Method for individual generation and trajectory fusion of vehicles in macro-micro conversion ghost fleet

Abstract

The application provides a macro-micro conversion individual generation and trajectory fusion method of ghost car fleet vehicles, and relates to the technical field of intelligent transportation. The method comprises the following steps: in a target trajectory generation period in multiple trajectory generation periods, obtaining a traffic state label of a target traffic flow according to a continuous field quantity buffer of the target traffic flow; generating a micro vehicle according to the traffic state label, and obtaining a multidimensional state of the micro vehicle; fusing a target trajectory with the highest candidate weight in multiple candidate trajectories generated in a first area and a pre-acquired measured trajectory to obtain a completed trajectory; taking the completed trajectory and a micro vehicle trajectory as a vehicle trajectory generation result of the target trajectory generation period, and storing the vehicle trajectory generation result in a discrete individual quantity buffer; wherein the micro vehicle trajectory is generated according to the multidimensional state. Compared with a macro-micro conversion method of randomly scattering points, the method can reduce false speed waves and false queues, and improve the consistency of macro-micro conversion, data fusion accuracy and trajectory completion quality.

Description

Technical Field

[0001] This invention relates to the field of intelligent transportation technology, and in particular to a method for individualized generation and trajectory fusion of ghost fleet vehicles in macro-micro transformation. Background Technology

[0002] Traffic flow can be described at two scales: macroscopic (continuous field quantities such as density, speed, and flow rate) and microscopic (individual vehicle position, speed, and acceleration). Conversion between these two scales is necessary: ​​macroscopic to microscopic requires generating a batch of specific vehicles based on the macroscopic state, while microscopic to macroscopic requires aggregating individual vehicle trajectories into continuous field quantities. Existing conversion methods employ "forward random point scattering" for macroscopic to microscopic conversion, resulting in vehicles whose initial states deviate from dynamic equilibrium. After startup, these vehicles spontaneously generate and amplify speed fluctuations, a phenomenon known as the "ghost convoy effect." Summary of the Invention

[0003] The purpose of this invention is to provide a method for individualized generation and trajectory fusion of ghost convoy vehicles in macro-micro transformation, which solves the problem of ghost convoy effect caused by the positive random point scattering method in the prior art when macro-to-micro transformation is carried out.

[0004] To achieve the above objectives, embodiments of the present invention provide a method for individualized generation and trajectory fusion of ghost convoy vehicles in macro-micro transformation, comprising: Within the target trajectory generation cycle of multiple trajectory generation cycles, the traffic state label of the target traffic flow is obtained based on the continuous field quantity cache of the target traffic flow; wherein, the continuous field quantity cache includes the physical data of the target traffic flow at the cellular scale; Microscopic vehicles are generated based on the traffic state labels, and the multidimensional state of the microscopic vehicles is obtained. The target trajectory with the highest candidate weight among multiple candidate trajectories generated in the first region is fused with the pre-acquired measured trajectory to obtain the completed trajectory; wherein, the first region is the region where the trajectory is discontinuous among multiple regions included in the target traffic flow; the multiple candidate trajectories are obtained by advancing according to the traffic wave propagation direction of the target traffic flow; the candidate weight is determined according to the comprehensive residual of the candidate trajectories; The completed trajectory and the micro vehicle trajectory are used together as the vehicle trajectory generation result of the target trajectory generation cycle and stored in the discrete individual quantity cache; wherein, the micro vehicle trajectory is generated based on the multidimensional state.

[0005] Optionally, the method further includes: Read the vehicle trajectory generation result from the discrete individual quantity cache, and obtain the traffic wave frequency domain features based on the vehicle trajectory generation result; The macroscopic velocity field of the target traffic flow is read from the continuous field quantity buffer; By performing two-dimensional spatiotemporal Fourier transforms on the frequency domain features of the traffic waves and the macroscopic velocity field, a ghost vehicle band consistency index is obtained. This ghost vehicle band consistency index represents the degree of matching between macroscopic basic map parameters and microscopic fluctuations. The macroscopic basic map parameters are the feature parameters of the macroscopic basic map corresponding to the target traffic flow. The microscopic fluctuations are the vehicle speed fluctuation statistics of the microscopic vehicles. If the band consistency index of the ghost vehicle fleet is less than a first preset threshold, target macroscopic basic map parameters are generated based on the band consistency index of the ghost vehicle fleet and historical macroscopic basic map parameters; wherein, the historical macroscopic basic map parameters are macroscopic basic map parameters obtained in a first trajectory generation cycle; the first trajectory generation cycle is a trajectory generation cycle that precedes the target trajectory generation cycle among the plurality of trajectory generation cycles; the target macroscopic basic map parameters are used to describe the macroscopic state of traffic flow in a second trajectory generation cycle; the second trajectory generation cycle is a trajectory generation cycle that follows the target trajectory generation cycle among the plurality of trajectory generation cycles.

[0006] Optionally, the method, wherein obtaining the ghost convoy band consistency index by performing two-dimensional spatiotemporal Fourier transforms on the traffic wave frequency domain features and the macroscopic velocity field respectively, includes: A two-dimensional spatiotemporal Fourier transform is performed on the frequency domain features of the traffic wave to obtain the transformed traffic wave frequency domain features; and a two-dimensional spatiotemporal Fourier transform is performed on the macroscopic velocity field to obtain the transformed macroscopic velocity field. Based on the frequency domain characteristics of the transformed traffic wave and the transformed macroscopic velocity field, the macro-micro frequency domain correlation is obtained; wherein, the macro-micro frequency domain correlation is used to characterize the frequency domain correlation between the macroscopic state corresponding to the target traffic flow and the microscopic state corresponding to the microscopic vehicle; The macro- and micro-frequency domain correlation is integrated within the characteristic beam interval of the ghost fleet to obtain the band consistency index of the ghost fleet.

[0007] Optionally, the method further includes: Based on the continuity equation of the target traffic flow, the traffic state to be estimated is obtained; Using the traffic state to be estimated as a constraint, the fused traffic state and vehicle state are obtained based on historical traffic state and multi-source observation data. If the target continuity residual generated by the target sensor among the multiple sensors corresponding to the multi-source observation data and the fused traffic state is greater than a second preset threshold, the sensor weight corresponding to the target sensor is reduced to obtain the final fused traffic state; wherein, the sensor weight is used to characterize the influence of the observation data obtained by the sensor on the final fused traffic state; The final fused traffic state is written into the continuous field quantity cache, and the vehicle state is written into the discrete individual quantity cache.

[0008] Optionally, the method, wherein, within a target trajectory generation cycle of multiple trajectory generation cycles, obtaining the traffic state label of the target traffic flow based on the continuous field volume cache of the target traffic flow includes: The downstream propagation speed, upstream propagation speed, and cross-flow propagation speed of the target traffic flow are obtained based on the continuous field quantity buffer. Based on the downstream propagation speed, the upstream propagation speed, and the cross-flow propagation speed, the traffic state of the target traffic flow corresponding to different cells is determined; The traffic status label is obtained based on the traffic status.

[0009] Optionally, the method, wherein generating micro-vehicles based on the traffic state labels and obtaining the multi-dimensional states of the micro-vehicles includes: Based on the traffic state label and the continuous field quantity cache, the number of vehicles to be generated within the target trajectory generation cycle is calculated at the transition boundary; wherein, the transition boundary is the boundary between the macro state corresponding to the target traffic flow and the micro state corresponding to the micro vehicle; Based on the equilibrium relationship of the car-following model, the micro-vehicles are generated according to the number of vehicles to be generated, and the initial state of the micro-vehicles is obtained. Based on the observation information of the micro-vehicle, the vehicle driving behavior type corresponding to the micro-vehicle is identified; wherein, different vehicle driving behavior types correspond to different driver parameters; Adjust the initial state of the vehicle according to the type of vehicle driving behavior to obtain the multi-dimensional state.

[0010] Optionally, the method, wherein fusing the target trajectory with the highest candidate weight among multiple candidate trajectories generated in the first region with the pre-acquired measured trajectory to obtain the completed trajectory, includes: In the first region, the multiple candidate trajectories are generated by advancing according to the traffic wave propagation direction under the traffic conditions corresponding to the sub-regions; wherein, the traffic conditions are determined according to the traffic condition labels; the first region is divided into multiple sub-regions according to the traffic condition labels; Based on the comprehensive residual, determine the candidate weight of each candidate trajectory among the plurality of candidate trajectories; Resampling is performed in the sub-region corresponding to each of the different traffic status labels to determine the candidate trajectory with the highest candidate weight as the target trajectory. The target trajectory is fused with the measured trajectory to obtain the completed trajectory.

[0011] To achieve the above objectives, embodiments of the present invention provide a device for individualized generation and trajectory fusion of ghost convoy vehicles in macro-micro transformation, comprising: The first acquisition module is used to acquire the traffic state label of the target traffic flow based on the continuous field quantity cache of the target traffic flow within the target trajectory generation cycle of multiple trajectory generation cycles; wherein, the continuous field quantity cache includes the physical data of the target traffic flow at the cellular scale. The second acquisition module is used to generate micro-vehicles based on the traffic state labels and acquire the multi-dimensional state of the micro-vehicles. The third acquisition module is used to fuse the target trajectory with the highest candidate weight among multiple candidate trajectories generated in the first region with the pre-acquired measured trajectory to obtain the completed trajectory; wherein, the first region is the region where the trajectory is discontinuous among multiple regions included in the target traffic flow; the multiple candidate trajectories are acquired according to the traffic wave propagation direction of the target traffic flow; the candidate weight is determined according to the comprehensive residual of the candidate trajectories; The first processing module is used to combine the completed trajectory and the micro vehicle trajectory as the vehicle trajectory generation result of the target trajectory generation cycle and store them in a discrete individual quantity cache; wherein, the micro vehicle trajectory is generated based on the multidimensional state.

