Macroscopic dynamic traffic flow generation and evolution method and system for mixed traffic flow

Abstract

The application discloses a macro dynamic traffic flow generation and evolution method and system for mixed traffic flow, and the method comprises the following steps: acquiring traffic flow basic data containing average speed, vehicle density, vehicle type ratio and automatic driving penetration rate; introducing a heterogeneous disturbance factor dynamically adjusted according to a heavy truck ratio in a continuity equation of a classical LWR model to obtain an improved traffic flow model to simulate fluctuation effects caused by mixed flow; calculating macro traffic flow by using the improved model; generating a virtual vehicle injection sequence based on a non-homogeneous Poisson process; detecting downstream density and speed changes in real time, dynamically correcting upstream injection intensity when approaching saturation, and suppressing reverse shock waves; deducing a start-end point matrix according to intersection flow by using a maximum entropy model; and finally outputting a virtual vehicle sequence with position, speed, vehicle type and target path, and injecting the virtual vehicle sequence into an automatic driving simulation platform. The application can accurately reflect congestion wave evolution characteristics under high heavy truck ratio mixed flow and improve test fidelity.

Description

Technical Field

[0001] This invention relates to the field of traffic simulation technology, and in particular to a method and system for generating and evolving macroscopic dynamic traffic flow for mixed traffic flow, which is applicable to virtual simulation testing of autonomous driving. Background Technology

[0002] In autonomous driving simulation testing, typical scenarios such as ports, logistics parks, and urban expressways are characterized by complex vehicle types, particularly a high proportion of heavy trucks and dynamic changes in autonomous driving penetration. Due to their physical characteristics such as long bodies, delayed start-up response, and extended braking distance, heavy trucks generate shockwave effects in mixed traffic flow that differ from those of passenger cars. Simultaneously, the time-varying nature of autonomous driving penetration further exacerbates the heterogeneity and dynamism of traffic behavior patterns. This mixed traffic environment places extremely high demands on the realism of macroscopic traffic flow models.

[0003] Currently widely used macroscopic traffic flow models are mostly based on the LWR (Lighthill-Whitham-Richards) continuous medium model. The classic LWR model, based on the assumption of homogeneous fluid, cannot characterize the impact of vehicle type composition differences on traffic wave propagation patterns. Some LWR models improved for intelligent connected environments differentiate between autonomous vehicles and traditional vehicles by introducing car-following behavior parameters or heterogeneous congestion densities for different vehicle types. However, these improvements still focus on differences in driving behavior and do not fully consider the fundamental impact of the physical properties of heavy trucks on macroscopic fluctuations. Furthermore, existing model structures generally lack fluid characteristic correction mechanisms that dynamically adjust with vehicle type proportions, making it difficult to accurately reproduce the generation, propagation, and dissipation of congestion waves in mixed traffic environments with a high proportion of heavy trucks and time-varying autonomous driving penetration rates.

[0004] Furthermore, existing simulation systems often use equal-interval or simple random methods to inject virtual vehicles when generating background traffic flow, without dynamically adjusting based on the real-time carrying capacity of downstream road segments, thus failing to simulate the real evolution of congestion spreading upstream. The origin-destination path allocation also largely relies on static presets, failing to maintain statistical consistency with macroscopically observed traffic flow, further reducing the realism and usability of mixed traffic flow simulation. Summary of the Invention

[0005] The technical problem this application aims to solve is: addressing the issues that existing macro-traffic flow models cannot accurately reflect the heterogeneous traffic characteristics caused by the high proportion of heavy trucks and the penetration rate of time-varying autonomous driving in mixed traffic flow, as well as the lack of dynamic feedback in virtual vehicle injection and the disconnect between path distribution and macro-data. This application provides a dynamic traffic flow generation and evolution method that can realistically simulate the shock wave effect and dynamic congestion spread process of mixed traffic flow, and provides statistically consistent path distribution.

