Intelligent generation method for autonomous ability test scene of ground unmanned equipment

Abstract

The invention discloses an intelligent generation method for an autonomous ability test scene of ground unmanned equipment, and the method comprises the steps: building a multi-level simulation scene terrain model through noise-based height map generation, terrain erosion simulation, landform parameterization modeling and terrain fusion, and far-end scene modeling; based on generation of dynamic illumination and weather conditions and generation of static simulation elements of terrain coverage elements, unstructured roads and scene vegetation, scene natural environment modeling with high authenticity and complexity is carried out; and automatically generating a test scene case aiming at the autonomous reconnaissance and maneuverability of the unmanned equipment by combining an agent behavior model and a dynamic triggering rule. According to the method, the test scene oriented to the autonomous reconnaissance capability and the autonomous maneuvering capability of the ground unmanned equipment is intelligently constructed in the controlled simulation environment, and various unmanned system simulation scenes are automatically constructed within second-level time, so that the problems of low construction efficiency and high cost of the current ground unmanned equipment test simulation scene are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of testing of unmanned equipment, and particularly relates to an intelligent generation method for an autonomous ability test scenario of ground unmanned equipment. Background Art

[0002] Comprehensive testing is a prerequisite for large-scale deployment and application of ground unmanned equipment. Among many evaluation methods, simulation testing can conduct all-factor and full-process training and testing on the autonomous ability of ground unmanned equipment in a virtual environment, which helps to improve the autonomous reconnaissance and automatic maneuvering abilities of ground unmanned equipment, thereby improving its combat effectiveness. During the simulation process of ground unmanned equipment, in addition to paying attention to the accumulation of test mileage, it is also necessary to attach importance to improving the scene quality. A diverse and highly covered simulation scene library has become the "cornerstone" for improving the intelligence level of ground unmanned equipment.

[0003] However, the deficiencies of the existing technology are that the current simulation testing technology still mainly relies on manual CAD modeling in scene construction, which is time-consuming and laborious to construct large-scale outdoor scenes, and lacks the ability to efficiently and autonomously generate diverse scenes. At the same time, high-quality simulation scenes should have a wide coverage and be able to autonomously generate marginal scenes that are difficult to obtain in real-world data collection, but the current simulation system has not fully explored the ability to generate extreme scenes. Summary of the Invention

[0004] The purpose of the present invention is to overcome the deficiencies of the existing technology. To achieve the above purpose, an intelligent generation method for an autonomous ability test scenario of ground unmanned equipment is adopted to solve the problems raised in the above background art.

[0005] An intelligent generation method for an autonomous ability test scenario of ground unmanned equipment includes the following steps: S1. Construct a multi-level simulation scene terrain model through noise-based height map generation, terrain erosion simulation, geomorphic parameterized modeling and terrain fusion, and distal scene modeling; S2. Then, conduct high-fidelity and complex scene natural environment modeling based on the generation of dynamic lighting and weather conditions, terrain coverage primitives, unstructured roads, and the generation of static simulation elements of scene vegetation; S3. Combine the intelligent agent behavior model with dynamic trigger rules to automatically generate test scenario cases for the autonomous reconnaissance and maneuvering abilities of unmanned equipment.

[0006] As a further solution of the present invention: The specific steps in S1 include: S11. Perform multiple Perlin noise calculations using different frequency and amplitude parameters through multi-layer fractal noise, and then superimpose the calculation results to generate a primary terrain height map; S12. Based on the obtained primary terrain height map, introduce the progressive gradient water erosion algorithm to simulate the erosion of the terrain by water and generate the secondary terrain height map; S13. Extract the height distribution from the geomorphic data of real events, introduce parameters such as mountain height, size, overall inclination, surface attenuation degree, and edge attenuation degree to parameterize the geomorphology, and obtain the parameterized geomorphology; Then introduce the prominence, saddle, and isolation degree parameters, and fuse the generated secondary terrain height map and the parameterized geomorphology to generate the tertiary terrain height map of the basic scene; S14. Import the generated tertiary terrain height map of the basic scene into the terrain module of UE5 to generate the core terrain model of the simulation scene, and with the core terrain model as the center, expand and add the distal fourth-level scene that does not participate in the actual physical simulation, and finally form the terrain model of the simulation scene.

[0007] As a further solution of the present invention: The specific steps in S11 include: Adopt a layered method, that is, use different frequency and amplitude parameters to perform multiple calculations of Perlin noise, and then combine the results of multiple noises; The Perlin noise with fractal geometric characteristics generated after multiple superpositions, and then according to the limitations of the terrain data for a specific terrain, simulate a three-dimensional realistic terrain.

[0008] As a further solution of the present invention: The specific steps in S12 include: During the construction of the secondary terrain height map, it is composed of multiple water erosion iterations, and the original terrain is gradually modified through erosion and deposition; Use key parameters such as erosion rate, deposition rate, and evaporation rate to determine the water flow path and terrain changes.

[0009] As a further solution of the present invention: The specific steps in S2 include: S21. Based on the graphics engine, construct a diverse lighting and sky system, as well as a high-fidelity simulation of various weather conditions; and classify after setting the control parameters of each natural environment; S22. For the types of field operation scenarios, based on the pre-constructed digital asset terrain material library, automatically assign the corresponding scene terrain materials to the terrain on the terrain model of the simulation scene constructed in step S1; For the generation of field scene vegetation, combine the slope and altitude parameters of the terrain of the simulation scene constructed in step S1 to dynamically adjust the weights, control the distribution of vegetation aggregation and sparse areas, and set multiple levels of vegetation ecological levels. Use road networks, etc. as non-generation area constraints, and use the procedural content generation module of UE5 to automatically construct the scene vegetation distribution; S23. Generate unstructured roads for the scenario. Based on the geometric structure of the simulation scenario terrain model constructed in step S1, automate the construction of the scenario road network by integrating the randomization of path point parameters and the constraints of the simulation terrain of the scenario.

[0010] As a further solution of the present invention: The specific steps in S21 include: After setting each natural environment control parameter, classify them into a random parameter group and a level parameter group; Among them, the random parameter group is randomly controlled by a seed random generator; the level parameter group generates the challenge level of the scenario for the autonomous ability control to be tested.