[0012] To achieve the above objectives, embodiments of the present invention provide a device for individualized generation and trajectory fusion of ghost convoy vehicles in macro-micro transformation, comprising: a processor, a memory, and a program or instructions stored in the memory and executable on the processor; wherein, when the processor executes the program or instructions, it implements the method for individualized generation and trajectory fusion of ghost convoy vehicles in macro-micro transformation as described above.

[0013] To achieve the above objectives, embodiments of the present invention provide a readable storage medium storing a program or instructions thereon, wherein the program or instructions, when executed by a processor, implement the steps in the method for individualizing and fusion trajectories of ghost convoy vehicles in macro-micro transformation as described above.

[0014] To achieve the above objectives, embodiments of the present invention provide a computer program product, which includes computer instructions that, when executed by a processor, implement the steps of the method for individualizing and fusion trajectories of ghost convoy vehicles in macro-micro transformation as described above.

[0015] The beneficial effects of the above-described technical solution of the present invention are as follows: In this embodiment of the invention, traffic state labels are determined based on the physical data of the target traffic flow within the target trajectory generation cycle, thereby generating micro-vehicles. The target trajectory obtained in the region of discontinuous trajectory in the target traffic flow is advanced according to the direction of traffic wave propagation and fused with the pre-acquired measured trajectory. Together with the trajectory of the micro-vehicles, it serves as the vehicle trajectory generation result. Compared with the method of randomly scattering points from macro to micro, this method can reduce false speed waves and false queues, and improve the consistency of macro-micro conversion, data fusion accuracy, and trajectory completion quality. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the method for individualizing and fusing the trajectory of ghost convoy vehicles in macro-micro transformation according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the equilibrium state inverse mapping in the method for individualizing and fusing ghost convoy vehicles in macro-micro transformation according to an embodiment of the present invention; Figure 3 This is a schematic diagram of the device for individualizing and fusionting the trajectory of ghost convoy vehicles in macro-micro transformation according to an embodiment of the present invention. Detailed Implementation

[0017] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0018] It should be understood that the phrase "one embodiment" or "an embodiment" throughout the specification means that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of the invention. Therefore, "in one embodiment" or "in an embodiment" appearing throughout the specification do not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments.

[0019] In various embodiments of the present invention, it should be understood that the sequence number of each process described below does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0020] In addition, the terms "system" and "network" are often used interchangeably in this article.

[0021] In the embodiments provided by this invention, it should be understood that "B corresponding to A" means that B is associated with A, and B can be determined based on A. However, it should also be understood that determining B based on A does not mean that B is determined solely based on A; B can also be determined based on A and / or other information.

[0022] For ease of understanding, the following describes some aspects of the embodiments of the present invention: like Figure 1 As shown in the embodiment of the present invention, a method for individualized generation and trajectory fusion of ghost convoy vehicles in macro-micro transformation includes: Step S10: Within the target trajectory generation cycle of multiple trajectory generation cycles, obtain the traffic state label of the target traffic flow based on the continuous field quantity cache of the target traffic flow; wherein, the continuous field quantity cache includes the physical data of the target traffic flow at the cellular scale. It should be noted that the trajectory generation system corresponding to the method for individualized generation and trajectory fusion of ghost fleet vehicles in the macro-micro transformation described in this embodiment of the invention includes a traffic wave phase space state pipeline module, namely the first module, which is mainly responsible for establishing the unified data structure and state cache required for system operation. The traffic wave phase space state pipeline module unifies the continuous field quantity cache, micro-discrete vehicle states, statistical feature parameters, and spatial index relationships, providing a common data interface for subsequent modules. Among them, the continuous field quantity cache is used to record the density, speed, flow, and traffic state of each lane-cell; the discrete individual quantity cache is used to record the micro-information such as the position, speed, acceleration, lane, cell, and path of each vehicle; the statistical feature cache is used to record driver parameters, traffic wave frequency domain features, ghost fleet consistency index, and macro-basic graph parameters. The traffic wave phase space state pipeline module is equivalent to the data foundation and intermediate state expression layer of the entire system, and all subsequent calculation results are read and written back around this module. The first module establishes a unified state transfer standard between the macro, meso, and micro levels. Its function is not to create a separate, complex database, but rather to organize the traffic states that subsequent modules actually need into three types of caches within each macro-micro transition cycle: continuous field quantity cache, discrete individual quantity cache, and statistical feature cache. These caches are linked through the same timestamp, the same sub-region boundary number, and the same lane-cell index, enabling other modules to read and write states under the same physical standard.

[0023] Continuous field quantity buffer is denoted as Used to store density, speed, flow, and traffic status labels at the lane-cell scale: ; Where t is the current transition time; i is the lane-cell index; Let be the density of the i-th cell; This is the average velocity of the cell; This is the cell flow rate; Traffic status labels, including free flow, synchronized flow, or congestion; The effective traffic wave propagation speed or propagation mode is calculated in Module 2.

[0024] The discrete individual quantity buffer, denoted as D(t), is used to store the individual states of microscopic vehicles and vehicles to be generated: ; Where n is the vehicle index; A unique vehicle identification number; Number the route to which the vehicle belongs; and These are the vehicle's lane and cell, respectively; This refers to the seven-dimensional state of the vehicle (i.e., the multi-dimensional state). This is the net distance between this vehicle and the vehicle in front. The discrete individual quantity cache is written by the third module to generate the vehicle, the driving behavior parameters are updated by the fourth module, the vehicle state is read or corrected by the sixth and eighth modules, and the trajectory is written by the fifth module.

[0025] Statistical feature cache is denoted as This is used to store statistics that cannot be expressed by a single cell or a single vehicle: ; in, The posterior or classification parameters of the driver parameters estimated in the fourth module; The traffic wave frequency domain feature signature extracted from the sixth module; The macroscopic fundamental diagram parameters after calibration for the seventh module include free flow velocity, critical density, and stopping wave velocity.

[0026] The update sequence of the first module is as follows: first, it receives macroscopic path flow, mesoscopic cell occupancy, and microscopic vehicle status; then, it maps data from different sources to a unified lane-cell index; next, it performs consistency checks on vehicle count, boundary flow, and timestamps; finally, it publishes the state cache for the current transition cycle. Downstream modules only read the published cache and do not directly read the intermediate state being written, thereby avoiding spatiotemporal mismatches between macroscopic field quantities, microscopic vehicles, and sensor observations.

[0027] The traffic state label of the target traffic flow is obtained based on the continuous field quantity buffer of the target traffic flow. This is completed in the traffic wave propagation speed calculation and traffic state discrimination module, i.e., the second module. The second module reads the macroscopic density, speed, flow rate, and speed difference between adjacent lanes from the first module (i.e., the continuous field quantity buffer includes the physical data of the target traffic flow at the cellular scale), calculates the downstream propagation speed, the upstream return speed, and the lateral disturbance propagation speed, and determines whether the current lane-cell belongs to free flow, synchronous flow, or congestion state based on these propagation speeds. That is, the traffic state label includes free flow, synchronous flow, and congestion.

[0028] Step S20: Generate micro-vehicles based on the traffic state labels and obtain the multi-dimensional state of the micro-vehicles; It should be noted that the third module in the trajectory generation system is the equilibrium-state inverse mapping vehicle generation module, which mainly completes the generation from macroscopic traffic conditions to microscopic vehicle individuals. The third module reads the macroscopic target flow, density, and speed from the first module and the traffic state labels from the second module, and calculates the number of vehicles to be generated in the current cycle at the transition boundary. Subsequently, based on the equilibrium relationship of the car-following model, module three inversely solves the macroscopic density and speed into the expected spacing and speed of microscopic vehicles, and restricts the generation of disturbances through stable mode projection, ensuring that the newly generated vehicles fall within the stable attraction domain of car-following dynamics. The seven-dimensional states of the generated vehicles, including position, speed, heading angle, and acceleration, are written into the discrete individual quantity cache of the first module. In the fourth module, the driving behavior parameter identification and injection module, based on observation information such as microscopic vehicle trajectory, local density, speed difference, net distance, and acceleration, the vehicle driving behavior type is identified, such as conservative, neutral, and aggressive. Driver parameters such as expected speed, safe headway, maximum acceleration, comfortable deceleration, and reaction time are matched for different driving states, thereby updating the seven-dimensional state output in the third module to obtain the multi-dimensional state. This allows newly generated vehicles to no longer use fixed parameters, but are dynamically generated based on local traffic conditions (i.e., the corresponding traffic condition labels) and driving behavior distribution, thereby improving the realism and stability of the generated vehicle behavior.

[0029] Step S30: The target trajectory with the highest candidate weight among the multiple candidate trajectories generated in the first region is fused with the pre-acquired measured trajectory to obtain the completed trajectory; wherein, the first region is the region where the trajectory is discontinuous among the multiple regions included in the target traffic flow; the multiple candidate trajectories are obtained by advancing according to the traffic wave propagation direction of the target traffic flow; the candidate weight is determined according to the comprehensive residual of the candidate trajectories; It should be noted that in the fifth module, the ghost vehicle trajectory completion and fusion module, the target trajectory with the highest candidate weight among multiple candidate trajectories generated in the first region is fused with the pre-acquired measured trajectory to obtain the completed trajectory. The fifth module is used to generate candidate trajectories based on the traffic wave propagation direction and local traffic conditions in areas where microscopic trajectories are missing, discontinuous, or sparsely observed (i.e., the first region). The fifth module reads the traffic conditions (i.e., the traffic condition labels) and propagation speed output by the second module, advances the candidate trajectory along the downstream, lateral coupling, or upstream direction, and calculates the candidate weight based on the matching degree between the candidate trajectory and the measured vehicle in terms of spatial position, speed, wave energy, and continuity residual (i.e., the comprehensive residual). Finally, the fifth module fuses the high-weight candidate trajectory with the existing measured trajectory to complete the missing vehicle trajectory.