[0006] To solve the above-mentioned technical problems, the technical solution adopted in this application is as follows: A macroscopic dynamic traffic flow generation and evolution method for mixed traffic flows includes: Acquire basic traffic flow data of the target road network, including average speed, vehicle density, vehicle type ratio, and autonomous driving penetration rate, wherein the vehicle type ratio includes at least the proportion of heavy trucks µ. Based on the acquired traffic flow data, a heterogeneous disturbance factor is introduced into the continuity equation of the classic LWR model. An improved traffic flow model is obtained, wherein the heterogeneous disturbance factor The ratio µ of heavy trucks is dynamically adjusted to simulate the traffic flow fluctuation effect caused by heavy trucks. Macroscopic traffic flow is calculated using the improved traffic flow model. Based on the macroscopic traffic flow, a non-homogeneous Poisson process is used to generate the probability distribution of vehicle arrivals and to produce a virtual vehicle injection sequence. Real-time detection of vehicle density and speed in downstream road sections. When the vehicle density approaches saturation density and the downstream speed reduction rate meets preset conditions, a queuing flow limiting operator is introduced at the upstream control node to dynamically correct the intensity parameters of the non-homogeneous Poisson process, so as to build a deceleration buffer in advance at the upstream boundary and suppress the reverse shock wave. Using the maximum entropy model, the origin-destination (OD) matrix of the entire network's vehicle travel paths is automatically derived based on the observed intersection traffic flow. The traffic flow evolution results are output, which include at least the virtual vehicle injection sequence, wherein the virtual vehicle injection sequence contains the position, initial velocity, vehicle type, and target path determined from the OD matrix of each virtual vehicle, and is used to inject into the autonomous driving simulation platform.

[0007] In one embodiment, the heterogeneous perturbation factor Constructed through a nonlinear composite mapping, its functional form is: ; Where α is the weighting coefficient for the mixed-flow heterogeneity adjustment. This is the damping attenuation factor for heavy-duty trucks. The value of γ is the autonomous driving penetration compensation index, and it monotonically increases with the autonomous driving penetration rate. This nonlinear composite mapping is used to characterize the situation when the proportion of heavy trucks... The nonlinear accelerated decay characteristics of the overall traffic flow capacity when the elevation increases.

[0008] In one embodiment, the formula for calculating the macroscopic traffic flow q in the improved traffic flow model is: ; Where k(x,t) represents the vehicle density of a specific road segment at time t. The free-flow velocity of the road segment. Blocking density; It is a heterogeneous perturbation factor.

[0009] In one embodiment, the steps of real-time detection of vehicle density and speed in downstream road sections, and when the vehicle density approaches saturation density and the downstream speed reduction rate meets a preset condition, introducing a queuing and flow-limiting operator at the upstream control node to dynamically correct the intensity parameters of the non-homogeneous Poisson process specifically include: Real-time monitoring of the carrying capacity of downstream road sections; if the current density... Satisfy the judgment conditions And the rate of decline in downstream speed At that time, reverse shock wave suppression is activated, and the queuing current limiting operator R(t) is: ; And the intensity parameters of the upstream non-homogeneous Poisson process. Revised to: ; Where ω is the reverse shock wave suppression attenuation coefficient; θ is a preset velocity descent rate threshold; For blocking density.

[0010] In one embodiment, the acquisition of basic traffic flow data of the target road network further includes: acquiring at least one spatiotemporal constraint from the traffic light timing schemes at each intersection, road speed limit gradients, and tidal lane status. The acquisition of basic traffic flow data for the target road network also includes: acquiring real-time congestion index and / or weather impact factors as external triggers to adjust traffic flow parameters.

[0011] In one embodiment, the step of acquiring basic traffic flow data of the target road network further includes spatiotemporal alignment of multi-source heterogeneous data, specifically including: Timestamp synchronization: The IEEE 1588 PTP precise time protocol is used to align the time of the intersection checkpoint data with the test vehicle's sensing data, and interpolation compensation is used to eliminate transmission delay errors; Spatial coordinate unification: For the physical coordinates of the checkpoint and the coordinates of the perceived target in the vehicle body coordinate system of the test vehicle, a multi-parameter spatial affine transformation model is used to project both onto the global grid coordinate system of the high-precision map through transformation matrix and translation vector.

[0012] In one embodiment, the output traffic flow evolution results further include: a road network heat map, key performance indicators, and anomaly scenario reports; The road network heat map represents the dynamic saturation distribution of all road segments; the key performance indicators include average queue length and intersection delay time; and the abnormal scenario report includes automatically identified congestion bottlenecks and their duration.

[0013] In one embodiment, the method is applied to the virtual simulation verification of an autonomous driving system in a mixed traffic environment with a high proportion of heavy trucks in a port or logistics park.