[0011] As a further solution of the present invention: The random parameter group of illumination includes cloud type, cloud height, and sun type; The level parameter group of illumination includes cloud cover and sun intensity; The random parameter group of weather conditions includes rain, snow type, dust type, and wind direction and intensity; The level parameter group of weather conditions includes the intensity of rain, snow, haze, and dust.

[0012] As a further solution of the present invention: The specific steps of the challenge level are to normalize the level parameters of illumination conditions and weather conditions and divide them into multiple challenge levels.

[0013] As a further solution of the present invention: The specific steps in S3 include: S31. According to the constructed simulation scenario model, configure the unmanned equipment simulation scenario model based on the task type of the actual ground unmanned equipment. S32. For the autonomous reconnaissance ability test of ground unmanned equipment, construct dynamic task events for three scenarios: randomization of reconnaissance target positions, short-term exposure of reconnaissance targets, and dynamic mixing of friendly and enemy forces. Define and automatically generate the corresponding terrain model and natural environment style for each autonomous reconnaissance task scenario, as well as the models and actions of hostile target agents in each reconnaissance scenario. S33. For the autonomous maneuverability test of ground unmanned equipment, construct dynamic task events for three scenarios: unmanned vehicle logistics supply and transportation, unmanned vehicle rapid maneuver rescue, and satellite denial positioning information loss. Define and automatically generate the corresponding terrain model and natural environment style for each autonomous maneuver task scenario, as well as static and dynamic obstacles and threat areas in the maneuver scenario.

[0014] As a further solution of the present invention: The specific steps of constructing the reconnaissance target in S32 include: The construction of reconnaissance targets in the simulation scenario includes the setting of multi-type targets, the dynamic behavior of targets, and the camouflage method, so as to simulate the complex conditions of the real battlefield; The movement modes of the reconnaissance targets are diverse, including three states: static state, slow movement, and fast movement. Among them, the static targets are usually located in the shelters and hidden positions of buildings.

[0015] Compared with the prior art, the present invention has the following technical effects: Adopting the above technical solution, the intelligent generation method of the autonomous ability test scenario gets rid of the dependence on the manual construction of the simulation scenario, greatly reduces the construction cost of the unmanned equipment simulation scenario, and can realize the automatic generation of the test scenario in seconds, significantly improving the construction efficiency of the simulation scenario; Utilize various terrain and landform elements such as plains and mountains, various terrain distributions such as forests and grasslands, various weather environments such as sunny days, rain, snow, and sandstorms, and various scene dynamic events for reconnaissance tasks and maneuver tasks generated by autonomous control. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The following will describe in detail the specific embodiments of the present invention with reference to the accompanying drawings: Figure 1 It is a schematic diagram of the steps of the intelligent generation method of the disclosed embodiment of the present application; Figure 2 It is a schematic diagram of the implementation process of the automatic generation of the simulation scenario terrain model of the disclosed embodiment of the present application; Figure 3 It is a schematic diagram of the scene dynamic agent modeling process and typical test scenario types of the disclosed embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0018] Please refer to Figure 1 , in the embodiment of the present invention, an intelligent generation method for the autonomous ability test scenario of a ground unmanned equipment includes the following steps: S1. Construct a multi-level simulation scenario terrain model through noise-based height map generation, terrain erosion simulation, geomorphic parameterization modeling, terrain fusion, and remote scene modeling; In this embodiment, as Figure 2As shown in the figure, it is a schematic diagram of the implementation process for the automated generation of a terrain model for a simulation scenario; through a procedural terrain construction method that combines multi-layered fractal noise and parameterized landforms, the automated generation of a terrain model for a simulation scenario is carried out. This method includes the following steps: constructing a height map based on Perlin noise to generate a primary terrain; using a high-fidelity terrain erosion model to generate a secondary terrain; parametrically modeling typical landforms and fusing them with the secondary terrain to generate a tertiary terrain; fusing with a static mesh terrain model at the far end of the scenario to generate a quaternary final terrain; In the specific implementation steps, the specific steps include: S11. Perform multiple Perlin noise calculations using different frequency and amplitude parameters through multi-layered fractal noise, and then superimpose the calculation results to generate a primary terrain height map; Specifically, if the height map is filled with random values when automatically generating the terrain, the terrain generated in this way will only be random spikes and not resemble a terrain. In this embodiment, Perlin noise is used to generate basic terrain height data, and Perlin noise is constructed by superimposing multiple noise functions with different wavelengths and amplitudes; Although the Perlin noise for generating the basic terrain height map simulates natural noise to a certain extent, it still does not fully exhibit the irregularity of natural noise. In this embodiment, a further layering method is adopted, that is, multiple Perlin noise calculations are performed using different frequency and amplitude parameters, and then the results of multiple noises are combined; The Perlin noise generated through multiple superpositions has strong fractal geometric characteristics, and three-dimensional realistic terrain can be simulated according to the limitations of terrain data for a specific terrain.

[0019] The automated generation of the primary terrain height profile is controlled by frequency and amplitude parameters, thereby simulating different types of terrain profiles. The noise waves of mountainous terrains are simulated by noise waves with larger amplitudes and smaller frequencies, and those with larger frequencies and smaller amplitudes are used to simulate flat terrains; The degree to which the noise functions are superimposed is controlled by the parameter octave. During the automated generation of the primary terrain height map, the larger the octave, the more detailed and refined the terrain details simulated, and the greater the computational load; S12. According to the obtained primary terrain height map, introduce a progressive gradient water erosion algorithm to simulate the erosion of the terrain by water and generate a secondary terrain height map; Specifically, in order to further simulate the impact of the erosion and deposition processes of natural factors such as water flow on terrain changes in the real world, a progressive gradient water erosion algorithm is further introduced in this embodiment to simulate the flow of water droplets on the terrain surface and their erosion of the terrain, obtaining a more natural and diverse secondary terrain height map.