[0030] Step S40: The completed trajectory and the micro vehicle trajectory are used together as the vehicle trajectory generation result of the target trajectory generation cycle and stored in the discrete individual quantity cache; wherein, the micro vehicle trajectory is generated based on the multidimensional state; It should be noted that in the fifth module, the obtained completed trajectory is written into the discrete individual quantity cache of the first module, and together with the micro vehicle trajectory generated by the multidimensional state of the micro vehicle, it serves as the vehicle trajectory generation result and is stored in the discrete individual quantity cache.

[0031] In this embodiment, traffic state labels are determined based on the physical data of the target traffic flow within the target trajectory generation cycle, thereby generating micro-vehicles. The target trajectory obtained in the region of discontinuous trajectory in the target traffic flow is advanced according to the direction of traffic wave propagation and fused with the pre-acquired measured trajectory. Together with the trajectory of the micro-vehicles, it serves as the vehicle trajectory generation result. Compared with the method of randomly scattering points from macro to micro, this can reduce false speed waves and false queues, and improve the consistency of macro-micro conversion, data fusion accuracy, and trajectory completion quality.

[0032] The trajectory generation system corresponding to the method for individualizing and fusing vehicle convoys in the macro-micro transformation described in this invention consists of eight modules. These modules do not communicate directly but instead use a unified three-level buffer of a "state pipeline" for cascading constraint transmission. The macro, meso, and micro layers of the trajectory generation system all correspond to existing simulation engines, with bidirectional transformations occurring at the junctions between them. The macro-level layer is a network traffic allocation based on the origin-destination matrix (OD) and column flow, providing the flow rate and travel demand for each path. The meso-level layer is a lane-cell loading engine that expands each physical link into multiple lanes, and each lane is further divided into cells, ensuring that each vehicle has a precise position in the "link-lane-cell" triplet. It uses cellular automata-style, provably conserved occupancy updates to drive traffic flow, directly characterizing stop-line queuing under signal control, spillback from upstream links, multi-lane occupancy differences, and detector-level flow / speed / occupancy time series. The micro-level layer is a single-vehicle-level engine driven by the path files allocated from the macro-simulation, with each vehicle in a seven-dimensional state. The progression is recorded every 0.1 seconds, driven by a hybrid of neural networks and rules. It includes seven states. The components are: , The coordinates (in meters) represent the vehicle's global position. The orientation angle (in radians). , For velocity components (m / s). , The acceleration component is (m / s²).

[0033] The so-called macro-micro bidirectional conversion refers to two types of mappings completed at the subarea boundary (i.e., the conversion boundary): converting the density / velocity / flow field carried by the meso-level cells into individual micro-level vehicles and their initial states (macro / meso-level → micro-level), and aggregating the micro-level seven-dimensional trajectories back into the meso-level cell field and macro-level moment parameters (micro-level → macro / meso-level). The ghost convoy effect occurs precisely at this subarea boundary—when converting to the micro-level, vehicles appear out of nowhere at the boundary, and the initial states are inconsistent with the equilibrium flow, causing false shock waves; when converting to the macro-level, the trajectory of a single vehicle and the statistical moment information are averaged out and lost. The modules of this system revolve around "artifact-free bidirectional conversion at the subarea boundary."

[0034] From the data interface perspective, the inputs of this system are: OD and path flow at the macro level, cell-by-cell occupancy of the meso-level cellular engine (based on which density / velocity / flow field is calculated), seven-dimensional vehicle state and traffic light state at the micro level, and multi-source observations such as coils / video / detection vehicles; the outputs are: initial vehicle state and completed trajectory generated at the sub-region boundary in the micro-direction, density / velocity / flow field and traffic wave frequency domain feature signature obtained by aggregation in the macro direction, and macro-micro consistency index and calibrated basic graph parameters during runtime. Each module is cascaded in a fixed time sequence: in each conversion cycle, module 2 first determines the traffic status and writes it into the cache; in the macro to micro direction, module 3 generates the initial vehicle state, module 4 injects heterogeneous driving behavior, and module 5 completes the individual trajectory when necessary; in the micro to macro direction, module 6 performs moment-preserving kernel aggregation; then module 7 performs frequency domain consistency verification and triggers directional calibration, and module 8 performs conservation constraint fusion on multi-source observations; all intermediate results are passed between modules through the three-level cache of module 1, thereby connecting the three types of engines—macro-allocation, meso-cell, and micro-vehicle—that have been implemented on a unified boundary.

[0035] The path file that drives the micro-level to generate vehicles comes directly from the traffic assignment results of the macro-simulation model and is not a separately agreed-upon special format. This file consists of a set of path records, each corresponding to an OD path obtained from the macro-assignment. The main fields include: (1) Path number, which is unique in the entire network and serves as an index for vehicle generation and road network matching on this path; (2) Start point, which is the starting position of the path, represented by the starting traffic cell / node or the starting link; (3) End point, which is the target position of the path, represented by the target traffic cell / node or the ending link; (4) Number of vehicles, which is the total number of vehicles to be generated on this path during the simulation period, obtained by multiplying the traffic flow assigned to this path by the time period length, and determining the number and rhythm of vehicles boarding along this path; (5) The driving path corresponding to the road network, which is the sequence of links (or lanes) arranged in driving order from the start point to the end point. It is the ordered road network unit that the vehicle passes through, and the micro-engine uses this to attach the vehicle to the specific lane sequence for driving. In addition, optional fields such as departure time distribution, initial velocity, and start point offset may be included. After reading the file, the micro-engine generates vehicles sequentially at the starting point of the corresponding driving path according to the path number, and drives to the destination along the ordered link sequence; adjusting the requirements of a certain OD only requires modifying the vehicle number segment of the corresponding path record, without modifying the engine itself.

[0036] The connection between the macroscopic continuous field and the mesoscopic cellular state is achieved through occupancy conversion: the number of vehicles in each cell is normalized according to the number of lanes and the cell length to obtain the local density field, and the average of the vehicle speeds in the cell is taken to obtain the local velocity field, which serves as the input for the macroscopic state generated by the second module's discrimination and the third module's inverse mapping. ; In the formula: , , The first Local density, average velocity, and flow rate at each cell; This represents the number of vehicles within the cell (along the cross section); The vertical length of the cell; For the first in a cell The vehicle's speed. This conversion ensures a one-to-one correspondence between the mesoscopic cell occupancy and the macroscopic LWR field quantity, while the microscopic side directly provides the position and velocity from the vehicle's seven-dimensional state. All three layers maintain consistency in scope under the same set of field quantity definitions.

[0037] Optionally, the method further includes: Read the vehicle trajectory generation result from the discrete individual quantity cache, and obtain the traffic wave frequency domain features based on the vehicle trajectory generation result; The macroscopic velocity field of the target traffic flow is read from the continuous field quantity buffer; By performing two-dimensional spatiotemporal Fourier transforms on the frequency domain features of the traffic waves and the macroscopic velocity field, a ghost vehicle band consistency index is obtained. This ghost vehicle band consistency index represents the degree of matching between macroscopic basic map parameters and microscopic fluctuations. The macroscopic basic map parameters are the feature parameters of the macroscopic basic map corresponding to the target traffic flow. The microscopic fluctuations are the vehicle speed fluctuation statistics of the microscopic vehicles. If the band consistency index of the ghost vehicle fleet is less than a first preset threshold, target macroscopic basic map parameters are generated based on the band consistency index of the ghost vehicle fleet and historical macroscopic basic map parameters; wherein, the historical macroscopic basic map parameters are macroscopic basic map parameters obtained in a first trajectory generation cycle; the first trajectory generation cycle is a trajectory generation cycle that precedes the target trajectory generation cycle among the plurality of trajectory generation cycles; the target macroscopic basic map parameters are used to describe the macroscopic state of traffic flow in a second trajectory generation cycle; the second trajectory generation cycle is a trajectory generation cycle that follows the target trajectory generation cycle among the plurality of trajectory generation cycles.

[0038] In this embodiment, in the sixth module of the trajectory generation system, the micro-trajectory aggregation and traffic wave feature extraction module, the traffic wave frequency domain features are obtained from the vehicle trajectory generation results cached in the discrete individual quantity. Micro-vehicle trajectories (i.e., the vehicle trajectory generation results) and vehicle states (i.e., the multi-dimensional states) are read from the first module. Vehicles are spatially aggregated according to lane-cell, and density, average speed, flow rate, and speed fluctuation intensity are calculated, achieving a micro-to-macro state write-back. Simultaneously, the sixth module further performs traffic wave response kernel decomposition or frequency domain analysis on the speed time series, extracting traffic wave energy, phase, and band features to form a traffic wave frequency domain feature signature. The sixth module transforms the micro-vehicle operation results into a macro-continuous field and traffic wave features for subsequent consistency verification between micro and macro levels, adjusting traffic flow in the next trajectory generation cycle, and achieving bidirectional conversion between micro and macro levels.

