Simulation method and device for determining risk field based on particle diffusion model

Abstract

The invention provides an analogue simulation method and device for determining a risk field based on a particle diffusion model, and the method comprises the steps: carrying out the analysis of the current sensing data collected by each target sensing device on a target vehicle in a simulation scene, and determining the risk particles included in each current sensing data; performing particle diffusion model updating processing by using each risk particle and the current environmental influence data, and determining at least one candidate particle diffusion model; for each candidate particle diffusion model, re-updating the candidate particle diffusion model by using effective secondary particles generated by chain reaction after collision to obtain at least one optimal particle diffusion model; fusing all the optimal particle diffusion models to determine a target particle diffusion model; and outputting a target risk field according to particle information in the target particle diffusion model. Therefore, the particle diffusion model and dynamics are combined to realize dynamic interactive modeling, so that the dynamic accurate evaluation of the risk field can be effectively realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of automatic driving, in particular to a simulation method and device for determining a risk field based on a particle diffusion model. BACKGROUND

[0002] With the rapid development of automatic driving technology, safety performance testing based on simulation has become a core means of system verification and regulation formulation. By constructing a complex traffic scene that is repeatable and highly realistic, the response and decision-making ability of the automatic driving system in dynamic risks can be systematically evaluated. At present, the mainstream risk modeling method in simulation testing mainly relies on the Euclidean distance field or the Gaussian diffusion model. These methods are computationally efficient in simple structured scenes and can provide basic risk distribution information.

[0003] However, in the context of complex simulation testing environments that are highly realistic and have multiple physical interactions, the existing methods have significant limitations. First, in simulating geometrically complex scenes such as curves and irregular intersections, the static metric based on Euclidean distance cannot conform to the actual road curvature and driving path, resulting in distortion of the risk field and possible false positives or false negatives. Second, when simulating sensor performance degradation under adverse weather conditions such as rain, fog, and snow, the existing probabilistic models have difficulty in fully depicting the nonlinear and multimodal dynamic evolution of perception errors, reducing the credibility of risk assessment.

[0004] More fundamentally, the existing methods lack the ability to deeply simulate dynamic physical processes after the triggering of risk events. They mostly remain at the static or simple probability level and cannot effectively simulate key processes such as collision energy transfer, complex interactions between vehicles and vulnerable road users, and chain accidents. This results in insufficient coverage of extreme high-risk scenarios such as highway rear-end collisions in simulation testing, forming a blind spot in safety verification.

[0005] Therefore, there is an urgent need to develop a new high-precision dynamic risk scene modeling technology to overcome the limitations of existing methods and to build a more realistic and comprehensive dynamic risk deduction system in the simulation environment, providing a more solid and reliable foundation for automatic driving safety assessment. SUMMARY

[0006] Therefore, the purpose of the present application is to provide a simulation method and device for determining a risk field based on a particle diffusion model, which combines particle diffusion models with dynamics to achieve dynamic interaction modeling, thereby effectively achieving dynamic and accurate risk field assessment in simulation scenarios.

[0007] The simulation method for determining a risk field based on a particle diffusion model provided by the embodiments of the present application comprises: In the simulation scene, the current perception data collected by each target perception device on the target vehicle is analyzed respectively to determine the risk particles included in each current perception data; For each current perception data, a particle diffusion model update process is performed according to the risk particles included in the current perception data and the current environment influence data to determine at least one candidate particle diffusion model; For each candidate particle diffusion model, the particles in the candidate particle diffusion model are subjected to collision detection, and the candidate particle diffusion model is updated again according to the effective secondary particles generated by the chain reaction after the collision is detected, to obtain at least one optimal particle diffusion model; All optimal particle diffusion models are fused to determine a target particle diffusion model; According to the particle information in the target particle diffusion model, a risk field generation process is performed, and the load consumption during the risk field generation process is monitored; When the load consumption meets the preset load requirement, a target risk field is output.

[0008] Optionally, the analysis of the current perception data collected by each target perception device on the target vehicle to determine the risk particles included in each current perception data comprises: For the current perception data collected by each target perception device on the target vehicle, the object information included in the current perception data is identified to determine whether it includes a target object of a specified type; According to the specific data of the target object of the specified type and in combination with a preset particle configuration rule, the risk particles included in the current perception data are determined.

[0009] Optionally, the current environment influence data includes a current environment diffusion coefficient and a current road potential field gradient matrix, and the current environment influence data is determined by the following steps: A high-precision map is obtained in advance; A grid map is generated according to the road features in the high-precision map; A current road potential field gradient matrix including a preset size area centered on the target vehicle is constructed according to the grid position attributes in the grid map; The current environment diffusion coefficient is determined according to the target type environment data around the target vehicle.

[0010] Optionally, when the particles in the candidate particle diffusion model are subjected to collision detection and it is determined that no collision occurs, the simulation method further comprises: The candidate particle diffusion model is determined as an optimal particle diffusion model, and all optimal particle diffusion models are fused to determine a target particle diffusion model.

