safety boundaries described in Embodiment 1, including: The vehicle-level digital twin mirror system construction module 410 is suitable for building a vehicle-level digital twin mirror system. It synchronizes the dynamic status data of the vehicle and surrounding vehicles in real time through vehicle network communication, and establishes a simulation and simulation environment in virtual space that is consistent with the physical environment, thus forming a digital twin environment. The candidate scheme set generation module 420 is suitable for generating hierarchically screened candidate lane-changing schemes in the digital twin environment when a vehicle's lane-changing intention is detected. This includes pre-screening the candidate start time based on the collision time and minimum predicted distance, eliminating schemes that do not meet the safety threshold, and performing accelerated simulation and deduction on the remaining schemes to form a candidate scheme set. The comprehensive disturbance score generation module 430 is suitable for accelerating simulation and deduction of each scheme in the candidate scheme set in a digital twin environment, predicting the disturbance degree of the scheme to the vehicle, surrounding vehicles and local traffic flow after the scheme is executed, and generating a comprehensive disturbance score corresponding to each scheme. The safety boundary generation module 440 is suitable for selecting the candidate scheme with the least impact on traffic flow as the benchmark execution scheme based on the comprehensive disturbance score, and calculating the dynamic safety boundary formed by the benchmark execution scheme during the simulation process; the dynamic safety boundary includes two levels: hard boundary and soft boundary, wherein the hard boundary is calculated based on the minimum safe following distance standard, and the soft boundary is generated based on the hard boundary superimposed with the safety margin. The cooperative communication module 450 is suitable for using the dynamic safety boundary as a cooperative constraint parameter, broadcasting it to surrounding vehicles within the communication range via the vehicle network, and establishing a two-way commitment using a request-confirmation handshake mechanism, sending a virtual trajectory commitment data packet containing lane-changing time window, expected trajectory characteristics, and liability clause fields to surrounding vehicles. The execution and correction module 460 is suitable for the vehicle to perform lane-changing operations under the hard constraints of the broadcast dynamic safety boundary. During the execution process, it continuously compares the deviation value between the actual trajectory and the virtual trajectory commitment signal, corrects the safety boundary parameters in real time according to the deviation value, and dynamically adjusts the abort threshold according to the lane-changing completion degree. The data storage module 470 is suitable for desensitizing and storing lane-changing event records after a vehicle completes a lane change, as training samples for subsequent decision optimization.

[0081] It is not difficult to see that this embodiment is a system implementation corresponding to the first embodiment, and this embodiment can be implemented in conjunction with the first embodiment. The relevant technical details mentioned in the first embodiment are still valid in this embodiment, and will not be repeated here to reduce repetition. Accordingly, the relevant technical details mentioned in this embodiment can also be applied to the first embodiment.

[0082] It is worth mentioning that all modules involved in this embodiment are logical units. In practical applications, a logical unit can be a physical unit, a part of a physical unit, or a combination of multiple physical units. Furthermore, to highlight the innovative aspects of this invention, this embodiment does not introduce units that are not closely related to solving the technical problem proposed by this invention; however, this does not mean that other units are absent from this embodiment.

[0083] Example 3 Please see Figure 5 The present invention also provides an electronic device, including: a memory and a processor; the memory stores at least one program instruction; the processor loads and executes the at least one program instruction to implement the lane-changing collaborative decision-making method based on digital twin and dynamic security boundary provided in Embodiment 1.

[0084] The memory 702 and processor 701 are connected via a bus, which may include any number of interconnecting buses and bridges, connecting various circuits of one or more processors 701 and memory 702 together. The bus may also connect various other circuits, such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and therefore will not be described further herein. A bus interface provides an interface between the bus and the transceiver. The transceiver may be a single element or multiple elements, such as multiple receivers and transmitters, providing a unit for communicating with various other devices over a transmission medium. Data processed by processor 701 is transmitted over a wireless medium via an antenna, which further receives data and transmits it to processor 701.