[0039] Traffic wave frequency domain features are obtained based on the vehicle trajectory generation results. These features are used to retain both first-order average field quantity and second-order fluctuation information during the conversion of microscopic vehicle trajectories to macroscopic or mesoscopic field quantities. Traditional aggregation typically only calculates average density and average speed, easily losing dynamic features such as speed waves, parking waves, and ghost vehicle wavebands. The sixth module adds traffic wave response kernel decomposition in addition to first-order aggregation, allowing microscopic fluctuations to be included in subsequent frequency domain consistency checks. The first-layer aggregation is performed within a lane-cell spatiotemporal grid. For the i-th cell, if its length is... The number of vehicles within a cell is Then the density, average velocity, and flow rate are respectively: , , ; in, The set of vehicles located within cell i; Let n be the speed of vehicle n; , , These represent the density, average velocity, and flow rate obtained from polymerization, respectively. At this time, the cell velocity can be obtained by interpolation from adjacent cells or by preserving the velocity from the previous time step, and is marked as low confidence in the cache. To preserve second-order fluctuation information, the sixth module calculates the intra-cell velocity variance: ; in, The velocity fluctuation intensity is used to represent the degree of dispersion of vehicle velocity around the average velocity within the same cell.

[0040] The second layer uses traffic wave response to decompose the velocity time series. Let... Here is the traffic wave response kernel at scale a: , ; Where c is the characteristic propagation velocity of the traffic wave, which can be given by the second module; D is the equivalent diffusion coefficient; a is the decomposition scale; and h(x,t;c,D) is the impulse response function of the linearized traffic wave equation. Unlike general wavelets, this kernel function explicitly includes the traffic wave propagation velocity and diffusion characteristics. For the cellular velocity sequence... Perform convolution to obtain the response sequence at scale a: .

[0041] Then, the response energy and phase at this scale are calculated: , ; in, The traffic wave response energy at scale a; For the corresponding phase; The time sampling interval; This is a Fast Fourier Transform. Ultimately, this module generates a signature for the traffic wave frequency domain characteristics (i.e., the traffic wave frequency domain features): .

[0042] The signature is written into the statistical feature cache of the first module for the seventh module to perform macro-micro consistency verification in the feature band of the ghost car fleet.

[0043] In the seventh module, the Ghost Fleet Consistency Calibration Module, a consistent index for the ghost fleet bands is obtained by performing two-dimensional spatiotemporal Fourier transforms on the traffic wave frequency domain characteristics and the macroscopic velocity field, respectively. The seventh module reads the macroscopic velocity field and the velocity fluctuation field obtained from the aggregation of microscopic trajectories, compares them in the spatiotemporal frequency domain, and calculates the consistency index of the macroscopic and microscopic aspects within the characteristic bands of the ghost fleet. When the consistency index is below a threshold, it indicates that the macroscopic basic map parameters or microscopic behavioral parameters do not match the actual microscopic fluctuations. Based on this, the seventh module performs directional calibration on macroscopic basic map parameters such as free-flow velocity, critical density, and parking wave velocity. The calibrated parameters are written back to the first module and used by the second and third modules in the next cycle (i.e., the second trajectory generation cycle) to describe the macroscopic state of traffic flow in the second trajectory generation cycle, thus forming a parameter closed loop driven by traffic wave consistency.

[0044] When the ghost fleet band consistency index Below the threshold When the threshold (i.e., the first preset threshold) is reached, this module applies macroscopic basic graph parameters. Perform orientation calibration, among which For free flow velocity, Let be the critical density, and w be the propagation velocity of the parking wave. The calibration objective is to reduce the band inconsistency loss of the ghost convoy. , ; in, Losses due to inconsistency in frequency bands of the ghost convoy. For parameter calibration step size, For the loss function against the basic graph parameters The gradient. After calibration. The statistical feature cache of the first module is written back and used by the second and third modules in the next conversion cycle (i.e. the second trajectory generation cycle), thus forming a runtime closed loop.

[0045] Optionally, the method, wherein obtaining the ghost convoy band consistency index by performing two-dimensional spatiotemporal Fourier transforms on the traffic wave frequency domain features and the macroscopic velocity field respectively, includes: A two-dimensional spatiotemporal Fourier transform is performed on the frequency domain features of the traffic wave to obtain the transformed traffic wave frequency domain features; and a two-dimensional spatiotemporal Fourier transform is performed on the macroscopic velocity field to obtain the transformed macroscopic velocity field. Based on the frequency domain characteristics of the transformed traffic wave and the transformed macroscopic velocity field, the macro-micro frequency domain correlation is obtained; wherein, the macro-micro frequency domain correlation is used to characterize the frequency domain correlation between the macroscopic state corresponding to the target traffic flow and the microscopic state corresponding to the microscopic vehicle; The macro- and micro-frequency domain correlation is integrated within the characteristic beam interval of the ghost fleet to obtain the band consistency index of the ghost fleet.

[0046] In this embodiment, the seventh module compares whether the macroscopic velocity field is consistent with the frequency domain characteristics of the traffic wave obtained by aggregating microscopic trajectories, and triggers directional calibration of the macroscopic basic map parameters on the characteristic band of the ghost convoy. The core idea is that the fluctuations of the ghost convoy are both a non-physical phenomenon that needs to be suppressed and a sensitive probe for the consistency between macroscopic and microscopic transformations.

[0047] The seventh module reads the macroscopic velocity field from the continuous field buffer. Read the velocity fluctuation field obtained from the micro-trajectory from the statistical feature cache. Or the signature of the transformed traffic wave frequency domain characteristics. A two-dimensional spatiotemporal Fourier transform is performed on both (in the formula, the velocity wave field obtained from the microscopic trajectory is used). For example, the signature of the traffic wave frequency domain feature is also used in the same way. , ; Where k is the space wavenumber, For time frequency, This indicates that a Fourier transform is performed on both the spatial and temporal dimensions.

[0048] Macro- and micro-frequency domain correlation is defined as amplitude coherence: ; in, For macroscopic and microscopic velocity fields at wavenumber k and frequency The relevance is measured in terms of the number of points, ranging from 0 to 1. ∠ represents complex conjugate; ∠<·> represents the average over a local wavenumber-frequency window or multiple sampling periods.

[0049] The characteristics of ghost convoys typically correspond to spatial wavelengths of approximately 300 to 2000 meters. (Note:) The characteristic wavenumber range of the ghost convoy is: .

[0050] This module calculates the integral consistency index in the characteristic wavenumber range of ghost car teams: ; in, For the consistency index of ghost car fleets across different bands; For frequency The wavenumber correlation obtained after averaging or selecting the target frequency band (i.e., the macro-micro frequency domain correlation). The closer it is to 1, the more consistent the macroscopic and microscopic phenomena are in the ghost convoy band; The lower the value, the more it indicates a mismatch between the macroscopic fundamental parameters or microscopic behavioral parameters and the actual microscopic fluctuations.

[0051] Optionally, the method further includes: Based on the continuity equation of the target traffic flow, the traffic state to be estimated is obtained; Using the traffic state to be estimated as a constraint, the fused traffic state and vehicle state are obtained based on historical traffic state and multi-source observation data. If the target continuity residual generated by the target sensor among the multiple sensors corresponding to the multi-source observation data and the fused traffic state is greater than a second preset threshold, the sensor weight corresponding to the target sensor is reduced to obtain the final fused traffic state; wherein, the sensor weight is used to characterize the influence of the observation data obtained by the sensor on the final fused traffic state; The final fused traffic state is written into the continuous field quantity cache, and the vehicle state is written into the discrete individual quantity cache.

[0052] In this embodiment, the eighth module of the trajectory generation system, the continuity equation-constrained state estimation module, receives multi-source observation data from coils, video, detection vehicles, and simulation outputs, and maps them onto a unified lane-cell grid. The eighth module introduces the traffic flow continuity equation as a constraint during the fusion process to check whether the density change and flow rate difference satisfy the conservation relationship. For observation sources with large continuity residuals, the eighth module amplifies their observation noise covariance, reducing their weight in the fusion process, thereby suppressing the impact of abnormal observations on the system state. The fused density, speed, flow rate, and vehicle state are written back to the first module, providing more reliable state input for subsequent vehicle generation, traffic wave recognition, and trajectory completion. The system fuses multi-source observations from detection vehicle GPS, fixed coils, video detectors, and microscopic simulation trajectories, and enforces the satisfaction of the traffic flow continuity equation during the fusion process to avoid non-physical phenomena such as non-conservation of vehicle numbers, abrupt density changes, or discontinuous flow rates in the fusion results. The traffic flow continuity equation is: ; in, Let be the density, q be the flow rate, x be the spatial location, and t be the time. Discretizing this on a lane-cell mesh yields the continuity residual of the i-th cell at the k-th time step: ; Where Y is the set of traffic states to be estimated (i.e., the traffic states to be estimated); For time step; The spatial step size; and Let be the density and flux of the i-th cell at the k-th time step, respectively.

[0053] This module expresses state estimation as an optimization problem with equality constraints: , ; in, P represents the prior state (i.e., the historical traffic state) obtained from the previous time step or model prediction; P is the prior error covariance. For observations by the m-th type of sensor; This is the observation matrix that maps state Y to the sensor's observation space; The measurement noise covariance of this sensor. The first constraint ensures the estimation result does not deviate from the predicted prior; the second constraint ensures the estimation result closely approximates multi-source observations; equality constraints apply. Traffic flow conservation is ensured. This constraint optimization can be solved using the alternating direction multiplier method: first, update the state Y with the continuity constraint multiplier fixed, then update the multiplier based on the continuity residual, and repeat the iteration until the residual is below the threshold.

[0054] For anomalous sensors, this module automatically reduces weights based on continuous residuals. If the m-th sensor observation produces a residual after being individually substituted into the state estimation: And satisfy This amplifies the noise covariance of the sensor. ; in, Let m be the continuous residual corresponding to the m-th sensor; The residual threshold (i.e., the second preset threshold); This is the noise amplification factor. Because... After the increase, the sensor weight in the optimization is reduced, thus weakening the influence of anomalous sensors on the final fused state. The fused density, velocity, flow rate (i.e., the final fused traffic state) and vehicle state are written back to the continuous field quantity cache and the discrete individual quantity cache, respectively.