[0014] This invention also proposes a macroscopic dynamic traffic flow generation and evolution system for mixed traffic flows, used to implement the methods described above, including: The data acquisition module is used to acquire basic traffic flow data of the target road network. The basic traffic flow data includes average speed, vehicle density, vehicle type ratio, and autonomous driving penetration rate, wherein the vehicle type ratio includes at least the proportion of heavy trucks. The model building module is used to introduce heterogeneous disturbance factors into the continuity equation of the classic LWR model based on the acquired traffic flow basic data to obtain an improved traffic flow model. The heterogeneous disturbance factors are dynamically adjusted according to the proportion of heavy trucks in the vehicle type ratio to simulate the traffic flow fluctuation effect caused by heavy trucks. A macro-flow calculation module is used to calculate macro-flow using the improved traffic flow model; The vehicle injection generation module is used to generate a probability distribution of vehicle arrivals based on the macroscopic traffic flow using a non-homogeneous Poisson process, and to generate a virtual vehicle injection sequence. The shock wave suppression module is used to detect the vehicle density and speed of the downstream road section in real time. When the vehicle density is close to the saturation density and the downstream speed reduction rate meets the preset conditions, a queuing flow limiting operator is introduced at the upstream control node to dynamically correct the intensity parameters of the non-homogeneous Poisson process, so as to build a deceleration buffer in advance at the upstream boundary and suppress the reverse shock wave. The OD (Origin-Destination) extrapolation module is used to automatically extrapolate the origin-destination (OD) matrix of the entire network's vehicle travel paths based on the observed intersection traffic flow using the maximum entropy model. The output module is used to output traffic flow evolution results, which include at least the virtual vehicle injection sequence. The virtual vehicle injection sequence contains the position, initial velocity, vehicle type, and target path determined from the OD matrix of each virtual vehicle, and is used to inject the virtual vehicle into the autonomous driving simulation platform.

[0015] In one embodiment, the data acquisition module is further configured to acquire at least one spatiotemporal constraint from the traffic light timing scheme at each intersection, the road speed limit gradient, and the tidal lane status, as well as to acquire the real-time congestion index and / or weather influencing factors as external triggering sources.

[0016] Compared with existing technologies, this solution has the following advantages: By introducing a heterogeneous disturbance factor linked to heterogeneous parameters such as the proportion of heavy trucks into the classic LWR model, the model can sensitively reflect the nonlinear decay of traffic capacity caused by the increase in the proportion of heavy trucks, and accurately simulate the congestion wave and shock wave effects unique to mixed traffic flow. At the same time, through the combined application of non-homogeneous Poisson injection, active suppression of reverse shock waves, and maximum entropy OD derivation, a highly realistic microscopic vehicle sequence is generated under the drive of macroscopic dynamic traffic flow. This provides a heterogeneous traffic flow background environment with sudden disturbance characteristics for the autonomous driving simulation platform, significantly improving the realism and effectiveness of simulation testing. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the macroscopic dynamic traffic flow generation and evolution method for mixed traffic flow according to an embodiment of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the invention.

[0019] like Figure 1 As shown in the figure, the macroscopic dynamic traffic flow generation and evolution method for mixed traffic flow provided in this embodiment includes the following steps in its overall process: S1. Obtain basic traffic flow data.

[0020] The data obtained includes: ①Basic flow characteristics: average speed v, vehicle density k.

[0021] ②Specific parameters of the mixed flow: vehicle type ratio, from which the proportion of heavy trucks µ (0≤µ≤1) is extracted; autonomous driving penetration rate p AV (0≤) AV ≤1), representing the percentage of autonomous vehicles among all vehicles.

[0022] ③ Spatiotemporal constraints (optional): Traffic light timing schemes at each intersection, road speed limit gradients, and tidal lane status.

[0023] ④ External triggering sources (optional): Real-time congestion index and weather impact factor from Gaode / Baidu API (as a reduction factor).

[0024] In this step, in order to incorporate raw traffic data from different sources and with different sampling periods into a unified computational framework, spatiotemporal alignment of multi-source heterogeneous data is required: Timestamp synchronization adopts the IEEE 1588 PTP precise time protocol to align the second-level granular data of the checkpoint with the millisecond-level GNSS / INS clock source of the test vehicle, and eliminates transmission delay errors through interpolation compensation.

[0025] Unified spatial coordinates: Let the physical coordinates of the checkpoint be P. cam (Lat, Lon), the target perceived by the test vehicle is P in the vehicle coordinate system. veh (x,y), P cam By transforming conventional geodetic projection to a local planar coordinate system, for motion detection data, a multi-parameter spatial affine transformation model is used, introducing a transformation matrix M and a translation vector T, to uniformly project the vehicle coordinates to the global grid coordinate system P of the high-precision map. global : .

[0026] S2, Key Feature Extraction.

[0027] Extract or compute the following key features from spatiotemporally aligned data: Traffic flow density distribution k(x, t): The number of vehicles per unit length of a road segment at spatial coordinates x and time t, i.e., the vehicle density of a specific road segment at time t.

[0028] Heavy truck proportion µ: This represents the percentage of heavy trucks in all vehicles and is the most significant driver of subsequent heterogeneous disturbance factors. Autonomous driving penetration rate. AV This is used to calibrate the compensation index. .