[0085] Processor 701 is responsible for managing the bus and general processing, and can also provide various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions. Memory 702 can be used to store data used by processor 701 during operation.

[0086] Example 4 This invention also proposes a storage medium storing a lane-switching collaborative decision-making method based on digital twins and dynamic security boundaries. When executed by a processor, the lane-switching collaborative decision-making program based on digital twins and dynamic security boundaries implements the steps of the lane-switching collaborative decision-making method based on digital twins and dynamic security boundaries as described above. Since this storage medium adopts all the technical solutions of all the above embodiments, it possesses at least all the beneficial effects brought about by the technical solutions of the above embodiments, which will not be elaborated upon here.

[0087] The above descriptions are merely embodiments of the present invention. Commonly known structures and characteristics are not described in detail here. Those skilled in the art are aware of all common technical knowledge in the field prior to the application date or priority date, are aware of all existing technologies in that field, and have the ability to apply conventional experimental methods prior to that date. Those skilled in the art can, based on the guidance provided in this application, improve and implement this solution in combination with their own capabilities. Some typical known structures or methods should not be obstacles for those skilled in the art to implement this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the structure of the present invention. These should also be considered within the scope of protection of the present invention, and will not affect the effectiveness of the implementation of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. A lane-changing collaborative decision-making method based on digital twins and dynamic safety boundaries, characterized in that, Includes the following steps: Step S1: Construct a vehicle-level digital twin mirror system. Through vehicle-to-everything (V2X) communication, synchronize the dynamic status data of the vehicle and surrounding vehicles in real time. Establish a simulation and deduction environment in virtual space that is consistent with the physical environment to form a digital twin environment. Step S2: When the intention of the vehicle to change lanes is detected, a hierarchical candidate lane-changing scheme is generated in the digital twin environment, including pre-screening the candidate start time based on the collision time and the minimum predicted distance, eliminating schemes that do not meet the safety threshold, and performing accelerated simulation on the remaining schemes to form a set of candidate schemes. Step S3: Perform accelerated simulation and deduction on each of the candidate schemes in the digital twin environment, predict the degree of disturbance to the vehicle, surrounding vehicles and local traffic flow after the scheme is executed, and generate a comprehensive disturbance score for each scheme. Step S4: Based on the comprehensive disturbance score, select the candidate scheme with the least impact on traffic flow as the benchmark execution scheme, and calculate the dynamic safety boundary formed by the benchmark execution scheme during the simulation process; the dynamic safety boundary includes two levels: hard boundary and soft boundary, wherein the hard boundary is calculated based on the minimum safe following distance standard, and the soft boundary is generated based on the hard boundary superimposed with the safety margin. Step S5: The dynamic safety boundary is used as a collaborative constraint parameter and broadcast to surrounding vehicles within the communication range via the vehicle network. A two-way commitment is established using a request-confirmation handshake mechanism, and a virtual trajectory commitment data packet containing lane-changing time window, expected trajectory characteristics, and liability clause fields is sent to surrounding vehicles. Step S6: The vehicle performs a lane change operation under the hard constraints of the broadcast dynamic safety boundary. During the operation, the deviation value between the actual trajectory and the virtual trajectory commitment signal is continuously compared. The safety boundary parameters are corrected in real time according to the deviation value, and the abort threshold is dynamically adjusted according to the lane change completion degree. Step S7: After the vehicle completes the lane change, the lane change event record is anonymized and stored as a training sample for subsequent decision optimization.