[0020] The construction process of the secondary terrain height map consists of multiple water erosion iterations. Through erosion and deposition, the original terrain is gradually modified to make it more refined and realistic. Key parameters such as erosion rate, deposition rate, and evaporation rate determine the path of water flow and the changes in the terrain. The key parameters such as erosion rate, deposition rate, and evaporation rate are encapsulated in an interface, and the simulation intensity of terrain erosion can be controlled and generated by a random number seed through the interface. S13. Extract the height distribution from the geomorphic data of real events, introduce parameters such as mountain height, size, overall inclination, surface attenuation degree, and edge attenuation degree to parameterize the geomorphology, and obtain the parameterized geomorphology. Then, introduce prominence, saddle, and isolation parameters, and fuse the generated secondary terrain height map and the parameterized geomorphology to generate the tertiary terrain height map of the basic scene. Specifically, the procedural terrain model generation method based on multi-layer fractal noise can realize terrain model generation. However, due to the complexity of the outdoor real world, the difference in regional expression of the scene terrain model constructed by the secondary terrain height map is small, and the overall macro terrain structure is still relatively similar, and it cannot perfectly express the terrain characteristics of the real world.

[0021] Therefore, the present invention innovatively further fuses real terrain data information according to the constructed secondary terrain height map, and performs parametric modeling on the real terrain data to generate more real and complex and diverse terrains and landforms.

[0022] This embodiment takes the mountain terrain as an example to show the process of real terrain parameterization.

[0023] From the publicly available real mountain terrain DEM database, different mountain terrains are parameterized by linear interpolation to form various parameterized mountain templates. Each type of mountain template is controlled by parameters such as mountain height, size, overall inclination, surface attenuation degree, and edge attenuation degree. During the automatic generation of the mountain, when a certain mountain type is selected from the template library, the appearance and erosion degree of the mountain are controlled by this parameter group. Fuse the generated parameterized geomorphology height map with the secondary terrain height map, and use prominence, saddle, and isolation parameters to control the relative position of the parameterized geomorphology in the secondary topographic map. During the height fusion process of the overlapping area, the highest value is used as the final value of the generated tertiary terrain height map. S14. Import the generated tertiary terrain height map of the basic scene into the terrain module of UE5 to generate the core terrain model of the simulation scene, and expand and add a distal fourth-level scene that does not participate in the actual physical simulation with the core terrain model as the center to finally form the simulation scene terrain model. Specifically, the generated terrain height map is imported into the simulation engine to generate the final scene terrain model. In this embodiment, taking the UE5 engine as an example, the three-level terrain height map is imported into the "landscape" mode of the engine, and partitioned and segmented to generate the terrain model of the simulation scene.

[0024] The combat scene of ground unmanned equipment is a typical large-scale outdoor scene. Due to the limitation of computer memory, if the terrain at the extremely far end that does not participate in physical simulation interaction is included in the three-level terrain, it will cause a decrease in the simulation operation efficiency.

[0025] To solve this problem, in this embodiment, the generated three-level terrain is used as the core simulation scene, and a preset terrain model is extended outward. The preset terrain model exists in the form of a fixed static mesh; the core scene generated by the three-level terrain supports the automatic generation of subsequent scene elements and the simulation test tasks of ground unmanned equipment. The extended scene does not participate in the physical interaction of ground unmanned equipment and only exists for visual rendering.

[0026] The externally extended terrain model can be extended in multiple levels, and each level of the externally extended terrain adjusts the LOD (Level of Detail) level according to the specific core simulation scene. Usually, 2-3 levels of LOD can be set for the externally extended terrain. For example, 70% of the model mesh faces are reserved for the middle distance in the first level, and it is reduced to 30% for the long distance in the second level. In this way, when the terrain details are reduced in the distance, the GPU load can be significantly reduced; In terms of the material of the externally extended terrain static mesh, try to reuse the material instance of the main terrain as much as possible, or create a simplified version of the material, such as removing the parallax map and reducing texture sampling to reduce the shader complexity.

[0027] For the externally extended terrain that does not require interaction, directly disable the collision in the static mesh properties; and set an impassable fence wall at the edge of the core terrain scene to prevent the passage of ground unmanned equipment. Since there may be drones participating in the simulation task in the simulation, for example, the combat space of a fixed-wing drone may cover the entire map and does not need to interact with the ground terrain, so no fence wall is set to block the drone; S2. Then, based on the generation of dynamic lighting and weather conditions, terrain coverage primitives, unstructured roads, and the generation of static simulation elements of scene vegetation, a highly realistic and complex scene natural environment is modeled; In this embodiment, after the terrain model of the simulation scene for ground unmanned equipment is generated, a highly realistic scene natural environment is further constructed based on the scene terrain. For the autonomous capabilities to be tested of ground unmanned equipment, static scene simulation elements such as scene lighting and weather, and environmental primitives are automatically generated in the early scene; The specific steps in S2 include: S21. Based on the graphics engine, construct diverse lighting and sky systems, as well as high-fidelity simulations of various weather conditions; And classify the natural environment control parameters into a random parameter group and a level parameter group; Among them, the random parameter group is randomly controlled by a seed random generator; the level parameter group generates the challenge level of the scene for the autonomous ability to be tested. Specifically, the generation of lighting and weather is as follows: The autonomous reconnaissance ability test of ground unmanned equipment requires the construction of diverse lighting and weather conditions to generate sensor data such as cameras and radars in different environments, and conduct comprehensive robustness tests on algorithm models such as target recognition and target tracking.

[0028] In this implementation example, based on the UE5 engine, which is the fifth-generation game engine developed, the natural environment system of the ground unmanned equipment autonomous ability test scenario is designed to provide rich and diverse natural environment conditions for the autonomous reconnaissance ability test.

[0029] The generation of lighting first involves the realization of sky effects. In this implementation example, based on the graphics rendering system, the generation and control of natural elements such as clouds and the sun light source are realized.

[0030] In this implementation example, the BP_Sky_Sphere blueprint integrated in UE5 and the Ultra Dynamic Sky plug-in are used for design to accelerate the development process of the natural environment system.

[0031] Classify the parameters of each sky effect into a random parameter group and a level parameter group. The parameter types of the random parameter group have little impact on the autonomous reconnaissance ability of unmanned equipment, and the level parameter group will directly affect the challenge level of the autonomous reconnaissance ability test scenario.