[0055] Optionally, the method, wherein step S10 includes: The downstream propagation speed, upstream propagation speed, and cross-flow propagation speed of the target traffic flow are obtained based on the continuous field quantity buffer. Based on the downstream propagation speed, the upstream propagation speed, and the cross-flow propagation speed, the traffic state of the target traffic flow corresponding to different cells is determined; The traffic status label is obtained based on the traffic status.

[0056] In this embodiment, the continuous field quantity cache from the first module... Read density ,speed Based on the speed difference with adjacent lanes, the effective disturbance propagation speed in the upstream, downstream, and lateral directions is calculated, and the current lane-cell is determined to be in a free-flow, synchronous flow, or congested state. The traffic state label... It will serve as the input for the third module vehicle generation, the fourth module driver parameter switching, and the fifth module trajectory completion.

[0057] For the i-th lane-cell, the propagation velocity of small disturbances in the downstream direction is represented by the characteristic linear velocity of the macroscopic fundamental graph. Under the Greenshields type fundamental graph: , ; in, For density is Traffic flow at that time; For free flow velocity; Blocking density; Let be the downstream propagation velocity of the i-th cell. When the density is low, A positive value usually indicates that the disturbance propagates downstream with the traffic flow; when the density is high, It can turn negative, indicating that the disturbance has a tendency to propagate upstream.

[0058] The reverse transmission speed is no longer simply fixed as a negative value, but is controlled by the congestion activation coefficient: , ; in, is the critical density; w is a positive value of the parking wave return velocity; This represents the congestion activation coefficient. This indicates that the value is limited to between 0 and 1. When the free-flow density is below the critical density, When the congestion density approaches zero, the backflow mode is not activated; when the congestion density increases, As the flow rate increases, the countercurrent return mode gradually strengthens. This helps prevent the countercurrent return mode from being falsely activated in free-flow conditions.

[0059] Lateral disturbance propagation velocity is used to characterize the diffusion of disturbances caused by lane changes or speed differences between adjacent lanes: ; in, This is the lateral disturbance influence coefficient; The average speed difference between adjacent lanes; Lane width; This refers to the lateral propagation speed.

[0060] This module combines the above three propagation quantities into a set of effective propagation modes: .

[0061] The traffic condition determination rule is: when , When the lateral velocity difference is less than the threshold and the lateral velocity difference is also less than the threshold, it is considered free flow; when Still positive but or When the threshold is exceeded, it is determined to be a synchronous stream; when and When the threshold is exceeded, it is determined to be congested. The resulting state is then analyzed. (i.e., the traffic status label) is written back to the continuous field quantity cache.

[0062] Optionally, the method, wherein step S20 includes: Based on the traffic state label and the continuous field quantity cache, the number of vehicles to be generated within the target trajectory generation cycle is calculated at the transition boundary; wherein, the transition boundary is the boundary between the macro state corresponding to the target traffic flow and the micro state corresponding to the micro vehicle; Based on the equilibrium relationship of the car-following model, the micro-vehicles are generated according to the number of vehicles to be generated, and the initial state of the micro-vehicles is obtained. Based on the observation information of the micro-vehicle, the vehicle driving behavior type corresponding to the micro-vehicle is identified; wherein, different vehicle driving behavior types correspond to different driver parameters; Adjust the initial state of the vehicle according to the type of vehicle driving behavior to obtain the multi-dimensional state.

[0063] In this embodiment, the third module calculates the number of vehicles to be generated within the target trajectory generation cycle at the transition boundary based on the traffic state label and the continuous field quantity cache. Furthermore, based on the equilibrium relationship of the car-following model, the micro-vehicles are generated according to the number of vehicles to be generated, and the initial state of the micro-vehicles is obtained. When transitioning from macroscopic or mesoscopic traffic states to microscopic vehicle states, a set of micro-vehicles with clear location, speed, acceleration, lane affiliation, and path affiliation is generated based on density, speed, flow rate, and lane capacity constraints at the sub-region boundary. Unlike traditional vehicle generation methods that randomly scatter points based on density, this module first calculates the equilibrium distance of vehicles under the current traffic state based on the car-following dynamics model, and then constrains the initial state of the vehicles within a stable modal subspace near the equilibrium point. This ensures that the generated vehicles satisfy local dynamic stability conditions upon entering the microscopic simulation region, suppressing non-physical speed fluctuations and ghost convoy effects caused by inconsistencies in initial distance and speed from the source. Figure 2 As shown, in the equilibrium inverse mapping: the initial state is arranged along the stable mode to eliminate the ghost convoy. For the region boundary b, this module in the current conversion cycle... Internally, based on the macro target flow Calculate the number of vehicles to be generated: ; in, The number of vehicles that should be generated within boundary b in the current conversion cycle (i.e., the target trajectory generation cycle); This is the macro-micro transformation cycle; This is the vehicle count margin generated by rounding in the previous conversion cycle (i.e., the second trajectory generation cycle), used to ensure that the cumulative number of vehicles generated across multiple cycles (i.e., the multiple trajectory generation cycles) is consistent with the macro traffic flow.

[0064] This module uses an Intelligent Driver Model (IDM) to describe vehicle acceleration while following the car: , ; in, v is the acceleration due to following the car; v is the speed of the vehicle itself. s represents the desired speed; s represents the net distance between this vehicle and the vehicle in front. This is the speed difference between this vehicle and the vehicle in front; The speed of the vehicle in front; b is the maximum acceleration; α is the comfortable deceleration; delta is the acceleration exponent. is the minimum clear distance; T is the safe headway.

[0065] In equilibrium, the speed of this vehicle is the same as that of the vehicle in front, that is... And the vehicle's acceleration is zero, that is Therefore, the net equilibrium spacing corresponding to the velocity v can be obtained by inverse solving: ; in, This is the net clearance for equilibrium at velocity v. When v approaches... If the denominator is too small, this module will truncate v to a value slightly smaller than 1. The effective velocity is calculated first to avoid divergence of the balance spacing.

[0066] For lane-cell i, if the macroscopic density Calculated in "vehicles / meters / lane", the target net spacing is then converted from density: ; in, The target net spacing is calculated from the macro density. This refers to the average length of the vehicle. To prevent extremely small positive numbers with zero density, the macroscopic target state and the carousel equilibrium state are constructed as follows: , , ; in, The target state vector is constructed from macroscopic density and velocity; This is the equilibrium state vector corresponding to the car-following model; This represents the deviation of the macroscopic state from the race-following equilibrium state.

[0067] In equilibrium state In the vicinity, the carousel dynamics will be linearized. Let the local state vector be... Its dynamic form is Then the Jacobian matrix at the equilibrium point is: .

[0068] right Find the eigenvalues ​​and eigenvectors: ; in, For the m-th eigenvalue, For the corresponding eigenvectors. When When this eigenvector is selected, it is considered a stable mode direction. All stable mode directions form the stable mode matrix. .

[0069] To prevent macroscopic state deviations from being injected into the microscopic vehicle along the unstable direction, this module will Projected onto the stable mode subspace: , ; in, For the stable modal projection matrix, This refers to the component of the macroscopic deviation located within the stable mode subspace. If no eigenvectors satisfying the stability condition exist, a conservative generation method is adopted: using... Based on the basic spacing, Set the base speed and the disturbance amplitude to the preset minimum value.

[0070] To ensure that the generated vehicles satisfy both stability and individual variability, this module introduces small random perturbations into the stable mode subspace. The perturbation strength is determined by the local convergence rate near the equilibrium point. Sure: , , ; in, The standard deviation of the spacing disturbance; The standard deviation of the velocity disturbance; and The standard deviation of the baseline disturbance; This is the convergence rate adjustment coefficient; This indicates that the disturbance will be limited to a preset upper and lower limit. When generating the m-th vehicle, the disturbance is sampled first. Then, the perturbation is projected onto the stable mode subspace to obtain the initial spacing and velocity: .

[0071] The vehicles are then positioned along the lane centerline. Let the coordinate function of the lane centerline be... d is the arc length coordinate. Let the coordinates of the initial arc length at the boundary be , then the position of the m-th vehicle is: , , , .

[0072] If the initial state is taken as equilibrium acceleration, then , If it is necessary to preserve local acceleration differences, then... Tangential decomposition along the lane yields and This yields the vehicle's seven-dimensional state (i.e., the initial state). .

[0073] Before generating the vehicle and writing it into the microscopic simulation region, safety constraints and capacity constraints must be met: ; ; in, Minimum safe clearance; The maximum speed allowed on the road; The maximum permissible acceleration or deceleration amplitude; This represents the number of vehicles already in lane l. This represents the number of vehicles to be generated in the current period. Let lane l be the maximum number of vehicles allowed to be accommodated within the current spatial range. Candidate vehicles that do not meet the constraints are not written into the microscopic simulation region, and the number of ungenerated vehicles is carried over as a margin to the next conversion cycle (i.e., the second trajectory generation cycle).

[0074] To ensure that the micro-level generated flow rate is consistent with the macro-level target flow rate, this module employs model predictive control at the transition boundary to adjust the vehicle generation rate. : , ; in, To predict the actual microscopic flow generated or passing through the boundary in step k; For macroscopic target flow; To prevent extremely small positive numbers with a denominator of zero; gamma is the smoothing weight for the generation rate; The allowable relative flow error threshold, for example, 3%.

[0075] In the fourth module, based on the observation information of the micro-vehicles, the driving behavior type corresponding to the micro-vehicles is identified, and the initial state of the vehicle is adjusted according to the driving behavior type to obtain the multi-dimensional state. Based on the micro-vehicle trajectory and local traffic conditions, the driving behavior type is estimated in real time, and the estimated driver parameters are fed back to the third module and the micro-car-following model. This is to avoid caching with the statistical features of the first module. Confusion, the driver's hidden state (i.e., the type of vehicle driving behavior) is denoted as Its values ​​include conservative C, neutral N, and aggressive A. The observation (i.e., the observation information) is denoted as... Including the speed of this vehicle Speed ​​difference relative to the vehicle in front Net spacing The acceleration of this vehicle and local density .