[0029] Temporal occupancy rate ρ (optional): The proportion of time a lane is physically occupied by vehicles. This parameter can be used to verify density estimates and as an auxiliary indicator for saturation evaluation, but it is not a necessary calculation condition.

[0030] S3. Improve the LWR model and construct macroscopic traffic flow equations.

[0031] The core continuity equations of the classic LWR model are retained to ensure the conservation of vehicle numbers during simulation. Specifically, the continuity equations are utilized... ; (1) Where k=k(x,t) represents vehicle density, q=q(x,t) represents traffic flow, t is time, and x is the spatial coordinate along the road direction. This represents the rate of change of density at a certain location over time. This represents the gradient of traffic flow as a function of space. The equation as a whole expresses the conservation of vehicle numbers: on road sections where no vehicles are generated or disappear (no entrance or exit ramps), the increase in density at a certain point must be equal to the difference in traffic flow between the upstream and downstream points at that point. In other words, vehicles cannot appear or disappear out of thin air.

[0032] The improvement of the LWR model in this invention is that a heterogeneous disturbance factor Φ is introduced into the basic flow-density diagram to reflect the coupling effect of heavy trucks and autonomous vehicles in mixed traffic flow.

[0033] Φ is constructed through a nonlinear composite mapping, whose parameters include: a heterogeneous adjustment weighting coefficient for mixed flow, a heavy truck damping attenuation factor, and an autonomous driving penetration compensation index. This nonlinear mapping causes the overall traffic flow capacity to exhibit a nonlinear accelerated attenuation characteristic when the proportion of heavy trucks increases, thereby accurately reproducing the shock wave propagation law in actual mixed flow.

[0034] Specifically, the macroscopic traffic flow after introducing the heterogeneous disturbance factor Φ is: ; (2)

[0035] in, The free-flow velocity of the road segment can be calibrated by the input average velocity v under low-density conditions, or determined based on the road speed limit gradient. The congestion density, i.e., the maximum density when vehicles come to a complete stop, is a preset constant or set according to the number of lanes; The heterogeneous disturbance factor is a fluid reduction coefficient dynamically adjusted based on the heavy truck ratio µ. It simulates the shockwave effect on the traffic flow behind the trucks caused by the slow start and long braking of heavy trucks, and adjusts with the penetration rate of autonomous driving.

[0036] The specific construction formula is limited as follows: (3)

[0037] in, The weighting coefficients for heterogeneous mixing flow control the relative effects of the exponential decay term and the logarithmic compensation term. This is the damping attenuation factor for heavy trucks; the larger the value, the more significant the erosion of traffic capacity by the proportion of heavy trucks. This is the autonomous driving penetration compensation index. Its value varies with the autonomous driving penetration rate. AV Monotonically increasing; when AV At higher levels, Larger, making This reduces the nonlinear decay of traffic capacity, reflecting the compensatory effect of autonomous vehicles on the stability of mixed flow.

[0038] This formula accurately characterizes the nonlinear accelerated decay of overall traffic flow capacity as the proportion of heavy trucks (µ) increases through a nonlinear composite mapping. Specifically, this construction forms a mapping mechanism distinct from conventional linear reduction or single exponential correction through the nonlinear composite of exponential decay and logarithmic compensation terms. In the range where the proportion of heavy trucks (µ) is low, the capacity decay is relatively gradual; when µ exceeds a certain threshold, the compensation capacity of the logarithmic term tends to saturate, and the exponential decay term dominates, resulting in a nonlinear accelerated decay of capacity, accurately replicating the soft threshold effect of the proportion of heavy trucks increasing from low to high in real traffic. Simultaneously, the introduction of the autonomous driving penetration compensation index γ can dynamically adjust the inflection point of the decay curve according to the penetration level of autonomous vehicles, resulting in differentiated capacity impacts for the same proportion of heavy trucks under different autonomous driving penetration rates. This provides higher scenario adaptability and simulation realism compared to existing LWR heterogeneous flow improvement methods (such as introducing dissipative terms or simple speed reduction).

[0039] Using the improved traffic flow model described above, combined with real-time input data, the macroscopic traffic flow of each road segment is calculated.

[0040] Substituting equation (2) into equation (1), we obtain the result regarding density. ( , The one-dimensional nonlinear hyperbolic partial differential equation is combined with the initial density field. ( Given the conditions (0) (provided by real-time monitoring data) and boundary conditions (upstream input flow, downstream output capacity), the density can be obtained by solving the equation numerically. ( , The spatiotemporal evolution distribution of traffic flow is analyzed, and then the macroscopic traffic flow at any location and time is calculated according to equation (2). ( , ).

