Unmanned aerial vehicle racing training method based on virtual-real confrontation mechanism

Abstract

The present application relates to the technical field of unmanned aerial vehicle simulation training, and is an unmanned aerial vehicle racing training method based on virtual and real confrontation mechanism, which specifically comprises: layering analysis on a real multi-layer three-dimensional race track structure to generate a hierarchical space reference set; generating a diving speed compression area, a visual angle size contraction and occlusion state, and a dynamic height compression area to obtain a vertical compression state set; evaluating the height reference collapse intensity and performing aggregated analysis on the space positioning imbalance area in different diving stages; strengthening and correcting the virtual vertical compression intensity for the direction reference of the lower obstacle door to obtain a layered space correction result; regenerating a vertical switching racing scene, associating and correcting the training behavior with the real racing behavior, and outputting the unmanned aerial vehicle three-dimensional racing training result and the real racing space adaptability evaluation result. The present application solves the problem of space positioning imbalance caused by diving layer switching in real racing.

Description

Technical Field

[0001] This invention belongs to the field of drone simulation training, and more specifically, relates to a drone racing training method based on a virtual-real confrontation mechanism. Background Technology

[0002] Existing drone racing simulation training methods often construct the track as a continuous and smooth three-dimensional space, failing to realistically reproduce the spatial reference abrupt changes and perceptual pressure caused by the dive-and-cutting of layers in multi-layered three-dimensional tracks. This results in a serious disconnect between training scenarios and actual competitions. In real three-dimensional racing, when the operator dives from a high level to a lower level obstacle gate, they face multiple pressures, including the rapidly enlarging size of the obstacle gate within their field of vision, the rapid convergence of its boundaries, and the brief obstruction of the lower level entrance by the transition structure. The basis for altitude judgment changes drastically accordingly, easily leading to the phenomenon of incorrectly inheriting the flight direction from the upper level to the lower level crossing stage, directly affecting obstacle gate crossing safety and racing rhythm. However, existing training technologies lack quantitative identification methods for this misinheritance of spatial direction and the resulting altitude reference collapse phenomenon, failing to effectively capture the spatial abrupt changes that occur to the operator at each stage of the dive-and-cutting, mid-level obstruction, and lower level approach. The existing methods suffer from several shortcomings. First, they fail to dynamically aggregate and categorize imbalanced areas. Second, they lack comprehensive modeling of multiple perceptual factors, including dynamic height pressure intensity, hierarchical reference takeover intensity, and visual pressure during the dive, making it difficult to reproduce the height judgment failure caused by reference switching gaps in real racing. Third, at the training and correction level, current technologies fail to correlate and adjust the lower obstacle gate direction reference, dive area spatial relationships, hierarchical reference switching rhythm, and virtual vertical pressure intensity based on the operator's actual spatial offset and collapse intensity. They often use fixed scene parameters, failing to form a closed-loop reinforcement from offset assessment to hierarchical spatial reconstruction. This results in operators, even after completing a traverse in a simulated environment, struggling to effectively transfer spatial adaptability to real 3D racing events, leading to a lack of reliable correlation between training effectiveness and real-world performance. Summary of the Invention

[0003] To address the shortcomings of existing technologies, the present invention aims to overcome the aforementioned deficiencies and propose a drone racing training method based on a virtual-real confrontation mechanism.

[0004] The present invention adopts the following technical solution; The first aspect of this invention discloses a drone racing training method based on a virtual-real adversarial mechanism, as follows: S1: By performing layered analysis on the real multi-layered three-dimensional track structure, we independently construct virtual and real racing spaces for each layer and define its visual reference structure, boundary compression structure and spatial direction benchmark, generating a set of layered spatial references. S2: Based on the hierarchical spatial reference set, the dive velocity compression region, the view size contraction and occlusion state, and the dynamic height compression region are generated sequentially. The process of switching between high-level and low-level references is compressed to obtain the vertical compression state set. S3: Based on the set of vertical compression states, continuously collect the spatial positioning status of the operator during the subduction and shearing process, identify the phenomenon of misinheritance of spatial orientation in high-rise space, evaluate the intensity of height reference collapse, and perform aggregate analysis on the spatial positioning imbalance area in different subduction stages to obtain the spatial offset evaluation results and the set of height reference collapse areas. S4: Based on the spatial offset assessment results and the set of height reference collapse areas, the layered spatial correction results are obtained by strengthening the lower barrier gate direction reference, remapping the spatial relationship of the diving area, adjusting the reference switching rhythm, and correcting the virtual vertical compression intensity. S5: Based on the hierarchical spatial correction results, regenerate the vertical switching racing scenario, comprehensively analyze the operator's various capabilities in continuous dive-cutting, continuously enhance the operator's spatial adaptation status in different levels, and correlate and correct the training behavior with the real racing behavior, outputting the UAV 3D racing training results and the real race spatial adaptation ability evaluation results.