[0032] The random parameter group of lighting conditions includes: cloud types (static clouds, 3D volumes, 2D dynamic clouds, etc.) and cloud heights, sun types (sun height, sun size), etc.; The level parameter group of lighting conditions includes: cloud cover (the higher the value, the higher the challenge level), sun intensity (the lower the value, the higher the challenge level), etc.; The generation of weather is designed using the Ultra Dynamic Weather plug-in of UE5 in this implementation example, which can greatly simplify the implementation of various weather environments. In this implementation example, the relevant parameter interfaces are extracted, specifically including: rain, snow, haze, and dust.

[0033] Classify the parameters of each weather effect into a random parameter group and a level parameter group. The parameter types of the random parameter group have little impact on the autonomous reconnaissance ability of unmanned equipment, and the level parameter group will directly affect the challenge level of the autonomous reconnaissance ability test scenario.

[0034] The random parameter groups of weather conditions include: rain, snow type (particle size, material, etc.), dust type (particle size, material, etc.), wind direction and strength, etc.; The weather conditions level parameter group includes: the intensity of rain, snow, haze and dust (the higher the value, the higher the challenging level); To improve rendering efficiency, this implementation dynamically adjusts the complexity and number of particles based on the distance between the camera and the particles. Particles at a distance are simplified into a small amount of geometry or lower-resolution materials to reduce GPU calculations and ensure performance when rendering a large number of particles in the scene at the same time.

[0035] The natural environment of the scene is automatically generated, and the control parameters of lighting and weather may cross-control the same natural condition. This implementation example sets the priority set of the control parameters generated by weather to be higher than the control parameters generated by lighting.

[0036] Normalize the level parameters of lighting conditions and weather conditions and divide them into multiple challenge levels. In the automatic generation process of autonomous reconnaissance capability test scenarios, in addition to directly controlling each parameter, you can also directly input the challenge level and randomly select the value in the corresponding level parameter space; and the control parameters of the random parameter group are randomly generated when they are not specified; S22, for the type of field operation scene, based on the pre-built digital asset terrain material library, automatically assign the corresponding scene terrain material on the terrain on the simulation scene terrain model built in step S1; For the generation of outdoor scene vegetation, the slope and altitude parameters of the simulated scene terrain constructed in step S1 are combined to dynamically adjust the weights, control the clustering and sparse distribution of vegetation, set multiple levels of vegetation ecological levels, use road networks as ungenerable area constraints, and use UE5's procedural content generation module to automatically construct scene vegetation distribution; S23, for the generation of unstructured roads in the scene, based on the geometric structure of the simulated scene terrain model constructed in step S1, the scene road network is automatically constructed by fusing the randomization of path point parameters with the scene simulation terrain constraints; Specifically, the automatic construction of scene static simulation elements is as follows: static simulation elements include terrain coverage primitives, unstructured roads, scene vegetation elements, etc.

[0037] (1) Generation of terrain covering primitives. The terrain covering primitives first set the terrain material. This implementation example uses the preset terrain material to directly assign it based on the scene type. The preset scene terrain materials include four types: forest, grassland, desert, and snow.

[0038] This example takes the terrain coverage primitives of a forest scene as an example. The production of the preset terrain materials relies on the rich data assets in the game industry. Download terrain textures, as well as various digital assets such as grass, moss, and rocks from the Quixel Bridge library, which also includes rich albedo maps, normal maps, distance maps, and other information; more than 20 types of digital assets such as grass and rocks need to be configured, and information such as the transformation range of the set size, appearance probability, and distribution coverage is set to ensure the random diversity generation of terrain coverage primitives. And set the wind and collision trampling effects on the vegetation meshes of the grass to achieve a more realistic physical simulation of the scene.

[0039] (2) Generation of scene vegetation elements. First, construct a digital resource library containing diverse vegetation forms. The digital resources can be obtained from digital asset libraries such as Quixel Bridge, or independently constructed using software such as Speed Tree. Ensure that the forms of these digital resources conform to the natural growth characteristics, and generate multiple form variants and models at different growth stages for the same vegetation type to enhance scene diversity.

[0040] The automated generation of vegetation is developed based on the PCG function of UE5. First, set the non-vegetation-covered areas of the scene, such as the roads in the scene, where no vegetation is generated; the target area uses a random noise algorithm to generate the spatial distribution of vegetation density, and dynamically adjusts the weight in combination with the slope and altitude parameters of the terrain to control the distribution of vegetation aggregation and sparse areas; when the scene span is large enough, divide the scene into multiple vegetation groups, and each biome is associated with specific vegetation combinations and distribution constraint conditions. Further design the ecological dependence rules between vegetation, such as defining the symbiotic or repulsive relationships between specific vegetation, to ensure that the generation results conform to the natural ecological laws.

[0041] Execute the automated layout generation of vegetation in the virtual terrain environment. First, extract terrain feature data, including height, slope, and simulated humidity information, as the basic input for vegetation distribution. Adopt a hierarchical generation strategy, divide the vegetation into multiple categories according to height and ecological levels, such as the grass and rock layer at the bottom layer, the shrub layer in the middle layer, and the forest layer at the high layer, to avoid visual conflicts and optimize calculation efficiency. For each vegetation level, batch generate vegetation instances in the area that meets the ecological constraint conditions according to the distribution rules predefined by PCG.

[0042] (3) Generation of unstructured roads. The wild scene contains a large number of unstructured roads. This solution uses the Spline Mesh component of UE5 to generate the road structure based on the road network structure, and preset various material textures of the Spline Mesh, such as dirt roads, muddy roads, snow roads, etc., and then adapt to diverse scene types; The present invention realizes the automatic construction of the scene road network by integrating parameter randomization configuration and terrain constraint analysis. The system supports two modes: manual setting and random generation. In the manual mode, it receives the specified starting point, ending point, width, and curvature range of the road from the user; In the random mode, it automatically generates road parameters within the specified geographical range. During the road planning stage, the system projects the control points onto the DEM elevation map for terrain feature analysis. First, it identifies the elevation mutation areas of adjacent grids. If they exceed the preset threshold, they are marked as impassable areas. At the same time, it calculates the terrain slope field based on the three-dimensional coordinate difference and marks the impassable areas where the slope exceeds the engineering limit; The generation of the road network path adopts an improved A* algorithm. On the basis of the traditional path length cost, it introduces a terrain penalty mechanism. The algorithm evaluates the terrain slope, elevation change, and path curvature of the candidate path points, generates a terrain fitness index through weighted superposition, and dynamically adjusts the path search direction.