[0041] When considering spatiotemporal constraints, the boundary conditions or local conditions can be adjusted accordingly. free , jam Parameters such as traffic light timing schemes are transformed into periodic maximum throughput limits at intersections; tidal flow lane status is adjusted by changing the number of effective lanes on the road segment. free , jam External congestion index and weather reduction factor can be used to calculate... ( , )or Perform a second correction (such as multiplying by a reduction factor).

[0042] S4. Virtual vehicle injection based on non-homogeneous Poisson process.

[0043] Based on macroscopic traffic flow, a non-homogeneous Poisson process is used to calculate the probability distribution of vehicle arrivals, generating virtual vehicle arrival times and initial states at unequal intervals, forming a virtual vehicle injection sequence. Each virtual vehicle instruction includes information such as location, initial velocity, and vehicle type.

[0044] S5, Reverse shock wave suppression.

[0045] To prevent downstream congestion from spreading upstream and causing simulation lock-up, the density k and speed v of the downstream road segment are monitored in real time. This is achieved when the following density condition is simultaneously met: And the rate of decline in downstream speed At that time, reverse shock wave suppression is automatically activated. Among them, It is a preset speed decrease rate threshold, which can be calibrated according to the scenario.

[0046] A queuing and flow-limiting operator R(t) is introduced at the upstream control node. (4) Wherein, ω is the reverse shock wave suppression attenuation coefficient, which determines the sensitivity of the suppression intensity to changes in density deviation.

[0047] Intensity parameters of upstream nonhomogeneous Poisson processes Revised to: (5) Because R(t) ≤ 1, the downstream becomes more crowded ( The closer jam The smaller R(t) is, the lower the probability of upstream vehicle injection is, thus forming a deceleration buffer in advance upstream, actively reducing peaks and suppressing the backflow propagation of reverse shock waves, and avoiding global grid lock-up.

[0048] This reduces the injection probability of upstream vehicles, builds a deceleration buffer zone at the upstream boundary in advance, and actively reduces and suppresses the reverse shock wave propagating against the current at the macro level, thus avoiding irrational mesh lock-up in the simulation environment.

[0049] S6, OD matrix automated derivation.

[0050] Knowing only the injection flow rate of each road segment is insufficient to determine the vehicle's travel path. This invention utilizes a maximum entropy model to extrapolate the OD matrix based on observed intersection turning flows.

[0051] For a given control node or key intersection in the road network, the turning flow of each approach lane can be obtained through checkpoint and floating car data. Using this as a constraint, the maximum entropy model maximizes the information entropy of path flow while satisfying all observed flow constraints, thus reversing the most probable vehicle origin-destination (OD) distribution across the entire network. The resulting OD matrix provides a reasonable range of target paths for each type of virtual vehicle. This step generalizes traffic flow from point flow data to area path distribution, providing reasonable target paths for virtual vehicles.

[0052] The OD matrix derivation in this step is not an independent, one-off static planning process, but rather a real-time, iterative calculation embedded in the dynamic feedback loop of traffic flow evolution. Specifically: The system collects the latest observed turning traffic at each key intersection during each traffic update cycle (synchronized with or multiple of the time step of macro traffic calculation); Using this real-time traffic as a constraint, the entire network OD distribution is solved in real time through the maximum entropy model; The updated OD matrix immediately acts in reverse on the virtual vehicle generation module, providing target paths for vehicles injected in the next time period that match the current macroscopic traffic conditions.

[0053] This closed-loop dynamic mechanism, which involves observing traffic flow, OD inversion, path assignment, vehicle injection, and traffic flow updates to new observed traffic flow, enables virtual vehicles to continuously follow the evolution of real traffic conditions and automatically adapt to the detour effect and traffic redistribution in congested areas. Unlike conventional simulation methods that use OD as a fixed input, this significantly improves the self-consistency and real-time performance of macro-micro coupled simulation of mixed flow.

[0054] S7, Output traffic flow evolution results.

[0055] Finally, the following results are output to support the autonomous driving simulation verification: Road network heat map: Displays the dynamic saturation (V / C ratio) distribution of road segments across the entire area. Here, V (Volume) represents the actual traffic volume q, and C (Capacity) represents the saturation level of the road segment under consideration. Effective traffic capacity after attenuation. The heatmap is provided in the form of a spatiotemporal matrix.

[0056] Virtual vehicle generation sequence: includes the initial position, initial velocity, vehicle type (heavy truck / passenger car / AV), target path (assigned by OD matrix) of each virtual vehicle, etc., as the real-time vehicle injection command of the simulation platform.