2. The lane-changing collaborative decision-making method based on digital twin and dynamic safety boundary as described in claim 1, characterized in that, Step S1 includes: Step S101: Deploy a digital twin engine on edge computing nodes or in the cloud, adopt a layered computing architecture, and allocate an independent virtual mirror space for each vehicle connected to the system. Step S102: Obtain the dynamic status data of the vehicle and surrounding vehicles in real time through vehicle network communication, and write the obtained dynamic status data into the corresponding virtual mirror space so that the state synchronization error between the virtual mirror and the physical entity is controlled within the preset accuracy range. Step S103: Construct a static background layer containing road geometry and traffic sign information in the virtual space as the basic environment for simulation. Step S104: In response to communication anomalies, a communication timeout handling mechanism is set up, including: when the status data of a vehicle is continuously missing for more than a first time limit, the dependency weight of the vehicle in subsequent decisions is reduced; when the missing data is missing for more than a second time limit, the collaborative dependency on the vehicle is terminated and the system switches to an independent decision-making mode based on onboard sensor information.

3. The lane-changing collaborative decision-making method based on digital twin and dynamic safety boundary as described in claim 1, characterized in that, Step S2 includes: Step S201: Detect the driver's lane-changing intention through the steering wheel angle sensor or turn signal. If the turn signal is activated or the steering wheel angle exceeds a preset threshold, the candidate lane-changing scheme generation process is triggered. Step S202: Perform pre-screening, which includes first-level screening and second-level screening. The first-level screening process involves: generating multiple candidate sets of lane-changing start times with different delay times based on the current time; estimating the collision time between the vehicle and the vehicle behind in the target lane, as well as the minimum predicted distance between the vehicle and surrounding vehicles for each candidate start time; classifying the start times according to a preset collision time threshold: start times with collision times greater than or equal to the first time threshold are marked as safe; start times with collision times between the second and first time thresholds are marked as requiring caution; and start times with collision times less than the second time threshold are marked as dangerous. Simultaneously, if the minimum predicted distance is less than the hard boundary, it is directly marked as dangerous. All start times marked as dangerous are removed, and the safe and cautious start times are retained for the second-level screening; wherein, the second time threshold is less than the first time threshold. The second-level screening involves generating multiple lateral trajectory shapes with different degrees of aggression and multiple longitudinal acceleration curves for each starting moment that passes the first-level screening, and combining the two to form a candidate scheme set; wherein, different degrees of aggression in the lateral trajectory shapes correspond to different lane-changing durations, and the longitudinal acceleration curves include at least one of acceleration, constant speed, and deceleration. Step S203: Send each candidate scheme in the candidate scheme set into the digital twin engine, use a parallel computing architecture to simultaneously perform accelerated simulation and deduction of multiple schemes, and control the overall decision delay within a preset time range.

4. The lane-changing collaborative decision-making method based on digital twin and dynamic safety boundary as described in claim 1, characterized in that, The comprehensive disturbance score generated for each scheme in step S3 includes: Step S301: In the digital twin environment, starting from the current moment, each candidate solution in the candidate solution set is imported into the simulation engine and accelerated at a rate higher than the real-time speed for a predetermined time window. Step S302: During the simulation, record the disturbance indexes, which include: the minimum collision time between the vehicle and the vehicle in front, the minimum collision time between the vehicle and the vehicle behind in the target lane, the maximum deceleration of the vehicle behind in the target lane to avoid the vehicle, the speed gain of the vehicle in front in the original lane due to the vehicle leaving, and the standard deviation of the acceleration change rate of surrounding vehicles. Step S303: Normalize each disturbance index and linearly map it to the [0,1] interval; Step S304: Dynamically determine the weight coefficients of each indicator based on traffic flow density, and sum the normalized indicators by weight to obtain the comprehensive disturbance score of each candidate scheme; where the lower the score, the smaller the impact of the scheme on traffic flow.