[0043] During the path extension process, it synchronously checks the slope compliance and curvature continuity to ensure that the generated path conforms to the real-world characteristics. Finally, it smooths the path through a B-spline curve to obtain a single path trajectory; Within the passable areas of the scene terrain, starting points are randomly scattered within a certain distance range to generate path trajectories, and finally a three-dimensional road network model with both random characteristics and authenticity is output.

[0044] The automatic generation of the natural environment of the simulation scene enables the scene model designers not to care about the implementation details of the underlying technology. While quickly generating a variety of massive test scenes, it also greatly reduces the development cost of the test scenes; S3. Combining the intelligent agent behavior model with the dynamic trigger rule, automatically generate test scenario cases for the autonomous reconnaissance and maneuverability of unmanned equipment. The specific steps include: As Figure 3 shown, the figure shows the schematic diagram of the scene dynamic intelligent agent modeling process and typical test scenario types; S31. According to the constructed simulation scene model, based on the task type of the actual ground unmanned equipment, configure the unmanned equipment simulation scene model; In this embodiment, the configuration of the unmanned equipment simulation scene model: There are many task types for ground unmanned equipment. In this implementation example, taking the construction of a ground unmanned vehicle intelligent agent as an example, the vehicle motion simulation uses the vehicle module to configure the simulation equipment model, including the parameters of each subsystem assembly, the digital model and material texture map of the equipment, the dynamic animation blueprint of the simulation equipment, as well as the motion control mode and path tracking mode of the unmanned equipment simulation model, enabling the unmanned equipment to automatically control the simulation motion according to the planned trajectory.

[0045] Regarding the autonomous reconnaissance ability of unmanned equipment, in the simulation equipment configuration, attention should be focused on the simulation sensor modeling of unmanned equipment. In this implementation example, an optical sensor simulation model of unmanned equipment is configured based on the SceneCapture2D camera component of UE5. First, based on the actual sensor position distribution of each equipment, the SceneCapture2D camera component is configured to the same position and parameters such as the field of view angle are adjusted to be consistent; a RenderTarget class is created in UE, and parameters such as the size and color mode of the generated image are configured, and each RenderTarget is assigned to the corresponding simulation camera. Finally, the RGB values obtained by SceneCapture2D through the image rendering pipeline are synthesized into a visible light camera image, and the obtained depth D value will be used as the output image of the depth camera; for the infrared camera, the infrared Post Process material is assigned to the simulation camera, and different Post Process material effects are set for the reconnaissance target. The post-processed scene image presents an infrared effect and is output to the corresponding RenderTarget.

[0046] Regarding the autonomous maneuverability of unmanned equipment, in the simulation equipment configuration, attention should be focused on the motion performance of unmanned equipment. In this implementation example, the motion control of the simulation equipment is constructed based on the ChaosVehicle plug-in of UE5. First, based on the actual overall parameters of the actual equipment, parameters such as the vehicle mass, center of mass position, and height of the simulation model are configured; then, for the vehicle power train, transmission train, braking train, steering train, suspension train, etc., parameters such as the power speed-torque curve, transmission method and transmission ratio, braking acceleration, wheel steering ratio, and suspension stiffness of the simulation model are configured to make the simulation model simulate the motion state of the real vehicle with high fidelity as much as possible.

[0047] S32. For the autonomous reconnaissance ability test of ground unmanned equipment, three dynamic task events of randomization of reconnaissance target positions, short-term exposure of reconnaissance targets, and dynamic mixing of friendly and enemy forces are constructed; Define the styles of automatically generating corresponding terrain models and natural environments for each autonomous reconnaissance task scenario, as well as the models and actions of hostile target agents in each reconnaissance scenario; In this embodiment, the basic environment is set. The construction of reconnaissance targets in the simulation scenario includes multi-type target settings, target dynamic behaviors, and camouflage methods to simulate the complex conditions of the real battlefield. The movement methods of reconnaissance targets are diverse, covering three states: stationary, slow moving, and fast moving. Stationary targets are usually located in hidden positions such as bunkers and buildings, increasing the difficulty of discovery; slow-moving targets simulate patrol teams or standby vehicles, while fast-moving targets test the real-time reaction ability of the recognition system. Especially in the case of short-term exposure, the system needs to quickly identify and locate the target.

[0048] Multiple target types are set in the scenario, including enemy infantry, different types of vehicles (such as armored vehicles, transport vehicles, tactical command vehicles), and low-flying drones, etc. The corresponding simulation models are pre-stored in the simulation model library and automatically called based on the type of dynamic events. Multiple coatings and camouflages are configured for vehicles of the same type. For example, during the generation of the jungle scenario, the vehicle to be reconnoitered uses military green coating to increase the challenge of the test scenario.

[0049] Taking three scenario tasks of random reconnaissance target location, short-term exposure of reconnaissance targets, and dynamic mixing of friendly and enemy forces as examples, this implementation example demonstrates the automated generation method of specific task scenarios in the autonomous reconnaissance ability test of unmanned equipment, so as to evaluate the recognition effect and task completion efficiency of ground unmanned equipment in diverse scenarios.

[0050] (1) Randomization of reconnaissance target locations. The ground unmanned equipment needs to detect the targets in the area under unknown target distribution conditions, and assess the area search ability of the unmanned equipment, especially the recognition ability under limited vision in complex natural environments.

[0051] In the test scenario, the initial positions of all enemy targets are randomly distributed in different terrain areas, and the distribution positions are regenerated each time for testing, simulating the unpredictability of target distribution in the real battlefield.