[0057] Key performance indicators include average queue length (based on density). The length integral of the region exceeding a certain threshold), and the intersection delay time (obtained by comparing the free-flow passage time with the actual passage time).

[0058] Abnormal Scenario Report: Automatically identifies and outputs road segments that may cause congestion bottlenecks, along with the start time and estimated duration of congestion, facilitating focused attention during simulation testing.

[0059] Ultimately, the generated macroscopic dynamic traffic flow evolution results are seamlessly injected into the virtual digital space of the autonomous driving simulation platform, providing the tested autonomous vehicles with a heterogeneous traffic flow background environment that possesses high fidelity and sudden disturbance characteristics. This invention is specifically designed to meet the stringent accuracy requirements of autonomous driving simulation testing for the injection of microscopic vehicles and macroscopic spatiotemporal alignment, thereby strengthening the essential distinction from conventional LWR models used in general traffic engineering (such as urban traffic light timing optimization and road planning). This invention is particularly suitable for virtual simulation verification of autonomous driving systems in mixed traffic environments with a high proportion of heavy trucks, such as ports and logistics parks.

[0060] The present invention also provides a macroscopic dynamic traffic flow generation and evolution system for mixed traffic flow, which includes functional modules for implementing the above steps, such as a data acquisition module, a model building module, a macroscopic flow calculation module, a vehicle injection generation module, a shock wave suppression module, an OD calculation module, and an output module.

[0061] The data acquisition module is used to acquire basic traffic flow data of the target road network. The basic traffic flow data includes average speed, vehicle density, vehicle type ratio, and autonomous driving penetration rate. The vehicle type ratio includes at least the ratio of heavy trucks to passenger cars.

[0062] The model building module is used to introduce heterogeneous disturbance factors into the continuity equation of the classic LWR model based on the acquired traffic flow basic data to obtain an improved traffic flow model. The heterogeneous disturbance factors are dynamically adjusted according to the proportion of heavy trucks in the vehicle type ratio to simulate the traffic flow fluctuation effect caused by heavy trucks.

[0063] The macro-flow calculation module is used to calculate macro-flow using the improved traffic flow model.

[0064] The vehicle injection generation module is used to generate the probability distribution of vehicle arrivals using a non-homogeneous Poisson process based on the macroscopic traffic flow, and to generate a virtual vehicle injection sequence.

[0065] The shock wave suppression module is used to detect the vehicle density and speed of the downstream road section in real time. When the vehicle density is close to the saturation density and the downstream speed reduction rate meets the preset conditions, a queuing flow limiting operator is introduced at the upstream control node to dynamically correct the intensity parameters of the non-homogeneous Poisson process, so as to build a deceleration buffer in advance at the upstream boundary and suppress the reverse shock wave.

[0066] The OD (Origin-Destination) derivation module is used to automatically deduce the origin-end point OD matrix of the entire network's vehicle travel paths based on the observed intersection traffic flow using the maximum entropy model.

[0067] The output module is used to output traffic flow evolution results, which include at least the virtual vehicle injection sequence. The virtual vehicle injection sequence contains the position, initial velocity, vehicle type, and target path determined from the OD matrix of each virtual vehicle, and is used to inject the virtual vehicle into the autonomous driving simulation platform.

[0068] Furthermore, the data acquisition module is also used to acquire at least one spatiotemporal constraint from the traffic light timing schemes at each intersection, road speed limit gradients, and tidal lane status, as well as to acquire real-time congestion indices and / or weather influencing factors from Gaode / Baidu APIs as external trigger sources.

[0069] Furthermore, the model building module also calculates the vehicle composition ratio coefficient µ based on the vehicle type ratio and autonomous driving penetration rate, and calculates the traffic density distribution and spatiotemporal occupancy rate based on the vehicle density. The vehicle composition ratio coefficient µ serves as the input parameter of the heterogeneous disturbance factor. The heterogeneous disturbance factor is constructed through a nonlinear composite mapping. The parameters of the nonlinear composite mapping include the mixed flow heterogeneous adjustment weight coefficient, the heavy truck damping attenuation factor, and the autonomous driving penetration compensation index.

[0070] Furthermore, the output module is also used to output road network heatmaps, key performance indicators, and abnormal scenario reports. The road network heatmap represents the dynamic saturation distribution of all road segments; key performance indicators include average queue length and intersection delay time; and abnormal scenario reports include automatically identified congestion bottlenecks and their duration.