[0076] The hidden state transition probability varies with density. Let... Given the baseline probability of transitioning from state i to state j, the density-corrected unnormalized probability is: , ; in, The density of the cell containing the vehicle; It is the critical density; The coefficient representing the influence of density on state transition; This represents the normalized transition probability. It is set by... This can increase the probability of a shift from an aggressive to a neutral or conservative type under high density.

[0077] Each driver state s corresponds to a set of car-following parameters: ; in, Let be the desired velocity in state s; For safe headway; This is the maximum acceleration; Decelerate for comfort; This is the overexcitation coefficient for braking; This refers to the reaction time.

[0078] This module uses observational likelihood. Let represent the probability that a vehicle generates the current observation in state s, and use the Viterbi algorithm to find the most likely sequence of hidden states: ; in, This represents the most likely sequence of hidden states of the driver from the initial time to the current time. This corresponds to the observation sequence. After determining the current hidden state, this module... Inject vehicle following model.

[0079] To characterize the excessive braking of some drivers when approaching the vehicle in front, this module superimposes an excessive braking term onto the IDM acceleration: ; in, Let n be the final acceleration of vehicle n; The IDM acceleration is calculated using the parameters corresponding to state s; This is the speed of the vehicle; The speed of the vehicle in front; This is an indicator function; it takes the value 1 if the condition is met, and 0 otherwise. This term is activated only when the vehicle is faster than the vehicle in front, simulating the driver's additional deceleration response. Replace the acceleration in the initial state with... Thus, the multidimensional state is obtained.

[0080] When the sixth or seventh module detects that the local velocity gradient energy exceeds the threshold, the fourth module limits the generation of newly generated vehicles. Not exceeding the stable upper limit, and will Adjusting the level to be no less than the lower limit of stability reduces the risk of amplification of non-physical disturbances from a behavioral perspective.

[0081] Optionally, the method, wherein step S30 includes: In the first region, the multiple candidate trajectories are generated by advancing according to the traffic wave propagation direction under the traffic conditions corresponding to the sub-regions; wherein, the traffic conditions are determined according to the traffic condition labels; the first region is divided into multiple sub-regions according to the traffic condition labels; Based on the comprehensive residual, determine the candidate weight of each candidate trajectory among the plurality of candidate trajectories; Resampling is performed in the sub-region corresponding to each of the different traffic status labels to determine the candidate trajectory with the highest candidate weight as the target trajectory. The target trajectory is fused with the measured trajectory to obtain the completed trajectory.

[0082] In this embodiment, in the fifth module, when the macroscopic to microscopic or multi-source observations are incomplete, i.e., when there is a discontinuous trajectory in the first region, the missing individual vehicle trajectories are completed, thus obtaining the completed trajectory. Unlike free random walk trajectory generation, this module advances the candidate trajectory along the traffic wave propagation direction given by the second module, so that the completed trajectory conforms to the traffic wave propagation law and is simultaneously consistent with the measured trajectory, wave energy, and continuity constraints.

[0083] Let the position of the c-th candidate trajectory at time t be... The effective propagation speed of the traffic wave under the current traffic condition z is: The unit vectors along the propagation direction and perpendicular to the propagation direction are respectively and The candidate trajectory is then advanced according to the following formula: , ; in, The time step for trajectory generation; Let be the lateral disturbance intensity under traffic condition z; These are standard normally distributed random numbers. They are valid under free flow, synchronized flow, and congested conditions. Take the effective propagation speed of downstream, lateral coupling, or upstream output from module 2, respectively.

[0084] The weight of each candidate trajectory is determined by spatial deviation, velocity deviation, fluctuation energy deviation, and continuity residual (i.e., the comprehensive residual includes spatial deviation, velocity deviation, fluctuation energy deviation, and continuity residual). , , , , ; in, The candidate weights are those for the c-th candidate trajectory; and These represent the positions and speeds of nearby measured vehicles; For candidate trajectory velocities; The fluctuation energy corresponding to the candidate trajectory; This refers to the wave energy obtained through actual measurement or aggregation. The continuity residual introduced for this candidate trajectory; , , and These are the space, velocity, energy, and residual scale parameters, respectively.

[0085] After the candidate weights are normalized, resampling is performed in the fifth module within the free flow, synchronous flow, and congested zones, respectively. Each zone retains high-weight candidate trajectories to prevent candidate trajectories from occupying all samples in a particular traffic state area. The final output completed trajectory is written into the discrete individual quantity buffer and aggregated with the actual observed trajectory in the sixth module.

[0086] Here is an example of a hybrid simulation scenario for a highway: The simulated road segment is approximately 10 kilometers long, with three lanes in one direction. Several loop detectors, video detection equipment, and a small number of detection vehicles are deployed along the route, with a detection vehicle penetration rate of approximately 5%. The simulation period is the morning rush hour, lasting approximately 2 hours, with a density varying between 15 and 85 vehicles / km / lane, covering three traffic states: free flow, synchronized flow, and congestion. For the macroscopic region, a first-order macroscopic (continuous medium) (Lighthill-Whitham-Richards Model, LWR) model is used for discretization with a spatial step size of meters and a time step size of seconds. For the mesoscopic region, a lane-cell loading engine is used to update cell occupancy. For the microscopic region, an extended IDM is used to advance the single-vehicle trajectory with a step size of 0.1 seconds. The macro-micro transition boundary is set at several mileage locations within the road segment.

[0087] The main parameter can be taken as: free flow velocity km / h, congestion density Vehicles / km / lane, critical density Vehicles / km / lane, parking wave return speed km / h, lateral disturbance influence coefficient Average vehicle length Values ​​are taken from vehicle model. IDM parameters. T , b and delta are set separately for conservative, neutral, and aggressive drivers. The characteristic spatial wavelength of the ghost convoy is taken as 300 to 2000 meters, corresponding to the wavenumber range. Ghost car band consistency threshold 0.75 is acceptable.

[0088] The typical operation process is as follows: In the free flow phase, the second module calculates... , Approaching 0 and with a small lateral velocity difference, status label It is determined to be a free flow; the third module generates vehicles by inverse mapping of equilibrium state, and uses small spacing and speed perturbation to make the generated vehicles fall into the car-following stable attraction domain.

[0089] When the density rises to near the critical density, the congestion activation coefficient in the second module or lateral propagation speed The system determines that the increase is a synchronous flow; the fourth module increases the posterior probability of the conservative driver state based on the local density, and injects the corresponding safe headway, comfort deceleration and reaction time into the car-following model to suppress the amplification of non-physical disturbances.

[0090] When congestion occurs, the second module receives... and If the threshold is exceeded, the system determines it as congestion; the sixth module extracts traffic wave response energy and phase during micro-trajectory aggregation; the seventh module calculates the wave number range of ghost convoys. .when When the value is below the threshold, the seventh module... The basic map parameters related to this band are oriented for calibration. The calibrated parameters are written back to the statistical feature cache and used by the second and third modules in the next conversion cycle.

[0091] When the detection data of a certain coil or video is inconsistent with the continuity equation, the eighth module calculates its continuity residual. ;like Exceeding the threshold Then the sensor's observation noise covariance is updated to... This reduces the fusion weight. During the congestion dissipation phase, the fifth module advances the candidate trajectory along the effective traffic wave propagation direction and performs weighted and partitioned resampling based on spatial deviation, velocity deviation, wave energy deviation, and continuity residual. Finally, it merges with the measured vehicle trajectory to output a complete microscopic traffic flow field.

[0092] Compared to the "random point distribution + mean aggregation + ordinary filtering" approach, this invention satisfies car-following balance and stable mode constraints during vehicle generation, preserves second-order traffic wave fluctuation information during the micro-to-macro transition, utilizes ghost convoy bands as sensitive probes during consistency verification, and enforces continuity equations during multi-source fusion. Therefore, this invention reduces false speed waves and false queuing, improves consistency between macro and micro transitions, data fusion accuracy, and trajectory completion quality, and controls the time for a single transition to the tens of milliseconds, meeting real-time simulation requirements.

[0093] The advantages of the method for individualized generation and trajectory fusion of ghost convoy vehicles in macro-micro transformation described in this embodiment of the invention are as follows: (1) By generating vehicles through inverse mapping of equilibrium state, the initial spacing and speed of vehicles meet the equilibrium conditions of the car-following model, and the macroscopic state deviation is limited to the stable mode subspace, thereby reducing the ghost car effect caused by random point scattering from the source. (2) By effectively propagating the mode, the free flow, synchronous flow and congestion states are distinguished, and the reverse flow feedback mode is not mistakenly activated in the free flow state, so that the traffic state judgment is consistent with the subsequent vehicle generation, driver parameter switching and trajectory completion. (3) By using traffic wave response kernel aggregation, when converting microscopic trajectories into macroscopic field quantities, first-order field quantities such as density, velocity, and flow rate, as well as second-order fluctuation characteristics such as velocity variance, response energy, and phase, are retained at the same time, reducing the information loss caused by only performing mean aggregation. (4) The fluctuations of the ghost fleet are used as a macro-micro consistency verification probe. The frequency domain correlation is calculated in the wavenumber range corresponding to the characteristic wavelength of 300 to 2000 meters. Based on this, the basic graph parameters such as free flow velocity, critical density and parking wave velocity are calibrated in a directional manner to form a closed loop during operation. (5) By calibrating the hidden Markov driver, the hidden state of the driver is inferred based on the observation of density, speed, spacing and acceleration, and the car-following parameters corresponding to the conservative, neutral and aggressive types are injected into the micro model to suppress the amplification of non-physical disturbances from the perspective of behavioral heterogeneity. (6) By constraining the state estimation through the continuity equation, the fusion results of multi-source data such as coils, video, detection vehicles and simulation trajectories satisfy traffic flow conservation, and the weight of abnormal sensors is automatically reduced by utilizing the continuity residual; (7) The trajectory is generated by guiding the traffic wave characteristic line, so that the completed trajectory advances along the effective traffic wave propagation direction, and the sampling is weighted according to the spatial deviation, velocity deviation, wave energy deviation and continuity residual, thereby improving the fit between the virtual trajectory and the measured trajectory; (8) Through continuous field quantity caching, discrete individual quantity caching and statistical feature caching, state transfer between modules is realized, so that the three types of engines, namely macroscopic allocation, mesoscopic cell and microscopic vehicle, form a two-way conversion mechanism that is traceable, verifiable and can be updated in a closed loop at the unified boundary.