[0011] Optionally, for each candidate particle diffusion model, collision detection is performed on the particles in the candidate particle diffusion model, and the candidate particle diffusion model is updated again based on the effective secondary particles generated by the chain reaction after the collision, to obtain at least one optimal particle diffusion model, including: For each candidate particle diffusion model, collision detection is performed on the particles in the candidate particle diffusion model to determine whether a collision has occurred. If a collision occurs, determine whether a chain reaction occurs based on the collision energy generated during the collision. If a chain reaction occurs, identify the initial secondary particles produced when the chain reaction occurs; Invalid secondary particles are removed from the initial secondary particles according to the preset screening principle to obtain valid secondary particles; The candidate particle diffusion model is updated again using effective secondary particles to obtain the optimal particle diffusion model corresponding to the candidate particle diffusion model.

[0012] Optionally, when the load consumption does not meet the preset load requirement, and the load consumption is lower than the lower limit of the preset load requirement, the simulation method further includes: The current environmental diffusion coefficient in the current environmental impact data is updated, and the particle diffusion model is updated again based on the updated current environmental impact data to obtain the target particle diffusion model. Based on the particle information in the target particle diffusion model, the target risk field whose load consumption meets the preset load requirements is output.

[0013] Optionally, when the load consumption does not meet the preset load requirements, and the load consumption exceeds the upper limit of the preset load requirements, the simulation method further includes: Reacquire the current perception data collected by each target perception device on the target vehicle and generate the target risk field.

[0014] This application also provides a simulation device for a risk field based on a particle diffusion model, the simulation device comprising: The analysis module is used to analyze the current perception data collected by each target perception device on the target vehicle in a simulation scenario, and to determine the risk particles included in each type of current perception data. The first update module is used to perform particle diffusion model update processing for each type of current sensing data based on the risk particles included in the current sensing data and the current environmental impact data, and to determine at least one candidate particle diffusion model. The second update module is used to perform collision detection on the particles in each candidate particle diffusion model, and update the candidate particle diffusion model again based on the effective secondary particles generated by the chain reaction after the collision, so as to obtain at least one optimal particle diffusion model. The fusion module is used to fuse all the best particle diffusion models to determine the target particle diffusion model. The monitoring module is used to perform risk field generation processing based on the particle information in the target particle diffusion model, and to monitor the load consumption during the risk field generation process. The output module is used to output the target risk field when the load consumption meets the preset load requirements.

[0015] Optionally, when the analysis module analyzes the current sensing data collected by each target sensing device on the target vehicle to determine the risk particles included in each type of current sensing data, the analysis module is used to: For the current perception data collected by each target perception device on the target vehicle, the object information included in the current perception data is identified to determine whether it includes a target object of a specified type; Based on the specific data of the target objects of the specified type, and in conjunction with the preset particle configuration rules, determine the risk particles included in the current perception data.

[0016] Optionally, the current environmental impact data includes the current environmental diffusion coefficient and the current road potential field gradient matrix. The simulation device further includes a determination module, which is used to determine the current environmental impact data through the following steps: Obtain a pre-built high-precision map; Generate a grid map based on the road features in the high-precision map; Based on the grid location attributes in the grid map, construct the current road potential field gradient matrix centered on the target vehicle and including a region of a preset size; The current environmental diffusion coefficient is determined based on environmental data of the target type surrounding the target vehicle.

[0017] Optionally, the fusion module is further configured to: When collision detection is performed on the particles in the candidate particle diffusion model and it is determined that no collision occurs, the candidate particle diffusion model is determined as the optimal particle diffusion model. All optimal particle diffusion models are then fused together to determine the target particle diffusion model.

[0018] Optionally, when the second update module performs collision detection on the particles in each candidate particle diffusion model and updates the candidate particle diffusion model again based on the effective secondary particles generated by the chain reaction after the collision, to obtain at least one optimal particle diffusion model, the second update module is used to: For each candidate particle diffusion model, collision detection is performed on the particles in the candidate particle diffusion model to determine whether a collision has occurred. If a collision occurs, determine whether a chain reaction occurs based on the collision energy generated during the collision. If a chain reaction occurs, identify the initial secondary particles produced when the chain reaction occurs; Invalid secondary particles are removed from the initial secondary particles according to the preset screening principle to obtain valid secondary particles; The candidate particle diffusion model is updated again using effective secondary particles to obtain the optimal particle diffusion model corresponding to the candidate particle diffusion model.

[0019] Optionally, the simulation device is also used for: When the load consumption does not meet the preset load requirements, and the load consumption is lower than the lower limit of the preset load requirements, the current environmental diffusion coefficient in the current environmental impact data is updated, and the particle diffusion model is updated again based on the updated current environmental impact data to obtain the target particle diffusion model. Based on the particle information in the target particle diffusion model, the target risk field whose load consumption meets the preset load requirements is output.