[0005] Preferably, S1 includes: S11: First, taking a real drone racing track as the object, collect the height of each obstacle gate and the historical flight altitude change trajectory of the drone, analyze the height difference between the center of the obstacle gate and the track reference plane, and divide the track into a high-level racing area, a middle-level transition area and a lower-level obstacle crossing area to obtain the track level division results. S12: Based on the results of the track level division, further calibrate the vertical spacing, dive angle range, cut-in and cut-out distances, and turning compression areas between adjacent levels, record the key positions and compression center points of the dive segment, and form the vertical switching calibration results between levels. S13: Based on the vertical switching calibration results between layers, the high-level racing area, the middle-level transition area, and the lower-level obstacle crossing area of ​​the real track are mapped to the virtual training space respectively. At the same time, the compression intensity of the inter-layer switching is evaluated to obtain the result of the construction of the layered virtual and real racing space. S14: Based on the results of constructing a hierarchical virtual-real racing space, a visual reference structure, a boundary compression structure, and a spatial direction benchmark are defined for each level of space. At the same time, the hierarchical spatial reference intensity is quantified, and the visual reference structure, boundary compression structure, spatial direction benchmark, and hierarchical spatial reference intensity are combined into a hierarchical spatial reference set for output.

[0006] Preferably, S2 includes: S21: Based on the hierarchical spatial reference set, extract the vertical height difference between adjacent levels, the position of the dive entrance and exit, the distance of the lower obstacle gate entrance, and the dive flight speed of the UAV. Generate a dive speed compression region in the virtual training space, and evaluate the dive speed compression intensity of the region to form the dive speed compression region result. S22: Based on the results of the dive velocity compression region, the obstacle gate size change, boundary contraction change and hierarchical switching occlusion state in the first view of the UAV are generated synchronously, and the synchronous results of view size contraction and hierarchical occlusion are obtained through segmented control. S23: Based on the results of the dive velocity compression region, the results of the view size shrinkage and the hierarchical occlusion synchronization, the dive velocity compression intensity is extracted, and combined with the dynamic height compression intensity, a dynamic height compression region is established. The range and intensity of the compression region are adjusted to obtain the dynamic height compression region results. The dynamic height compression zone includes: the dive entrance compression zone, the middle layer shielding compression zone, and the lower layer obstacle compression zone; S24: Based on the dynamic height compression region results, the switching process between the upper-level spatial reference and the lower-level spatial reference is compressed, and the hierarchical reference takeover strength is quantified to obtain the hierarchical reference switching compression results. S25: Based on the results of the dive velocity compression region, evaluate the overall vertical compression intensity to quantify the total degree of compression caused by the current vertical slicing scene to the operator's spatial judgment. Combine the results of view size shrinkage and layer occlusion synchronization, dynamic height compression region results and layer reference switching compression results to form a set of vertical compression states, and output them in stages according to the dive entry stage, middle layer occlusion stage and lower layer approach stage.

[0007] Preferably, S3 includes: S31: Based on the set of vertical compression states, the spatial positioning status of the UAV is continuously collected in the dive entry stage, the middle layer obstruction stage, and the lower layer approach stage, and its dive-cutting layer spatial offset density is evaluated to determine whether the operator has experienced spatial positioning instability in the dive-cutting layer stage. Finally, the data is sorted in chronological order to obtain the dive-cutting layer spatial offset acquisition results. S32: Based on the spatial offset acquisition results of the dive-cut layer, compare the consistency between the operator's current flight direction and the main flight direction of the upper-level track and the passage direction of the lower-level obstacle gate to determine whether there is a high-level spatial direction mis-inheritance phenomenon, and evaluate the intensity of high-level spatial direction mis-inheritance to quantify the severity of mis-inheritance and obtain the high-level spatial direction mis-inheritance identification result. S33: Based on the high-altitude spatial orientation error inheritance identification results, the three conditions are comprehensively judged to generate the height reference collapse judgment result, and the height reference collapse intensity is given according to the dive stage. S34: Based on the height reference collapse determination results, the locations where the height reference collapse intensity continuously exceeds the preset threshold are aggregated into candidate imbalance regions according to time, space and stage consistency. Then, the real imbalance regions are filtered out by the spatial positioning imbalance aggregation intensity, and the imbalance region type is marked to obtain the spatial positioning imbalance region aggregation result.

[0008] Preferably, S3 further includes: S35: Based on the aggregation results of spatial positioning imbalance areas, by evaluating the comprehensive evaluation value of spatial offset, generate spatial offset evaluation results for hierarchical spatial benchmark reconstruction and a set of height reference collapse areas, and bind each collapse area to the corresponding hierarchical reference takeover status and the lower-level barrier gate spatial direction benchmark.