[0094] like Figure 3 As shown, to achieve the above objectives, embodiments of the present invention provide a device for individualized generation and trajectory fusion of ghost convoy vehicles in macro-micro transformation, comprising: The first acquisition module 301 is used to acquire the traffic state label of the target traffic flow based on the continuous field quantity cache of the target traffic flow within the target trajectory generation cycle of multiple trajectory generation cycles; wherein, the continuous field quantity cache includes the physical data of the target traffic flow at the cellular scale. The second acquisition module 302 is used to generate micro-vehicles based on the traffic state labels and acquire the multi-dimensional state of the micro-vehicles. The third acquisition module 303 is used to fuse the target trajectory with the highest candidate weight among multiple candidate trajectories generated in the first region with the pre-acquired measured trajectory to obtain a completed trajectory; wherein, the first region is a region where the trajectory is discontinuous among multiple regions included in the target traffic flow; the multiple candidate trajectories are acquired according to the traffic wave propagation direction of the target traffic flow; the candidate weight is determined according to the comprehensive residual of the candidate trajectories; The first processing module 304 is used to store the completed trajectory and the micro vehicle trajectory together as the vehicle trajectory generation result of the target trajectory generation cycle into a discrete individual quantity cache; wherein, the micro vehicle trajectory is generated based on the multidimensional state.

[0095] Optionally, the device further includes: The fourth acquisition module is used to read the vehicle trajectory generation result from the discrete individual quantity cache and obtain the traffic wave frequency domain features based on the vehicle trajectory generation result; The second processing module is used to read the macroscopic velocity field of the target traffic flow from the continuous field quantity buffer; The fifth acquisition module is used to acquire the ghost vehicle band consistency index by performing two-dimensional spatiotemporal Fourier transforms on the traffic wave frequency domain features and the macroscopic velocity field respectively; wherein, the ghost vehicle band consistency index is used to represent the degree of matching between macroscopic basic map parameters and microscopic fluctuations; the macroscopic basic map parameters are the feature parameters of the macroscopic basic map corresponding to the target traffic flow; the microscopic fluctuations are the vehicle speed fluctuation statistics of the microscopic vehicles; The first generation module is used to generate target macroscopic basic map parameters based on the ghost vehicle band consistency index and historical macroscopic basic map parameters when the ghost vehicle band consistency index is less than a first preset threshold; wherein, the historical macroscopic basic map parameters are macroscopic basic map parameters obtained in a first trajectory generation cycle; the first trajectory generation cycle is a trajectory generation cycle that precedes the target trajectory generation cycle among the plurality of trajectory generation cycles; the target macroscopic basic map parameters are used to describe the macroscopic state of traffic flow in a second trajectory generation cycle; the second trajectory generation cycle is a trajectory generation cycle that follows the target trajectory generation cycle among the plurality of trajectory generation cycles.

[0096] Optionally, in the aforementioned apparatus, the fifth acquisition module includes: The first acquisition unit is used to perform a two-dimensional spatiotemporal Fourier transform on the frequency domain features of the traffic wave to acquire the transformed traffic wave frequency domain features; and to perform a two-dimensional spatiotemporal Fourier transform on the macroscopic velocity field to acquire the transformed macroscopic velocity field. The second acquisition unit is used to acquire macro-micro frequency domain correlation based on the frequency domain characteristics of the transformed traffic wave and the transformed macro velocity field; wherein, the macro-micro frequency domain correlation is used to characterize the frequency domain correlation between the macro state corresponding to the target traffic flow and the micro state corresponding to the micro vehicle; The third acquisition unit is used to integrate the macro-micro frequency domain correlation within the characteristic beam interval of the ghost fleet to obtain the band consistency index of the ghost fleet.

[0097] Optionally, the device further includes: The sixth acquisition module is used to acquire the traffic state to be estimated based on the continuity equation of the target traffic flow; The seventh acquisition module is used to take the traffic state to be estimated as a constraint and acquire the fused traffic state and vehicle state based on historical traffic state and multi-source observation data. The eighth acquisition module is used to reduce the sensor weight corresponding to the target sensor when the target continuity residual generated by the target sensor in the fused traffic state from multiple sensors corresponding to the multi-source observation data is greater than a second preset threshold, thereby acquiring the final fused traffic state; wherein, the sensor weight is used to characterize the influence of the observation data acquired by the sensor on the final fused traffic state; The third processing module is used to write the final fused traffic state into the continuous field quantity cache, and to write the vehicle state into the discrete individual quantity cache.

[0098] Optionally, in the aforementioned apparatus, the first acquisition module 301 includes: The fourth acquisition unit is used to acquire the downstream propagation speed, upstream propagation speed and cross-flow propagation speed of the target traffic flow according to the continuous field quantity buffer. The first processing unit is used to determine the traffic state of the target traffic flow in different cells based on the downstream propagation speed, the upstream propagation speed, and the cross-current propagation speed. The fifth acquisition unit is used to acquire the traffic status tag based on the traffic status.

[0099] Optionally, in the aforementioned apparatus, the second acquisition module 302 includes: The second processing unit is used to calculate the number of vehicles to be generated within the target trajectory generation cycle at the transition boundary based on the traffic state label and the continuous field quantity cache; wherein, the transition boundary is the boundary between the macro state corresponding to the target traffic flow and the micro state corresponding to the micro vehicle; The first generation unit is used to generate the micro-vehicles based on the equilibrium relationship of the car-following model and according to the number of vehicles to be generated, and to obtain the initial state of the micro-vehicles. The third processing unit is used to identify the vehicle driving behavior type corresponding to the micro vehicle based on the observation information of the micro vehicle; wherein, different vehicle driving behavior types correspond to different driver parameters. The sixth acquisition unit is used to adjust the initial state of the vehicle according to the vehicle driving behavior type and acquire the multi-dimensional state.

[0100] Optionally, in the aforementioned apparatus, the third acquisition module 303 includes: The second generation unit is used to advance within the first region according to the traffic wave propagation direction of the traffic state corresponding to the sub-region, and generate the multiple candidate trajectories; wherein the traffic state is determined according to the traffic state label; the first region is divided into multiple sub-regions according to the traffic state label; The first determining unit is used to determine the candidate weight of each candidate trajectory among the plurality of candidate trajectories based on the comprehensive residual. The second determining unit is used to resample within the sub-region corresponding to each traffic state label in different traffic state labels, and determine the candidate trajectory with the highest candidate weight as the target trajectory. The seventh acquisition unit is used to fuse the target trajectory with the measured trajectory to obtain the completed trajectory.

[0101] It should be noted that the apparatus provided in this embodiment of the invention can implement all the method steps implemented in the above method embodiment and can achieve the same technical effect. Therefore, the parts and beneficial effects that are the same as those in the method embodiment will not be described in detail here.

[0102] To achieve the above objectives, embodiments of the present invention provide a device for individualized generation and trajectory fusion of ghost convoy vehicles in macro-micro transformation, comprising: a processor, a memory, and a program or instructions stored in the memory and executable on the processor; wherein, when the processor executes the program or instructions, it implements the method for individualized generation and trajectory fusion of ghost convoy vehicles in macro-micro transformation as described above.

[0103] To achieve the above objectives, embodiments of the present invention provide a readable storage medium storing a program or instructions thereon, wherein the program or instructions, when executed by a processor, implement the steps in the method for individualizing and fusion trajectories of ghost convoy vehicles in macro-micro transformation as described above.

[0104] To achieve the above objectives, embodiments of the present invention provide a computer program product, which includes computer instructions that, when executed by a processor, implement the steps of the method for individualizing and fusion trajectories of ghost convoy vehicles in macro-micro transformation as described above.

[0105] It should be further noted that the terminals described in this specification include, but are not limited to, smartphones, tablets, etc., and many of the functional components described are referred to as modules in order to emphasize the independence of their implementation.

[0106] In this embodiment of the invention, the module can be implemented in software so that it can be executed by various types of processors. For example, an identified executable code module may include one or more physical or logical blocks of computer instructions, which may be constructed as objects, procedures, or functions. Nevertheless, the executable code of the identified module does not need to be physically located together, but may include different instructions stored in different locations, which, when logically combined, constitute the module and achieve the module's intended purpose.

[0107] In practice, an executable code module can be a single instruction or many instructions, and can even be distributed across multiple different code segments, different programs, and across multiple memory devices. Similarly, operational data can be identified within the module and can be implemented in any suitable form and organized within any suitable data structure. This operational data can be collected as a single dataset or distributed across different locations (including different storage devices), and can exist, at least in part, solely as electronic signals within the system or network.