[0020] Optionally, the simulation device is also used for: When the load consumption does not meet the preset load requirements, and the load consumption is higher than the upper limit of the preset load requirements, the current perception data collected by each target perception device on the target vehicle is reacquired, and the target risk field is generated.

[0021] This application also provides an electronic device, including: a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory via the bus. When the machine-readable instructions are executed by the processor, the steps of the simulation method described above are performed.

[0022] This application also provides a computer-readable storage medium storing a computer program, which, when run by a processor, executes the steps of the simulation method described above.

[0023] This application provides a simulation method and apparatus for a risk field based on a particle diffusion model. The simulation method includes: analyzing the current perception data collected by each target perception device on the target vehicle in a simulation scenario to determine the risk particles included in each type of current perception data; updating the particle diffusion model for each type of current perception data based on the risk particles included in the current perception data and the current environmental impact data to determine at least one candidate particle diffusion model; performing collision detection on the particles in each candidate particle diffusion model and updating the candidate particle diffusion model again based on the effective secondary particles generated by the chain reaction after the collision, to obtain at least one optimal particle diffusion model; fusing all optimal particle diffusion models to determine the target particle diffusion model; performing risk field generation processing based on the particle information in the target particle diffusion model and monitoring the load consumption during the risk field generation process; and outputting the target risk field when the load consumption meets the preset load requirements. Thus, for the simulation of risk fields, this application effectively solves the problems of inaccurate risk field assessment and limited evaluation results in existing technologies by introducing risk particles and dynamically updated particle diffusion models, and innovatively integrating collision detection and chain reaction simulation mechanisms. Furthermore, the technical solution provided by this application also has the following beneficial effects: First, by directly generating risk particles adapted to specific scenarios from sensing data and updating the model in conjunction with current environmental impact data, it can better fit the actual geometric and dynamic characteristics of unstructured roads such as curves, thereby significantly reducing the positioning bias present in traditional Euclidean distance field methods. Second, by independently analyzing and generating candidate models from data from different sensing devices, and then fusing them into a target particle diffusion model, this process essentially achieves the probabilistic fusion of multi-source heterogeneous information and uncertainty propagation, which can more accurately characterize the nonlinear evolution of sensing errors under rainy and foggy weather, improving the credibility of risk assessment. Furthermore, this solution adds a step of updating the effective secondary particles generated by particle collision detection and chain reactions. This is the first time that explicit simulation of continuous dynamic physical processes such as collision energy propagation and derivative hazards has been introduced into risk field modeling. This enables effective simulation and prediction of complex high-risk scenarios such as multi-vehicle pileups on highways, fundamentally reducing the false negative rate in such scenarios. Finally, by monitoring load consumption and controlling the output with preset load requirements, the method ensures that it generates a high-precision dynamic risk field while possessing computational feasibility and engineering practicality, meeting the balance requirements of real-time performance and resource consumption for autonomous driving safety assessment systems.

[0024] In summary, this solution breaks through the limitations of traditional static distance field or simple probability diffusion models in terms of model mechanism, and realizes high-precision integrated modeling of dynamic risks, physical processes and uncertainties, providing a more reliable technical foundation for the safety assessment of autonomous driving systems.

[0025] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0026] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0027] Figure 1 A flowchart illustrating a simulation method for determining a risk field based on a particle diffusion model, provided as an embodiment of this application; Figure 2 This is a timing diagram of GPU-CPU parallel processing provided in this application; Figure 3 This is one of the structural schematic diagrams of a simulation device for determining the risk field based on a particle diffusion model, provided in an embodiment of this application. Figure 4 This is a second schematic diagram of a simulation device for determining a risk field based on a particle diffusion model, provided in an embodiment of this application. Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. Based on the embodiments of this application, every other embodiment obtained by those skilled in the art without inventive effort falls within the scope of protection of this application.

[0029] First, the applicable application scenarios for this application will be introduced. This application can be applied to the field of autonomous driving.

[0030] Research has revealed that with the rapid development of autonomous driving technology, its safety performance assessment has become a core aspect of industrial implementation and regulatory formulation. Accurate modeling and prediction of dynamic risk scenarios that vehicles may encounter during operation are crucial technological foundations for ensuring the reliability of autonomous driving systems. Currently, mainstream risk scenario modeling methods primarily rely on Euclidean Distance Transform (EDT) or probabilistic Gaussian diffusion models. Euclidean distance transformation (EDT) characterizes risk distribution by calculating the Euclidean distance from each point in the environment to the nearest obstacle; this method is intuitive and computationally efficient. Gaussian diffusion models, on the other hand, simulate the impact of uncertainty by probabilistically diffusing potential risk sources. These methods can provide some risk indication in simple, structured environments.