5. The lane-changing collaborative decision-making method based on digital twin and dynamic safety boundary as described in claim 4, characterized in that, The formula for calculating the comprehensive disturbance score is as follows: ； In the formula, , , , and These are the weighting coefficients. ; This is the normalized value of the minimum collision time between the vehicle and the vehicle in front. This is the normalized value of the minimum collision time between the vehicle and the vehicle following in the target lane; The normalized value of the maximum deceleration of the following vehicle in the target lane to avoid the following vehicle; This is the normalized value of the speed gain gained by the vehicle in front in the original lane due to leaving the lane. This is the normalized value of the standard deviation of the rate of change of acceleration of surrounding vehicles; It is a constant; The weighting coefficients are dynamically adjusted based on traffic flow density, including: reducing the weight of the maximum deceleration term of the vehicle following in the target lane when the traffic flow density is lower than a first density threshold; and increasing the weight of the collision time term and the maximum deceleration term of the vehicle following in the target lane when the traffic flow density is higher than a second density threshold.

6. The lane-changing collaborative decision-making method based on digital twin and dynamic safety boundary as described in claim 1, characterized in that, Step S4 includes: Step S401: Select the scheme with the lowest comprehensive disturbance score from the candidate scheme set as the benchmark execution scheme; Step S402: Based on the minimum safe following distance standard and combined with the maximum braking capacity of the vehicle, calculate the insurmountable physical limit safe distance as a hard boundary. Step S403: Take a preset multiple of the hard boundary as the longitudinal warning distance to generate a longitudinal soft boundary, and combine it with the lane width parameter to generate a lateral safety envelope as a lateral soft boundary; Step S404: Combine the hard boundary with the soft boundary to form a time-varying dynamic safety boundary function; wherein the soft boundary is located spatially outside the hard boundary.

7. The lane-changing collaborative decision-making method based on digital twin and dynamic safety boundary as described in claim 1, characterized in that, Step S5 includes: Step S501: Obtain the expected trajectory data contained in the benchmark execution scheme. The expected trajectory data is determined by the lateral trajectory shape and longitudinal acceleration curve of the candidate scheme generated in step S2, including the expected lane change time window and the expected trajectory point sequence. Step S502: Encapsulate the dynamic safety boundary and the expected trajectory data into a virtual trajectory commitment data packet. The virtual trajectory commitment data packet includes a contract unique identifier, an initiator vehicle desensitization identifier, expected lane change time window, expected trajectory point sequence, longitudinal safety distance hard boundary and soft boundary parameters, lateral safety envelope soft boundary parameters, and liability clause fields. Step S503: Send the virtual trajectory commitment data packet to the cooperating vehicle via the vehicle network communication channel, and broadcast the notification message to surrounding vehicles at the same time; Step S504: Receive confirmation information from the cooperating vehicle within a predetermined time, indicating acceptance of the commitment; Step S505: If a confirmation message is received, the cooperating vehicle uses the virtual trajectory commitment data packet as the input constraint of its own decision module and adjusts its driving strategy outside the safety boundary to form implicit cooperation with two-way confirmation; if no confirmation message is received, the initiator chooses to resend or abandon the scheme; if no confirmation message is received after more than a preset number of times, the scheme is automatically abandoned and the vehicle switches to the pre-calculated and pending suboptimal candidate scheme; for partial confirmation, the safety boundary is adjusted and execution continues.

8. The lane-changing collaborative decision-making method based on digital twin and dynamic safety boundary as described in claim 1, characterized in that, Step S6 includes: Step S601: During the lane change process, collect the actual motion state data of the vehicle at a preset frequency, including actual position, speed and acceleration; Step S602: Compare the actual trajectory with the expected trajectory points in the virtual trajectory commitment data packet point by point, and calculate the position deviation and velocity deviation; Step S603: If the position deviation does not exceed the predetermined deviation threshold, the current dynamic safety boundary is maintained; if the position deviation exceeds the predetermined deviation threshold, safety boundary correction is triggered: the dynamic safety boundary within the remaining time window is recalculated with the actual position as the new starting point, and smoothly transitioned to the originally promised endpoint state, and the corrected dynamic safety boundary is rebroadcast to the surrounding vehicles. Step S604: Dynamically adjust the abort threshold based on the lane change completion rate, including: when the lane change completion rate is less than a first predetermined completion rate threshold and the position deviation continues to exceed the predetermined deviation threshold for a predetermined number of times, trigger the lane change abort process; when the lane change completion rate is greater than a second predetermined completion rate threshold, even if the position deviation exists, the lane change is forcibly completed, and only the dynamic safety boundary is adjusted; when the lane change completion rate is between the first predetermined completion rate threshold and the second predetermined completion rate threshold, determine whether to abort based on the degree of deviation and the remaining space. The lane change completion rate is defined as 1 - (the distance between the current lateral position and the center line of the target lane / the distance between the center line of the original lane and the center line of the target lane).