[0071] The present invention has the following beneficial effects: By introducing a heterogeneous disturbance factor linked to heterogeneous parameters such as the proportion of heavy trucks into the classic LWR model, the model can sensitively reflect the nonlinear decay of traffic capacity caused by the increase in the proportion of heavy trucks, and accurately simulate the congestion wave and shock wave effects unique to mixed traffic flow. At the same time, through the combined application of non-homogeneous Poisson injection, active suppression of reverse shock waves, and maximum entropy OD derivation, a highly realistic microscopic vehicle sequence is generated under the drive of macroscopic dynamic traffic flow. This provides a heterogeneous traffic flow background environment with sudden disturbance characteristics for the autonomous driving simulation platform, significantly improving the realism and effectiveness of simulation testing.

[0072] In particular, the application of this invention in mixed traffic environments with a high proportion of heavy trucks, such as ports and logistics parks, has resulted in significant synergistic effects between the solution and the technical scenario: In port scenarios, the proportion of heavy trucks µ is usually very high. At this time, the nonlinear accelerated decay characteristic of Φ(µ) is fully excited, the exponential decay term drastically reduces the passage capacity, and the logarithmic compensation term gradually saturates, which can accurately reproduce the heavy truck towing queues and the strong shock wave propagation caused by them that are unique to this type of area. Meanwhile, the vehicle queuing density within the port's closed or semi-closed road network is extremely high, approaching the congestion density k. jam This situation occurs frequently. The reverse shock wave suppression algorithm of this invention is based on the real-time density k and the blocking density k jam The deviation is dynamically adjusted to reduce the upstream injection intensity, which effectively avoids the lock-up problem of traditional simulation under such extreme density and ensures the stability of continuous long-term simulation of port scenarios. The value of the autonomous driving penetration compensation index γ can be precisely linked to the actual penetration rate of autonomous vehicles (such as unmanned trucks) in ports, so that the same method can be effectively adapted to ports with different levels of automation.

[0073] The aforementioned synergistic effect indicates that the specific application scenario of ports / logistics parks is not a ubiquitous choice, but rather it is interdependent with and works together with the technical means of this invention (nonlinear Φ, active shock wave suppression) to improve simulation performance.

[0074] The embodiments described above are merely further illustrations of the present invention and are not intended to limit the present invention in any other way. The present invention may have many other embodiments. Without departing from the spirit and essence of the present invention, those skilled in the art can make various corresponding modifications and changes based on the present invention, but all such modifications and changes should fall within the protection scope of the present invention.

Claims

1. A method for macroscopic dynamic traffic flow generation and evolution oriented towards mixed traffic flow, characterized in that, include: Acquire basic traffic flow data of the target road network, including average speed, vehicle density, vehicle type ratio, and autonomous driving penetration rate, wherein the vehicle type ratio includes at least the proportion of heavy trucks µ. Based on the acquired traffic flow data, a heterogeneous disturbance factor is introduced into the continuity equation of the classic LWR model. An improved traffic flow model is obtained, wherein the heterogeneous disturbance factor The ratio µ of heavy trucks is dynamically adjusted to simulate the traffic flow fluctuation effect caused by heavy trucks. Macroscopic traffic flow is calculated using the improved traffic flow model. Based on the macroscopic traffic flow, a non-homogeneous Poisson process is used to generate the probability distribution of vehicle arrivals and to produce a virtual vehicle injection sequence. Real-time detection of vehicle density and speed in downstream road sections. When the vehicle density approaches saturation density and the downstream speed reduction rate meets preset conditions, a queuing flow limiting operator is introduced at the upstream control node to dynamically correct the intensity parameters of the non-homogeneous Poisson process, so as to build a deceleration buffer in advance at the upstream boundary and suppress the reverse shock wave. Using the maximum entropy model, the origin-destination (OD) matrix of the entire network's vehicle travel paths is automatically derived based on the observed intersection traffic flow. The traffic flow evolution results are output, which include at least the virtual vehicle injection sequence, wherein the virtual vehicle injection sequence contains the position, initial velocity, vehicle type, and target path determined from the OD matrix of each virtual vehicle, and is used to inject into the autonomous driving simulation platform.

2. The method according to claim 1, characterized in that, The heterogeneous perturbation factor Constructed through a nonlinear composite mapping, its functional form is: ； Where α is the weighting coefficient for the mixed-flow heterogeneity adjustment. This is the damping attenuation factor for heavy-duty trucks. The value of γ is the autonomous driving penetration compensation index, and it monotonically increases with the autonomous driving penetration rate. This nonlinear composite mapping is used to characterize the situation when the proportion of heavy trucks... The nonlinear accelerated decay characteristics of the overall traffic flow capacity when the elevation increases.

3. The method according to claim 2, characterized in that, In the improved traffic flow model, the formula for calculating the macroscopic traffic flow q is: ； Where k(x,t) represents the vehicle density of a specific road segment at time t, and x is the spatial coordinate along the road direction; The free-flow velocity of the road segment. Blocking density; It is a heterogeneous perturbation factor.