[0031] However, in practical road applications, especially in complex and ever-changing real-world traffic environments, existing technologies have revealed serious limitations, failing to meet the demands for high-precision safety assessments. Specifically, their shortcomings are mainly reflected in the following aspects: First, in scenarios with complex road geometry, such as continuous curves or irregular intersections, Euclidean distance-based metrics cannot accurately reflect the actual driving path and risk relationship along the road curvature. This leads to severely inaccurate risk localization, potentially resulting in erroneous hazard warnings or missed warnings. Second, in adverse weather conditions with poor visibility or impaired sensor performance (such as rain, fog, and snow), the uncertainty of environmental perception increases dramatically. Current probabilistic models are insufficient to fully characterize this nonlinear, multimodal evolution of perception errors, resulting in extremely low reliability of risk assessment results at critical moments.

[0032] Even more serious is the fundamental flaw in existing modeling methods regarding the deep characterization of the physical processes involved in risk events. Most remain at the level of static obstacle distances or simple probability distributions, failing to effectively simulate and extrapolate the dynamic physical processes following the triggering of risk events. For example, existing methods cannot simulate critical physical processes such as collision energy propagation and secondary accident chains, leading to a high rate of false negatives in high-risk scenarios like chain-reaction rear-end collisions on highways. This prevents safety testing of autonomous driving systems from covering such extreme but highly serious situations, leaving significant safety hazards.

[0033] Based on this, the embodiments of this application provide a simulation method and apparatus for determining the risk field based on a particle diffusion model. By combining the particle diffusion model with dynamics, dynamic interactive modeling is achieved, thereby effectively realizing the dynamic and accurate assessment of the risk field.

[0034] Please see Figure 1 , Figure 1This is a flowchart illustrating a simulation method for determining a risk field based on a particle diffusion model, provided as an embodiment of this application. Figure 1 As shown in the embodiments of this application, the simulation method includes: S101. In the simulation scenario, analyze the current perception data collected by each target perception device on the target vehicle to determine the risk particles included in each type of current perception data. S102. For each type of current sensing data, perform particle diffusion model update processing based on the risk particles included in the current sensing data and the current environmental impact data, and determine at least one candidate particle diffusion model. S103. For each candidate particle diffusion model, perform collision detection on the particles in the candidate particle diffusion model, and update the candidate particle diffusion model again based on the effective secondary particles generated by the chain reaction after the collision, so as to obtain at least one optimal particle diffusion model. S104. Merge all the optimal particle diffusion models to determine the target particle diffusion model.

[0035] S105. Based on the particle information in the target particle diffusion model, perform risk field generation processing and monitor the load consumption during the risk field generation process.

[0036] S106. When the load consumption meets the preset load requirements, output the target risk field.

[0037] The exemplary steps of the embodiments of this application are described below: Before executing step S101, it is necessary to determine the current environmental impact data that affects the particle diffusion model, that is, the dynamic risk field to be determined. This current environmental impact data will be dynamically adjusted according to the vehicle's movement, and can be simulation data.

[0038] For example, in one embodiment provided in this application, the current environmental impact data includes the current environmental diffusion coefficient and the current road potential field gradient matrix, and the current environmental impact data is determined through the following steps: S201. Obtain a pre-built high-precision map.

[0039] S202. Generate a grid map based on the road features in the high-precision map.

[0040] S203. Based on the grid location attributes in the grid map, construct the current road potential field gradient matrix centered on the target vehicle and including a preset size region.

[0041] S204. Determine the current environmental diffusion coefficient based on the environmental data of the target type around the target vehicle.

[0042] For step S201, first start the initialization, and then obtain a pre-built high-precision map including the specified area.

[0043] The initialization process typically includes powering on the system, loading the core driver, and initializing hardware modules and preset load requirements.

[0044] The initialization hardware module includes: setting up the sensor interface (millimeter-wave radar / camera), setting up the GPU computing unit, and setting up the road potential field processor (FPGA chip, i.e., semiconductor chip).

[0045] In this step, the high-precision map can be read from the vehicle's onboard storage.

[0046] In step S202, the extracted road features include road curvature features and road slope features. Then, based on the extracted road features, a grid map is constructed at a preset resolution.

[0047] Here, the preset resolution can be 0.2m × 0.2m, and the spatial index format of the constructed grid map can be set, such as GeoHash encoding.

[0048] The area included in the grid map can be the entire area of ​​the high-precision map or a partial area.

[0049] Regarding step S203, the specific calculation process of the road potential field gradient matrix is ​​as follows: First, the road potential field is calculated grid by grid. Then, calculate the spatial gradient. Finally, it is stored as a matrix structure: that is, the dimension is M×N (M=number of map rows, N=number of map columns), and the precision is single-precision floating point.