[0108] When a module can be implemented using software, considering the current level of hardware technology, modules that can be implemented in software can be implemented using hardware circuits by those skilled in the art to achieve the corresponding functions, without considering cost. These hardware circuits include conventional very-large-scale integrated circuits (VLSI) or gate arrays, as well as existing semiconductors such as logic chips and transistors, or other discrete components. Modules can also be implemented using programmable hardware devices, such as field-programmable gate arrays, programmable array logic, and programmable logic devices.

[0109] The exemplary embodiments described above are with reference to the accompanying drawings. Many different forms and embodiments are feasible without departing from the spirit and teachings of the invention. Therefore, the invention should not be construed as limiting the exemplary embodiments set forth herein. Rather, these exemplary embodiments are provided to make the invention complete and convey the scope of the invention to those skilled in the art. In these drawings, component dimensions and relative dimensions may be exaggerated for clarity. The terminology used herein is for the purpose of describing particular exemplary embodiments only and is not intended to be limiting. As used herein, unless clearly indicated otherwise, the singular forms “a,” “an,” and “the” are intended to include all such forms. It will be further understood that the terms “comprising” and / or “including”, when used in this specification, indicate the presence of the stated features, integers, steps, operations, components, and / or elements, but do not exclude the presence or addition of one or more other features, integers, steps, operations, components, and / or groups thereof. Unless otherwise indicated, when stated, a range of values ​​includes the upper and lower limits of the range and any subranges in between.

[0110] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for individualized generation and trajectory fusion of ghost convoy vehicles in macro-micro transformation, characterized in that, include: Within the target trajectory generation cycle of multiple trajectory generation cycles, the traffic state label of the target traffic flow is obtained based on the continuous field quantity cache of the target traffic flow; wherein, the continuous field quantity cache includes the physical data of the target traffic flow at the cellular scale; Microscopic vehicles are generated based on the traffic state labels, and the multidimensional state of the microscopic vehicles is obtained. The target trajectory with the highest candidate weight among multiple candidate trajectories generated in the first region is fused with the pre-acquired measured trajectory to obtain the completed trajectory; wherein, the first region is the region where the trajectory is discontinuous among multiple regions included in the target traffic flow; the multiple candidate trajectories are obtained by advancing according to the traffic wave propagation direction of the target traffic flow; the candidate weight is determined according to the comprehensive residual of the candidate trajectories; The completed trajectory and the micro vehicle trajectory are used together as the vehicle trajectory generation result of the target trajectory generation cycle and stored in the discrete individual quantity cache; wherein, the micro vehicle trajectory is generated based on the multidimensional state.

2. The method according to claim 1, characterized in that, The method further includes: Read the vehicle trajectory generation result from the discrete individual quantity cache, and obtain the traffic wave frequency domain features based on the vehicle trajectory generation result; The macroscopic velocity field of the target traffic flow is read from the continuous field quantity buffer; By performing two-dimensional spatiotemporal Fourier transforms on the frequency domain features of the traffic waves and the macroscopic velocity field, a ghost vehicle band consistency index is obtained. This ghost vehicle band consistency index represents the degree of matching between macroscopic basic map parameters and microscopic fluctuations. The macroscopic basic map parameters are the feature parameters of the macroscopic basic map corresponding to the target traffic flow. The microscopic fluctuations are the vehicle speed fluctuation statistics of the microscopic vehicles. If the band consistency index of the ghost vehicle fleet is less than a first preset threshold, target macroscopic basic map parameters are generated based on the band consistency index of the ghost vehicle fleet and historical macroscopic basic map parameters; wherein, the historical macroscopic basic map parameters are macroscopic basic map parameters obtained in a first trajectory generation cycle; the first trajectory generation cycle is a trajectory generation cycle that precedes the target trajectory generation cycle among the plurality of trajectory generation cycles; the target macroscopic basic map parameters are used to describe the macroscopic state of traffic flow in a second trajectory generation cycle; the second trajectory generation cycle is a trajectory generation cycle that follows the target trajectory generation cycle among the plurality of trajectory generation cycles.

3. The method according to claim 2, characterized in that, By performing two-dimensional spatiotemporal Fourier transforms on the traffic wave frequency domain characteristics and the macroscopic velocity field, respectively, the band consistency index of the ghost convoy is obtained, including: A two-dimensional spatiotemporal Fourier transform is performed on the frequency domain features of the traffic wave to obtain the transformed traffic wave frequency domain features; and a two-dimensional spatiotemporal Fourier transform is performed on the macroscopic velocity field to obtain the transformed macroscopic velocity field. Based on the frequency domain characteristics of the transformed traffic wave and the transformed macroscopic velocity field, the macro-micro frequency domain correlation is obtained; wherein, the macro-micro frequency domain correlation is used to characterize the frequency domain correlation between the macroscopic state corresponding to the target traffic flow and the microscopic state corresponding to the microscopic vehicle; The macro- and micro-frequency domain correlation is integrated within the characteristic beam interval of the ghost fleet to obtain the band consistency index of the ghost fleet.

4. The method according to claim 1, characterized in that, The method further includes: Based on the continuity equation of the target traffic flow, the traffic state to be estimated is obtained; Using the traffic state to be estimated as a constraint, the fused traffic state and vehicle state are obtained based on historical traffic state and multi-source observation data. If the target continuity residual generated by the target sensor among the multiple sensors corresponding to the multi-source observation data and the fused traffic state is greater than a second preset threshold, the sensor weight corresponding to the target sensor is reduced to obtain the final fused traffic state; wherein, the sensor weight is used to characterize the influence of the observation data obtained by the sensor on the final fused traffic state; The final fused traffic state is written into the continuous field quantity cache, and the vehicle state is written into the discrete individual quantity cache.

5. The method according to claim 1, characterized in that, Within the target trajectory generation cycle of multiple trajectory generation cycles, the traffic state label of the target traffic flow is obtained based on the continuous field volume cache of the target traffic flow, including: The downstream propagation speed, upstream propagation speed, and cross-flow propagation speed of the target traffic flow are obtained based on the continuous field quantity buffer. Based on the downstream propagation speed, the upstream propagation speed, and the cross-flow propagation speed, the traffic state of the target traffic flow corresponding to different cells is determined; The traffic status label is obtained based on the traffic status.

6. The method according to claim 1, characterized in that, Based on the traffic state labels, micro-vehicles are generated, and the multi-dimensional states of the micro-vehicles are obtained, including: Based on the traffic state label and the continuous field quantity cache, the number of vehicles to be generated within the target trajectory generation cycle is calculated at the transition boundary; wherein, the transition boundary is the boundary between the macro state corresponding to the target traffic flow and the micro state corresponding to the micro vehicle; Based on the equilibrium relationship of the car-following model, the micro-vehicles are generated according to the number of vehicles to be generated, and the initial state of the micro-vehicles is obtained. Based on the observation information of the micro-vehicle, the vehicle driving behavior type corresponding to the micro-vehicle is identified; wherein, different vehicle driving behavior types correspond to different driver parameters; Adjust the initial state of the vehicle according to the type of vehicle driving behavior to obtain the multi-dimensional state.

7. The method according to claim 1, characterized in that, The target trajectory with the highest candidate weight among multiple candidate trajectories generated in the first region is fused with the pre-acquired measured trajectory to obtain the completed trajectory, including: In the first region, the multiple candidate trajectories are generated by advancing according to the traffic wave propagation direction under the traffic conditions corresponding to the sub-regions; wherein, the traffic conditions are determined according to the traffic condition labels; the first region is divided into multiple sub-regions according to the traffic condition labels; Based on the comprehensive residual, determine the candidate weight of each candidate trajectory among the plurality of candidate trajectories; Resampling is performed in the sub-region corresponding to each of the different traffic status labels to determine the candidate trajectory with the highest candidate weight as the target trajectory. The target trajectory is fused with the measured trajectory to obtain the completed trajectory.

8. A device for individualized generation and trajectory fusion of ghost convoy vehicles in macro-micro transformation, characterized in that, include: The first acquisition module is used to acquire the traffic state label of the target traffic flow based on the continuous field quantity cache of the target traffic flow within the target trajectory generation cycle of multiple trajectory generation cycles; wherein, the continuous field quantity cache includes the physical data of the target traffic flow at the cellular scale. The second acquisition module is used to generate micro-vehicles based on the traffic state labels and acquire the multi-dimensional state of the micro-vehicles. The third acquisition module is used to fuse the target trajectory with the highest candidate weight among multiple candidate trajectories generated in the first region with the pre-acquired measured trajectory to obtain the completed trajectory; wherein, the first region is the region where the trajectory is discontinuous among multiple regions included in the target traffic flow; the multiple candidate trajectories are acquired according to the traffic wave propagation direction of the target traffic flow; the candidate weight is determined according to the comprehensive residual of the candidate trajectories; The first processing module is used to combine the completed trajectory and the micro vehicle trajectory as the vehicle trajectory generation result of the target trajectory generation cycle and store them in a discrete individual quantity cache; wherein, the micro vehicle trajectory is generated based on the multidimensional state.

9. A device for individualized generation and trajectory fusion of ghost convoy vehicles in macro-micro transformation, comprising: A processor, a memory, and a program or instructions stored in the memory and executable on the processor; characterized in that, when the processor executes the program or instructions, it implements the method for individualized generation and trajectory fusion of ghost convoy vehicles in macro-micro transformation as described in any one of claims 1-7.

10. A readable storage medium having a program or instructions stored thereon, characterized in that, When the program or instructions are executed by the processor, they implement the steps in the method for individualizing and fusion of ghost fleet vehicles in macro-micro transformation as described in any one of claims 1-7.

11. A computer program product, characterized in that, The method includes computer instructions that, when executed by a processor, implement the steps of the method for individualizing and fusion trajectories of ghost convoy vehicles in macro-micro transformation as described in any one of claims 1-7.

Images