9. The lane-changing collaborative decision-making method based on digital twin and dynamic safety boundary as described in claim 1, characterized in that, Step S7 includes: Step S701: After the lane change is completed, a lane change event record is generated. The record includes at least the selected candidate scheme identifier, the initial dynamic safety boundary parameters, and the actual execution trajectory point sequence. Step S702: Collect response data of surrounding vehicles during the lane change process. The response data includes at least the acceleration change records of each vehicle, whether emergency braking occurred, and whether evasive maneuvers were performed. Step S703: Desensitize the vehicle identification, remove information that can identify an individual, and store the desensitized event records in a local encrypted storage area. Step S704: Using federated learning technology, based on the stored event records, optimize the weight coefficients used to generate the comprehensive disturbance score and improve the prediction accuracy of the digital twin environment; wherein, only the weight coefficients are uploaded to the cloud for aggregation, and the original trajectory data is not left locally.

10. A lane-changing collaborative decision-making system based on digital twins and dynamic safety boundaries, implemented using the lane-changing collaborative decision-making method based on digital twins and dynamic safety boundaries as described in any one of claims 1-9, characterized in that, include: The vehicle-level digital twin mirror system construction module is suitable for building a vehicle-level digital twin mirror system. It synchronizes the dynamic status data of the vehicle and surrounding vehicles in real time through vehicle network communication, and establishes a simulation and simulation environment in virtual space that is consistent with the physical environment, thus forming a digital twin environment. The candidate scheme set generation module is suitable for generating hierarchical and screened candidate lane-changing schemes in the digital twin environment when a vehicle's lane-changing intention is detected. This includes pre-screening the candidate start time based on the collision time and minimum predicted distance, eliminating schemes that do not meet the safety threshold, and performing accelerated simulation on the remaining schemes to form a candidate scheme set. The comprehensive disturbance score generation module is suitable for accelerating simulation and deduction of each scheme in the candidate scheme set in a digital twin environment, predicting the degree of disturbance to the vehicle, surrounding vehicles and local traffic flow after the scheme is executed, and generating a comprehensive disturbance score corresponding to each scheme. The safety boundary generation module is suitable for selecting the candidate scheme with the least impact on traffic flow as the benchmark execution scheme based on the comprehensive disturbance score, and calculating the dynamic safety boundary formed by the benchmark execution scheme during the simulation process; the dynamic safety boundary includes two levels: hard boundary and soft boundary, wherein the hard boundary is calculated based on the minimum safe following distance standard, and the soft boundary is generated based on the hard boundary superimposed with the safety margin; The collaborative communication module is suitable for using the dynamic safety boundary as a collaborative constraint parameter, broadcasting it to surrounding vehicles within the communication range via the vehicle network, and establishing a two-way commitment using a request-confirmation handshake mechanism, sending a virtual trajectory commitment data packet containing lane-changing time window, expected trajectory characteristics, and liability clause fields to surrounding vehicles; The execution and correction module is suitable for vehicles to perform lane-changing operations under the hard constraints of broadcast dynamic safety boundaries. During the execution process, it continuously compares the deviation value between the actual trajectory and the virtual trajectory commitment signal, corrects the safety boundary parameters in real time based on the deviation value, and dynamically adjusts the abort threshold based on the lane-changing completion degree. The data storage module is suitable for desensitizing and storing lane-changing event records after a vehicle completes a lane change, serving as training samples for subsequent decision optimization.

Images