[0050] Road potential field gradient matrix The calculation formula is as follows:

[0051] in, For curvature weight, the example can be set to 1.5; γ is the slope weight, which can be set to 0.8 in the example; γ is the potential field normalization factor. Let be the curvature (rad / m) at position (x,y); Let be the slope angle (°) of position (x,y); i and j represent unit vectors in the standard Cartesian coordinate system, i representing the unit vector in the x-direction, i.e., the longitudinal direction of the road (the direction in which the vehicle moves forward); j representing the unit vector in the y-direction, i.e., the transverse direction of the road (the direction in which the vehicle deviates to the left or right).

[0052] Regarding step S204, the environmental data for the target type can specifically be rainfall intensity acquired in real time via millimeter-wave radar. and the road surface friction coefficient obtained through infrared cameras. .

[0053] In this step, the current environmental diffusion coefficient (also known as the environment-dependent diffusion coefficient) is determined based on the environmental data of the target type around the target vehicle. Specifically, it can be calculated using the following formula.

[0054] Environmental diffusion coefficient Calculation formula:

[0055] The meaning of each letter in the formula is shown in Table 1, which is a reference table for symbol information provided in this application.

[0056] Table 1:

[0057] Once the current environmental impact data is determined, the risk field can be dynamically and cyclically determined, and a real-time calculation cycle can be started. For example, this includes: activating the cycle timer and initializing the particle counter: current particle count = number of environmental obstacles × (1.5~5.0).

[0058] Specifically, step S101 may include: determining the number and type of target sensing devices installed on the target vehicle; fusing data collected by target sensing devices of the same type to obtain current sensing data collected by that type of target sensing device; and analyzing each type of current sensing data to obtain the risk particles included in each type of current sensing data. When used in a simulation scenario, the current sensing data can be determined through simulation.

[0059] Here, the types of target perception devices may include lidar (for acquiring point cloud coordinates of obstacles), visible light cameras (for determining the type / speed of moving targets), and IMU (for determining the vehicle's acceleration / yaw angle).

[0060] Furthermore, in one embodiment provided in this application, the step of analyzing the current perception data collected by each target perception device on the target vehicle to determine the risk particles included in each type of current perception data includes: S1011. For the current perception data collected by each target perception device on the target vehicle, identify the object information included in the current perception data and determine whether it includes a target object of a specified type.

[0061] S1012. Based on the specific data of the target objects of the specified type included, and in conjunction with the preset particle configuration rules, determine the risk particles included in the current perception data.

[0062] For step S1011, the target object of the specified type can be predetermined.

[0063] For example, when analyzing current sensing data, it is determined whether the current sensing data includes motor vehicles, pedestrians, and static obstacles.

[0064] For step S1012, please refer to Table 2 for the particle configuration rules. Table 2 is a particle configuration rule comparison table provided in this application.

[0065] Table 2:

[0066] The initialization parameters of the determined risk particle can be set as follows: position = object centroid coordinates; initial velocity = object velocity vector.

[0067] Specifically, step S102 may include: the CPU sending the risk particles and current environmental impact data included in the current sensing data to the GPU, and the GPU performing particle diffusion model update processing in parallel to determine at least one candidate particle diffusion model.

[0068] Here, when it is the first time to update the particle diffusion model, the initial particle diffusion model is updated; when it is not the first time to update the particle diffusion model, the previously determined target particle diffusion model is updated.

[0069] When using the GPU to update the particle model, the GPU will perform computational configuration, which may include, for example: thread allocation: 256 particles / thread block; time step: Δt=0.01s.

[0070] When updating the particle diffusion model, the following formula can be used:

[0071] Please refer to Table 1 for the meaning of each letter in the formula.

[0072] It should be noted that, in order to improve computing speed, this application provides a GPU-CPU parallel processing architecture. For an example, please refer to [link / reference needed]. Figure 2 , Figure 2This is a timing diagram illustrating the GPU-CPU parallel processing provided in this application. Figure 2 As shown, the timeliness of risk field output can be guaranteed by parallel processing of GPU and CPU.

[0073] Regarding step S103, in one embodiment provided in this application, for each candidate particle diffusion model, collision detection is performed on the particles in the candidate particle diffusion model, and the candidate particle diffusion model is updated again based on the effective secondary particles generated by the chain reaction after the collision, to obtain at least one optimal particle diffusion model, including: S1031. For each candidate particle diffusion model, perform collision detection on the particles in the candidate particle diffusion model to determine whether a collision has occurred.

[0074] S1032. If a collision occurs, determine whether a chain reaction occurs based on the collision energy generated during the collision.

[0075] S1033. If a chain reaction occurs, identify the initial secondary particles generated when the chain reaction occurs.

[0076] S1034. According to the preset screening principle, the invalid secondary particles in the initial secondary particles are removed to obtain the valid secondary particles.

[0077] S1035. Use effective secondary particles to update the candidate particle diffusion model again to obtain the optimal particle diffusion model corresponding to the candidate particle diffusion model.