[0052] Automatically create plains and various mountain terrains based on the step S1, and assign texture maps to the terrains; In addition to the differences in terrain features, for mountain terrains, further based on the automated construction of the scene static simulation elements in the step S22, generate various types of scene vegetation distribution elements to form forest scenes, shrubbery scenes, grassland scenes, barren mountain scenes, etc.; automatically construct scene roads in the road network generation on the plain terrain and assign road materials such as dirt roads; load the preset forest vegetation on both sides of the scene road network based on the vegetation generation in the step S22, and the distribution of scene static elements in the remaining areas is randomly controlled by the system to generate; Based on the step S21, generate lighting and weather. The scene conditions controlled by the random parameter group are controlled by the random number generator, and the scene conditions controlled by the level parameter group are randomly controlled within the specified scene challenge level. The scenes with high challenge levels include Among them, the parameters of each scene can also be adjusted manually; The challenge level of the test scenario is controlled by the level parameter group during the generation of the natural environment on the one hand, and determined by the location of the target to be reconnoitered on the other hand. Set the area where the reconnaissance target appears and control the probability of the target appearing in each area.

[0053] For low - level challenging target appearance areas, such as non - forest - covered areas in plain terrain, a random number controller is used to control the generation of reconnaissance targets; high - level challenging target appearance areas include areas with shrubs for lurking and hiding, rocks, and semi - occlusion by woods, etc. Control the appearance frequency of targets in the target appearance area with different probability densities, and randomly control the positions and quantities of targets to generate reconnaissance targets. Furthermore, according to the materials of the test scene, set the camouflage of the reconnaissance targets. For high - challenging levels, such as in grassland and forest scenes, the reconnaissance targets use dark green woodland camouflage; in adjacent and desert scenes, the reconnaissance targets use desert spot camouflage; in snowfield scenes, the reconnaissance targets use snow camouflage, etc.

[0054] During the test, there is no restriction on the movement trajectory of the unmanned equipment. It moves freely in the whole scene, and the target discovery rate and search time are used to test and evaluate the autonomous reconnaissance ability of the unmanned equipment.

[0055] (2)Reconnaissance targets are briefly exposed. Set specific targets to be briefly exposed in the field of vision at random positions, such as the target only briefly appears behind shelters such as rocks and trees and then quickly hides. This task tests the recognition and reaction ability of the system within limited information and time.

[0056] Test evaluation indicators: recognition reaction time, short - time recognition accuracy rate, and capture success rate during the exposure period.

[0057] Automatically create a scene terrain model based on step S1, and automatically assign terrain materials based on step S22. The fixed position of the ground unmanned equipment to be tested remains unchanged, maintaining a fixed reconnaissance area. Through the visible area of the reconnaissance camera sensor of the ground unmanned equipment, it is quasi - projected into the test scene based on the sensor space transformation matrix, so as to obtain the generation area of the target to be tested. In grassland and low - shrub scenes, the target to be reconnoitered performs actions from lying down to standing up; in forest scenes, the target to be reconnoitered hides, runs, and hides, such as in thick forests; in mountain scenes, the target to be reconnoitered runs and hides behind shelters such as rocks. Based on this target generation rule, use the following strategy to automatically generate targets: In grassland and low - shrub scenes, targets are randomly generated in the generation area determined by the reconnaissance camera. For targets in forest scenes, first, according to the scene generated in S22, obtain the positions where vegetation appears, and use a certain density value as the threshold to further divide the thick forest area. In the generation area determined by the reconnaissance camera, the targets are randomly generated at a short distance outside the divided forest area and quickly enter the woods scene to hide. In the mountain scenario, the target first obtains the positions of the shelters in the scene according to the scene generated in S22, and extracts the shelters with a total size larger than the target size as the hiding shelters for the target. The target is finally generated at the rear side relative to the position of the reconnaissance camera in the area determined by the camera. When the test task is executed, the target quickly hides behind the shelter.

[0058] (3) Dynamic mixing of friendly and enemy forces. Arrange friendly and enemy targets in the test scene to simulate a complex combat environment, so as to test the system's ability to identify friend or foe and avoid friendly fire.

[0059] Test evaluation indicators: accuracy rate of identifying friend or foe, recognition delay, misrecognition rate.

[0060] Automatically create a scene terrain model based on step S1, and automatically assign terrain materials based on step S22; The fixed position of the unmanned equipment to be tested remains unchanged, maintaining a fixed reconnaissance area; Through the visible area of the reconnaissance camera sensor of the unmanned equipment, it is quasi-projected into the test scene based on the sensor space transformation matrix, so as to obtain the generation area of the target to be tested; Arrange friendly and enemy targets with different camouflages and coatings in the same scene area, control the appearance frequency of friendly and enemy targets in the target generation area with different probability densities, and randomly control the target positions and quantities to generate reconnaissance targets; All targets are created as automatically moving agents by the Mass AI of UE5, and the movement trajectories of the agents have certain intersections and contacts, simulating the scene of friendly and enemy forces dynamically interlacing in the area in reality.

[0061] S33. For the autonomous mobility ability test of ground unmanned equipment, construct dynamic task events for three scenarios: unmanned vehicle logistics supply and transportation, unmanned vehicle rapid mobility rescue, and satellite denial positioning information loss; Define the styles of automatically generating corresponding terrain models and natural environments for each autonomous mobility task scenario, as well as static and dynamic obstacles and threat areas in the mobility scenario.

[0062] In this embodiment, the basic environment is set. The autonomous mobility ability test scenario will focus on the construction of terrain conditions, covering various terrain types, terrain parameters, static and dynamic obstacles, impassable threat areas, and highly dynamic randomized emergencies and other factors.

[0063] The terrain types cover various natural environments such as plains, mountains, forests, and grasslands, making the test scenarios have higher simulation authenticity. The plain provides a broad movement space, suitable for basic path planning and acceleration testing; the mountain contains steep slopes and altitude differences, simulating the difficulty of climbing and descending in complex terrains; the forest terrain has the characteristics of vision occlusion and path interference, challenging the vehicle's obstacle avoidance ability in vegetation-dense areas; the grassland combines the openness of the plain and the local complexity of the forest, testing the vision expansion and recognition effects under different speed and distance conditions. This invention patent can directly control multiple parameters of the simulation terrain, including mountain height, altitude change, vegetation coverage, etc., making the scene more layered. For example, the mountain height and altitude change create a real undulating terrain through steep slopes and height differences, increasing the challenges for vehicle path planning and autonomous driving.