4. The method according to claim 1, characterized in that, The steps of real-time detection of vehicle density and speed in downstream road sections, and when the vehicle density approaches saturation density and the downstream speed reduction rate meets preset conditions, to introduce a queuing and flow-limiting operator at the upstream control node to dynamically correct the intensity parameters of the non-homogeneous Poisson process, specifically include: Real-time monitoring of the carrying capacity of downstream road sections; if the current density... The judgment conditions are met. And the rate of decline in downstream speed At that time, reverse shock wave suppression is activated, and the queuing current limiting operator R(t) is: ; And the intensity parameters of the upstream non-homogeneous Poisson process. Revised to: ; Where ω is the reverse shock wave suppression attenuation coefficient; θ is a preset velocity descent rate threshold; For blocking density.

5. The method according to claim 1, characterized in that, The acquisition of basic traffic flow data for the target road network also includes: acquiring at least one spatiotemporal constraint from the following: traffic light timing schemes at each intersection, road speed limit gradients, and tidal lane status. The acquisition of basic traffic flow data for the target road network also includes: acquiring real-time congestion index and / or weather impact factors as external triggers to adjust traffic flow parameters.

6. The method according to claim 1, characterized in that, The step of obtaining basic traffic flow data of the target road network also includes spatiotemporal alignment of multi-source heterogeneous data, specifically including: Timestamp synchronization: The IEEE 1588 PTP precise time protocol is used to align the time of the intersection checkpoint data with the test vehicle's perception data, and interpolation compensation is used to eliminate transmission delay errors; Spatial coordinate unification: For the physical coordinates of the checkpoint and the coordinates of the perceived target in the vehicle body coordinate system of the test vehicle, a multi-parameter spatial affine transformation model is used to project both onto the global grid coordinate system of the high-precision map through transformation matrix and translation vector.

7. The method according to claim 1, characterized in that, The output traffic flow evolution results also include: road network heat map, key performance indicators, and abnormal scenario reports; The road network heat map represents the dynamic saturation distribution of all road segments; the key performance indicators include average queue length and intersection delay time; and the abnormal scenario report includes automatically identified congestion bottlenecks and their duration.

8. The method according to any one of claims 1-7, characterized in that, The method is applied to the virtual simulation verification of autonomous driving systems in mixed traffic environments with a high proportion of heavy trucks in ports or logistics parks.

9. A macroscopic dynamic traffic flow generation and evolution system for mixed traffic flows, characterized in that, For implementing the method as described in any one of claims 1-7, comprising: The data acquisition module is used to acquire basic traffic flow data of the target road network. The basic traffic flow data includes average speed, vehicle density, vehicle type ratio, and autonomous driving penetration rate, wherein the vehicle type ratio includes at least the proportion of heavy trucks. The model building module is used to introduce heterogeneous disturbance factors into the continuity equation of the classic LWR model based on the acquired traffic flow basic data to obtain an improved traffic flow model. The heterogeneous disturbance factors are dynamically adjusted according to the proportion of heavy trucks in the vehicle type ratio to simulate the traffic flow fluctuation effect caused by heavy trucks. A macro-flow calculation module is used to calculate macro-flow using the improved traffic flow model; The vehicle injection generation module is used to generate a probability distribution of vehicle arrivals based on the macroscopic traffic flow using a non-homogeneous Poisson process, and to generate a virtual vehicle injection sequence. The shock wave suppression module is used to detect the vehicle density and speed of the downstream road section in real time. When the vehicle density is close to the saturation density and the downstream speed reduction rate meets the preset conditions, a queuing flow limiting operator is introduced at the upstream control node to dynamically correct the intensity parameters of the non-homogeneous Poisson process, so as to build a deceleration buffer in advance at the upstream boundary and suppress the reverse shock wave. The OD (Origin-Destination) extrapolation module is used to automatically extrapolate the origin-destination (OD) matrix of the entire network's vehicle travel paths based on the observed intersection traffic flow using the maximum entropy model. The output module is used to output traffic flow evolution results, which include at least the virtual vehicle injection sequence. The virtual vehicle injection sequence contains the position, initial velocity, vehicle type, and target path determined from the OD matrix of each virtual vehicle, and is used to inject the virtual vehicle into the autonomous driving simulation platform.

10. The system according to claim 9, characterized in that, The data acquisition module is also used to acquire at least one spatiotemporal constraint from the traffic light timing scheme at each intersection, the road speed limit gradient, and the tidal lane status, as well as to acquire the real-time congestion index and / or weather influencing factors as external triggering sources.

Images