[0078] Specifically, step S1031 may include performing collision detection on the particles in each candidate particle diffusion model to determine whether there is collision energy, and if so, determining that a collision has occurred.

[0079] Here, collision detection can be performed based on the spatial segmentation method (BVH tree).

[0080] Among them, collision energy release The calculation formula is as follows:

[0081] in, The initial kinetic energy of the particle is (J). Let be the velocity vector (m / s) of the object being struck. Material conversion efficiency (steel = 0.85, plastic = 0.2).

[0082] Specifically, step S1032 may include: if the collision energy generated during the collision is greater than the energy threshold, then a chain reaction is determined to have occurred, and new particles (secondary particles) N will be generated.

[0083] Regarding step S1033, the formula for generating the initial secondary particles is as follows:

[0084]

[0085] in, The single-particle generation threshold is 10J. The longitudinal velocity transmission coefficient is (0.6-0.9). The transverse transfer coefficient is (0.3-0.7). The direction vector is random and uniformly distributed in [-1,1]. : This is a rounding function that returns the integer part of a value.

[0086] For step S1034, this step specifically includes: removing initial secondary particles that have exceeded the boundary or whose life cycle has ended, and determining the remaining secondary particles after removal as valid secondary particles.

[0087] For step S1035, for each effective secondary particle determined by the current sensing data, the corresponding candidate diffusion model is updated again using the effective secondary particle to obtain its corresponding optimal particle diffusion model.

[0088] For step S104, this step includes: fusing all the optimal particle diffusion models and deleting duplicate particles to determine the target particle diffusion model.

[0089] Furthermore, in one embodiment provided in this application, when collision detection is performed on the particles in the candidate particle diffusion model and it is determined that no collision occurs, the simulation method further includes: determining the candidate particle diffusion model as the optimal particle diffusion model, and fusing all optimal particle diffusion models to determine the target particle diffusion model.

[0090] In this embodiment, the target particle diffusion model is determined directly using a candidate particle model.

[0091] Specifically, step S105 may include: converting a high-speed moving particle swarm into a spatially continuous risk density field based on the particle information included in the target particle diffusion model, and monitoring whether the time taken to generate the risk field is within a preset time window.

[0092] The process of determining the risk field here includes the following steps: First, divide the space into grids (for example, set the grid size to 1.0m × 1.0m cells) to cover the minimum resolution of a standard lane width (3.5m); Then, set the spatial range: for example, ±50m (longitudinal) × ±20m (lateral) to cover the effective perception range of autonomous driving; Finally, calculate the grid risk density. Used to determine the risk field:

[0093] in, This is the function bandwidth / influence range parameter, with a default value of 2.0m, derived from experimental data on the impact range of collisions on urban roads; For particle weighting factors; | | represents the Euclidean distance from the grid center to the particle.

[0094] For step S106, check whether the time taken to generate the risk field is within the preset time window. If the preset load requirement is met, output the risk field to obtain the current target risk field.

[0095] Furthermore, in one embodiment provided in this application, when the load consumption does not meet the preset load requirement and the load consumption is lower than the lower limit of the preset load requirement, the simulation method further includes: updating the current environmental diffusion coefficient in the current environmental impact data, and re-updating the particle diffusion model based on the updated current environmental impact data to obtain a target particle diffusion model; and outputting a target risk field in which the load consumption meets the preset load requirement based on the particle information in the target particle diffusion model.

[0096] In this step, if the load consumption is determined to be lower than the preset load requirement, it means that the determination of the risk field is too fast and the accuracy may be low. Therefore, the accuracy can be improved and the target risk field can be determined again.

[0097] Specifically, the accuracy can be improved by changing the environmental diffusion coefficient in the current environmental impact data, and then the process returns to step S102.

[0098] Furthermore, in another embodiment provided in this application, when the load consumption does not meet the preset load requirement and the load consumption is higher than the upper limit of the preset load requirement, the simulation method further includes: reacquiring the current perception data collected by each target perception device on the target vehicle and generating a target risk field.

[0099] Here, if the target risk field cannot be generated within the predetermined window based on the currently collected sensing data, the current risk field generation is stopped, and the process returns to step S101 to regenerate the target risk field based on the newly collected sensing data.

[0100] In summary, this solution breaks through the limitations of traditional static distance field or simple probability diffusion models in terms of model mechanism, and realizes high-precision integrated modeling of dynamic risks, physical processes and uncertainties, providing a more reliable technical foundation for the safety assessment of autonomous driving systems.

[0101] Based on the same inventive concept, this application also provides a simulation device corresponding to the simulation method. Since the principle of the device in this application is similar to the simulation method described above, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be described again.