[0064] The automatic maneuverability test scenario sets up static obstacles such as abandoned military vehicles, sandbags, and cheval de frise to simulate common obstacles in combat scenarios. Large obstacles such as abandoned vehicles increase path obstacles, requiring the vehicle to be able to identify and adopt appropriate detour strategies; the settings of sandbags and cheval de frise can not only block the path but also be used to test the detour and obstacle avoidance capabilities of the vehicle path planning system. In the part of dynamic obstacles, the scenario includes vehicles that suddenly change lanes and randomly appearing pedestrians, etc. These dynamic elements add uncertainty to the test, making the intelligent land combat platform need to respond quickly and make avoidance decisions.

[0065] The threat areas include minefields, static or dynamic enemy radar areas, etc. These non-passable high-risk areas require unmanned vehicles to identify and detour. The minefield is set as an absolutely non-passable area, while the radar area simulates the enemy's reconnaissance threat according to the dynamic or static state, increasing the risk avoidance requirements in path planning. In addition, the dynamically unpredictable environmental settings introduce scenarios of sudden bridge or road damage, making ground unmanned equipment face sudden path interruptions and must quickly make emergency responses and re-plan the path. These random changes in the environment significantly enhance the complexity of the simulation scenario, thus comprehensively testing the intelligence and robustness of the unmanned equipment's autonomous maneuverability in emergency situations.

[0066] This implementation example is based on environmental conditions. Taking three scenario tasks of logistics supply transportation, rapid maneuver rescue, and satellite denial positioning information loss as examples, it demonstrates the automatic generation method of specific task scenarios in the test of unmanned equipment's autonomous maneuverability, so as to evaluate the autonomous maneuver planning ability and task completion efficiency of ground unmanned equipment in diverse scenarios.

[0067] (1) Logistics supply transportation by ground unmanned vehicles. Ground unmanned vehicles need to transport logistics supply materials in diverse terrain environments to ensure the efficient completion of the supply transportation task within the set time.

[0068] Test evaluation metrics: the planning time from task reception to path generation, the obstacle avoidance response time to dynamic obstacles, and the success rate of successfully completing material transportation within the specified time.

[0069] Automatically create plains and various mountain terrains based on step S1, and automatically assign terrain materials. Automatically generate scene weather conditions based on step S21, with good weather such as sunny days as the main condition. Based on step S22, first automatically create various primitives in the scene, and then automatically generate various vegetation elements in the scene; in this task mode, further generate an unstructured road network, automatically create a driving path for the ground unmanned equipment based on the road network, and set the starting and ending points of the unmanned equipment movement. Randomly distribute static obstacles (such as abandoned vehicles, sandbags) around the driving path, control the appearance frequency of obstacles in the set area with different probability densities, and randomly control the positions and quantities of obstacles. The generation of dynamic obstacles uses a state machine model to simulate the appearance of environmental vehicles and pedestrians within a specific time, and set conditional triggers to randomly generate dynamic obstacles to increase the challenge of the autonomous mobile logistics supply and transportation scene test.

[0070] (2) Rapid maneuver rescue of ground unmanned vehicles. The ground unmanned vehicle needs to reach the target position within a short time to complete the rescue, and evaluate its emergency response ability in a complex threat environment.

[0071] Test evaluation metrics: the response time from receiving the task to reaching the target, the accurate avoidance rate of the threat area, and the time to regenerate the path in case of sudden obstacles.

[0072] Automatically create plains and various mountain terrains based on step S1, and automatically assign terrain materials. Automatically generate scene weather conditions based on step S21, with good weather such as sunny days as the main condition. Based on step S22, first automatically create various primitives in the scene, and then automatically generate various vegetation elements in the scene; in this task mode, it can be randomly or manually preset to decide whether to use the generated unstructured road network. Use simulated threat areas (such as minefields, enemy radar areas) in the scene, randomly generate the positions and shapes of the threat areas, and send them to the autonomous mobile module of the unmanned equipment to be tested; and randomly distribute and generate static obstacles (such as abandoned vehicles, sandbags) in the scene, control the appearance frequency of obstacles in the set area with different probability densities, and randomly control the positions and quantities of static obstacles. The generation state machine model of dynamic events. When the unmanned ground equipment travels to a specific area, a conditional trigger is set to randomly generate impassable areas in front, simulating that the original passable area becomes impassable due to enemy attacks, and examining the replanning ability of the unmanned equipment's autonomous maneuverability.

[0073] (3) Loss of satellite-denied positioning information. The unmanned ground vehicle needs to use real-time mapping and positioning algorithms to successfully position in the event of a sudden loss of satellite positioning signals, and evaluate its relocalization ability in the case of electromagnetic interference causing satellite positioning interruption.

[0074] Test evaluation index: The accuracy of real-time relocalization of the unmanned ground equipment after signal loss.

[0075] Automatically create plains and various mountain terrains based on step S1, and automatically assign terrain materials; Automatically generate scene weather conditions based on step S21, with good weather such as sunny days as the main conditions; Based on step S22, first automatically create various primitives in the scene, and then automatically generate various vegetation elements in the scene; in this task mode, further generate an unstructured road network, automatically create a driving path for the unmanned ground equipment based on the road network, and connect to the autonomous path tracking control module of the unmanned equipment under test; During the path tracking of the unmanned ground equipment, control the position and duration of the lost positioning information with different probability densities. During the time of information loss, connect to the real-time mapping and positioning algorithm of the maneuverability under test to obtain the relocalization information of the unmanned vehicle. In the present invention, the signal loss in the autonomous maneuverability test scenario is achieved by cutting off the GetActorTransform function in UE5.

[0076] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents, and all should be included within the protection scope of the present invention.