[0102] Please see Figure 3 , Figure 4 , Figure 3 This is one of the structural schematic diagrams of a simulation device for determining the risk field based on a particle diffusion model, provided in an embodiment of this application. Figure 4 This is a second schematic diagram of a simulation device for determining a risk field based on a particle diffusion model, provided as an embodiment of this application. Figure 3 As shown, the simulation device 300 includes: Analysis module 310 is used to analyze the current perception data collected by each target perception device on the target vehicle in a simulation scenario, and determine the risk particles included in each type of current perception data. The first update module 320 is used to perform particle diffusion model update processing for each type of current sensing data based on the risk particles included in the current sensing data and the current environmental impact data, and to determine at least one candidate particle diffusion model. The second update module 330 is used to perform collision detection on the particles in each candidate particle diffusion model, and update the candidate particle diffusion model again based on the effective secondary particles generated by the chain reaction after the collision, so as to obtain at least one optimal particle diffusion model. The fusion module 340 is used to fuse all the best particle diffusion models to determine the target particle diffusion model; The monitoring module 350 is used to perform risk field generation processing based on the particle information in the target particle diffusion model, and to monitor the load consumption during the risk field generation process. The output module 360 ​​is used to output the target risk field when the load consumption meets the preset load requirements.

[0103] Optionally, when the analysis module 310 analyzes the current sensing data collected by each target sensing device on the target vehicle to determine the risk particles included in each type of current sensing data, the analysis module 310 is used to: For the current perception data collected by each target perception device on the target vehicle, the object information included in the current perception data is identified to determine whether it includes a target object of a specified type; Based on the specific data of the target objects of the specified type, and in conjunction with the preset particle configuration rules, determine the risk particles included in the current perception data.

[0104] Optional, such as Figure 4 As shown, the current environmental impact data includes the current environmental diffusion coefficient and the current road potential field gradient matrix. The simulation device 300 also includes a determination module 370, which is used to determine the current environmental impact data through the following steps: Obtain a pre-built high-precision map; Generate a grid map based on the road features in the high-precision map; Based on the grid location attributes in the grid map, construct the current road potential field gradient matrix centered on the target vehicle and including a region of a preset size; The current environmental diffusion coefficient is determined based on environmental data of the target type surrounding the target vehicle.

[0105] Optionally, the fusion module 340 is further configured to: When collision detection is performed on the particles in the candidate particle diffusion model and it is determined that no collision occurs, the candidate particle diffusion model is determined as the optimal particle diffusion model. All optimal particle diffusion models are then fused together to determine the target particle diffusion model.

[0106] Optionally, when the second update module 330 performs collision detection on the particles in each candidate particle diffusion model and updates the candidate particle diffusion model again based on the effective secondary particles generated by the chain reaction after the collision, to obtain at least one optimal particle diffusion model, the second update module 330 is used to: For each candidate particle diffusion model, collision detection is performed on the particles in the candidate particle diffusion model to determine whether a collision has occurred. If a collision occurs, determine whether a chain reaction occurs based on the collision energy generated during the collision. If a chain reaction occurs, identify the initial secondary particles produced when the chain reaction occurs; Invalid secondary particles are removed from the initial secondary particles according to the preset screening principle to obtain valid secondary particles; The candidate particle diffusion model is updated again using effective secondary particles to obtain the optimal particle diffusion model corresponding to the candidate particle diffusion model.

[0107] Optionally, the simulation device 300 is further used for: When the load consumption does not meet the preset load requirements, and the load consumption is lower than the lower limit of the preset load requirements, the current environmental diffusion coefficient in the current environmental impact data is updated, and the particle diffusion model is updated again based on the updated current environmental impact data to obtain the target particle diffusion model. Based on the particle information in the target particle diffusion model, the target risk field whose load consumption meets the preset load requirements is output.

[0108] Optionally, the simulation device 300 is further used for: When the load consumption does not meet the preset load requirements, and the load consumption is higher than the upper limit of the preset load requirements, the current perception data collected by each target perception device on the target vehicle is reacquired, and the target risk field is generated.

[0109] Please see Figure 5 , Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 5 As shown, the electronic device 500 includes a processor 510, a memory 520, and a bus 530.

[0110] The memory 520 stores machine-readable instructions executable by the processor 510. When the electronic device 500 is running, the processor 510 and the memory 520 communicate via the bus 530. When the machine-readable instructions are executed by the processor 510, they can perform the operations described above. Figure 1 as well as Figure 2 The steps in the method embodiment shown are specifically implemented in the method embodiment and will not be repeated here.

[0111] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, can perform the above-described actions. Figure 1 as well as Figure 2 The steps in the method embodiment shown are specifically implemented in the method embodiment and will not be repeated here.

[0112] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0113] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the shown or discussed mutual couplings, direct couplings, or communication connections may be through some communication interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.

[0114] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0115] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0116] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0117] Finally, it should be noted that the above-described embodiments are merely specific implementations of this application, used to illustrate the technical solutions of this application, and not to limit them. The scope of protection of this application is not limited thereto. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features, within the scope of the technology disclosed in this application. Such modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A simulation method for determining the risk field based on a particle diffusion model, characterized in that, The simulation method includes: In the simulation scenario, the current perception data collected by each target perception device on the target vehicle is analyzed to determine the risk particles included in each type of current perception data. For each type of current sensing data, the particle diffusion model is updated based on the risk particles included in the current sensing data and the current environmental impact data, and at least one candidate particle diffusion model is determined. For each candidate particle diffusion model, collision detection is performed on the particles in the candidate particle diffusion model, and the candidate particle diffusion model is updated again based on the effective secondary particles generated by the chain reaction after the collision occurs, so as to obtain at least one optimal particle diffusion model. By fusing all the optimal particle diffusion models, the target particle diffusion model is determined. Based on the particle information in the target particle diffusion model, risk field generation is performed, and the load consumption during the risk field generation process is monitored. When the load consumption meets the preset load requirements, the target risk field is output.

2. The simulation method according to claim 1, characterized in that, The analysis of the current sensing data collected by each target sensing device on the target vehicle to determine the risk particles included in each type of current sensing data includes: For the current perception data collected by each target perception device on the target vehicle, the object information included in the current perception data is identified to determine whether it includes a target object of a specified type; Based on the specific data of the target objects of the specified type, and in conjunction with the preset particle configuration rules, determine the risk particles included in the current perception data.

3. The simulation method according to claim 1, characterized in that, The current environmental impact data includes the current environmental diffusion coefficient and the current road potential field gradient matrix. The current environmental impact data is determined through the following steps: Obtain a pre-built high-precision map; Generate a grid map based on the road features in the high-precision map; Based on the grid location attributes in the grid map, construct the current road potential field gradient matrix centered on the target vehicle and including a region of a preset size; The current environmental diffusion coefficient is determined based on environmental data of the target type surrounding the target vehicle.

4. The simulation method according to claim 1, characterized in that, When collision detection is performed on particles in the candidate particle diffusion model and it is determined that no collision occurs, the simulation method further includes: The candidate particle diffusion model is selected as the optimal particle diffusion model, and all optimal particle diffusion models are fused to determine the target particle diffusion model.

5. The simulation method according to claim 1, characterized in that, For each candidate particle diffusion model, collision detection is performed on the particles in the candidate particle diffusion model, and the candidate particle diffusion model is updated based on the effective secondary particles generated by the chain reaction after the collision, to obtain at least one optimal particle diffusion model, including: For each candidate particle diffusion model, collision detection is performed on the particles in the candidate particle diffusion model to determine whether a collision has occurred. If a collision occurs, determine whether a chain reaction occurs based on the collision energy generated during the collision. If a chain reaction occurs, identify the initial secondary particles produced when the chain reaction occurs; Invalid secondary particles are removed from the initial secondary particles according to the preset screening principle to obtain valid secondary particles; The candidate particle diffusion model is updated again using effective secondary particles to obtain the optimal particle diffusion model corresponding to the candidate particle diffusion model.

6. The simulation method according to claim 1, characterized in that, When the load consumption does not meet the preset load requirement, and the load consumption is below the lower limit of the preset load requirement, the simulation method further includes: The current environmental diffusion coefficient in the current environmental impact data is updated, and the particle diffusion model is updated again based on the updated current environmental impact data to obtain the target particle diffusion model. Based on the particle information in the target particle diffusion model, the target risk field whose load consumption meets the preset load requirements is output.

7. The simulation method according to claim 1, characterized in that, When the load consumption does not meet the preset load requirement, and the load consumption exceeds the upper limit of the preset load requirement, the simulation method further includes: Reacquire the current perception data collected by each target perception device on the target vehicle and generate the target risk field.

8. A simulation device for a risk field based on a particle diffusion model, characterized in that, The simulation device includes: The analysis module is used to analyze the current perception data collected by each target perception device on the target vehicle in a simulation scenario, and to determine the risk particles included in each type of current perception data. The first update module is used to perform particle diffusion model update processing for each type of current sensing data based on the risk particles included in the current sensing data and the current environmental impact data, and to determine at least one candidate particle diffusion model. The second update module is used to perform collision detection on the particles in each candidate particle diffusion model, and update the candidate particle diffusion model again based on the effective secondary particles generated by the chain reaction after the collision, so as to obtain at least one optimal particle diffusion model. The fusion module is used to fuse all the best particle diffusion models to determine the target particle diffusion model. The monitoring module is used to perform risk field generation processing based on the particle information in the target particle diffusion model, and to monitor the load consumption during the risk field generation process. The output module is used to output the target risk field when the load consumption meets the preset load requirements.

9. An electronic device, characterized in that, include: The device includes a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory via the bus, and the machine-readable instructions are executed by the processor to perform the steps of the simulation method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, performs the steps of the simulation method as described in any one of claims 1 to 7.

Images