Claims

1. An intelligent generation method for the autonomous ability test scenario of ground unmanned equipment, characterized in that It includes the following steps: S1. Construct a multi-level simulation scene terrain model through height map generation based on noise, terrain erosion simulation, geomorphic parameterization modeling, terrain fusion, and far-end scene modeling; S2. Then, based on the generation of dynamic lighting and weather conditions, the generation of terrain coverage primitives, unstructured roads, and static simulation elements of scene vegetation, conduct high-fidelity and complex scene natural environment modeling; S3. Combine the agent behavior model with dynamic trigger rules to automatically generate test scenario cases for the autonomous reconnaissance and maneuverability of unmanned equipment.

2. The intelligent generation method of an autonomous ability test scenario for ground unmanned equipment according to claim 1, wherein, The specific steps in S1 include: S11. Perform multiple Perlin noise calculations using different frequency and amplitude parameters through multi-layer fractal noise, and then superimpose the calculation results to generate a primary terrain height map; S12. According to the obtained primary terrain height map, introduce a progressive gradient water erosion algorithm to simulate the erosion of the terrain by water and generate a secondary terrain height map; S13. Extract the height distribution from the geomorphic data of real events, introduce parameters such as mountain height, size, overall inclination, surface attenuation, and edge attenuation to parameterize the geomorphology, and obtain the parameterized geomorphology; Then introduce prominence, saddle, and isolation parameters to fuse the generated secondary terrain height map and the parameterized geomorphology to generate a tertiary terrain height map of the basic scene; S14. Import the generated tertiary terrain height map of the basic scene into the terrain module of UE5 to generate the core terrain model of the simulation scene, and centered on the core terrain model, expand and add a far-end quaternary scene that does not participate in actual physical simulation, and finally form a simulation scene terrain model.

3. The intelligent generation method of an autonomous ability test scenario for ground unmanned equipment according to claim 2, wherein, The specific steps in S11 include: Adopt a hierarchical method, that is, perform multiple Perlin noise calculations using different frequency and amplitude parameters, and then combine the results of multiple noises; The Perlin noise with fractal geometry characteristics generated after multiple superpositions, and then according to the limitations of terrain data for a specific terrain, simulate a three-dimensional realistic terrain.

4. The intelligent generation method of an autonomous ability test scenario for ground unmanned equipment according to claim 2, wherein The specific steps in S12 include: During the construction of the secondary terrain height map, it is composed of multiple water erosion iterations, and the original terrain is gradually modified through erosion and deposition; Use key parameters such as erosion rate, deposition rate, and evaporation rate to determine the water flow path and terrain changes.

5. The intelligent generation method of an autonomous ability test scenario for ground unmanned equipment according to claim 1, characterized in that, The specific steps in S2 include: S21. Based on the graphics engine, construct a diverse lighting and sky system, as well as a high-fidelity simulation of various weather conditions; and classify after setting the control parameters of each natural environment; S22. For the type of field operation scenario, based on the pre-constructed digital asset terrain material library, automatically assign the corresponding scene terrain material to the terrain on the simulation scene terrain model constructed in step S1; For the generation of field scene vegetation, combine the slope and altitude parameters of the simulation scene terrain constructed in step S1 to dynamically adjust the weight, control the distribution of vegetation aggregation and sparse areas, set multiple levels of vegetation ecological levels, use road networks, etc. as non-generation area constraints, and use the procedural content generation module of UE5 to automatically construct the scene vegetation distribution; S23. Generate unstructured roads for the scenario. Based on the geometric structure of the simulation scenario terrain model constructed in step S1, automate the construction of the scenario road network by integrating the randomization of path point parameters and the constraints of the scenario simulation terrain.

6. The intelligent generation method of an autonomous ability test scenario for a ground unmanned equipment according to claim 5, characterized in that, The specific steps in S21 include: After setting the control parameters of each natural environment, classify them into a random parameter group and a level parameter group; Among them, the random parameter group is randomly controlled by a seed random generator; the level parameter group generates the challenge level of the scenario for the autonomous ability control to be tested.

7. The intelligent generation method of an autonomous capability test scenario for a ground unmanned equipment according to claim 6, wherein The random parameter group of lighting includes cloud type, cloud height, and sun type; The level parameter group of lighting includes cloud cover and sun intensity; The random parameter group of weather conditions includes rain, snow type, dust storm type, and wind direction and intensity; The level parameter group of weather conditions includes the intensity of rain, snow, haze, and dust storm.

8. The intelligent generation method of an autonomous capability test scenario for ground unmanned equipment according to claim 7, characterized in that, The specific steps of the challenge level are to normalize the level parameters of lighting conditions and weather conditions and divide them into multiple challenge levels.

9. The intelligent generation method for the autonomous ability test scenario of a ground unmanned equipment according to claim 1, characterized in that, The specific steps in S3 include: S31. According to the constructed simulation scenario model, configure the unmanned equipment simulation scenario model based on the task type of the actual ground unmanned equipment; S32. For the autonomous reconnaissance ability test of ground unmanned equipment, construct dynamic task events for three scenarios: randomization of reconnaissance target positions, short-term exposure of reconnaissance targets, and dynamic mixing of friendly and enemy forces; Define and automatically generate the styles of the corresponding terrain model and natural environment for each autonomous reconnaissance task scenario, as well as the models and actions of hostile target agents in each reconnaissance scenario; S33. For the autonomous mobility ability test of ground unmanned equipment, construct dynamic task events for three scenarios: logistical supply transportation of unmanned vehicles, rapid mobility rescue of unmanned vehicles, and loss of satellite denial positioning information; Define and automatically generate the styles of the corresponding terrain model and natural environment for each autonomous mobility task scenario, as well as static and dynamic obstacles and threat areas in the mobility scenario.

10. The intelligent generation method of an autonomous capability test scenario for a ground unmanned equipment according to claim 9, characterized in that, The specific steps of constructing the reconnaissance target in S32 include: The construction of reconnaissance targets in the simulation scenario includes setting multiple types of targets, target dynamic behaviors, and camouflage methods to simulate the complex conditions of a real battlefield; The movement modes of the reconnaissance targets are diverse, including three states: stationary state, slow movement, and fast movement. Among them, stationary targets are usually located in hidden positions such as bunkers and buildings.

Images