[0009] Preferably, S4 includes: S41: Based on the spatial offset assessment results and the set of height reference collapse regions, extract the directional error inheritance intensity, spatial offset comprehensive situation, height reference collapse intensity, and lower-level barrier gate passage direction in the lower-level near-collapse region; comprehensively determine the reinforcement magnitude of the lower-level barrier gate directional reference, obtain the lower-level barrier gate directional reference reinforcement coefficient, and actually perform reinforcement processing in the virtual environment to obtain the lower-level barrier gate directional reference reinforcement result; S42: Based on the enhanced results of the lower barrier gate direction reference, combined with the hierarchical occlusion intensity, the aggregation intensity of the spatial positioning imbalance area, the remaining distance from the UAV to the lower barrier gate, and the entrance safety distance, the remapping intensity of the spatial relationship in the dive area is comprehensively determined; according to the remapping intensity, the dive process is divided into three continuous correction segments: dive entrance, middle layer occlusion, and lower layer approach, and correction processing is performed to obtain the remapping result of the spatial relationship in the dive area; S43: Based on the spatial relationship remapping results of the subduction area, and combined with the evaluation of the hierarchical reference switching rhythm adjustment intensity, the exit rhythm of the high-level spatial reference and the takeover rhythm of the lower-level spatial reference are dynamically adjusted to obtain the hierarchical reference switching rhythm adjustment results. S44: Based on the adjustment results of the hierarchical reference switching rhythm, and combined with the correction intensity of the virtual vertical pressure obtained from the evaluation, the vertical pressure intensity in the virtual racing environment is continuously corrected to obtain the virtual vertical pressure intensity correction result.

[0010] Preferably, S4 further includes: S45: The results of strengthening the lower barrier gate direction reference, remapping the spatial relationship of the diving area, adjusting the level reference switching rhythm, and correcting the virtual vertical pressure intensity are uniformly packaged. At the same time, the comprehensive correction value of the layered space is evaluated to quantify the magnitude of the overall correction and obtain the layered space correction result.

[0011] Preferably, S5 includes: S51: Based on the hierarchical spatial correction results and combined with the benchmark flight speed, the pressure intensity of the reconstructed scene is determined, and multiple parameters in the virtual training environment are reset to obtain the vertical switching racing scene. S52: Based on the reconstructed vertical switching racing scenario, continuously detect the operator's spatial positioning stability, obstacle gate crossing stability, and spatial orientation recovery ability during the process of entering the dive entrance, passing through the middle layer of obstruction, and passing through the lower layer obstacle gate, and determine the continuous dive layer cutting stability coefficient. S53: Based on the continuous subduction shearing stability coefficient, combined with the layer reference pipe strength, height reference collapse strength and high-level spatial direction misinheritance strength, the layered racing adaptability value is comprehensively determined, and the spatial adaptability of the operator in different layered racing environments is continuously strengthened to obtain the layered racing adaptability strengthening result. S54: Based on the results of hierarchical racing adaptation reinforcement, the spatial reference switching behavior formed during training is correlated and corrected with the real racing behavior, and key data of each round of training is recorded. Based on the recorded key data, the operator's adaptability to the real competition space is comprehensively determined, the operator's degree of adaptability to the real competition is quantified, and finally the results of the drone 3D racing training and the evaluation results of the adaptability to the real competition space are formed.

[0012] A second aspect of the present invention discloses a terminal, including a processor and a storage medium; characterized in that: The storage medium is used to store instructions; The processor is configured to operate according to the instructions to execute the steps of the drone racing training method based on the virtual-real confrontation mechanism described in the first aspect.

[0013] The third aspect of the present invention discloses a computer-readable storage medium having a computer program stored thereon, characterized in that, when the program is executed by a processor, it implements the steps of the UAV racing training method based on the virtual-real confrontation mechanism described in the first aspect.

[0014] The beneficial effects of the present invention are as follows: Compared with the prior art, the present invention has the following advantages: 1. This technical solution, through continuous spatial positioning data acquisition in three stages—dive entrance, mid-level obstruction, and lower-level approach—can quantify and capture the directional deviation caused by high-level reference residue, accurately distinguish between normal dive transition and high-level spatial directional misinheritance, and combine dynamic height pressure intensity and hierarchical reference takeover intensity to perform intensity calibration and regional aggregation on height judgment failure caused by reference switching gaps, transforming the originally difficult-to-describe spatial cognitive dissonance process into a measurable and labelable training and diagnostic object; 2. This technical solution establishes a dynamic correction closed loop from spatial offset assessment to layered spatial reference reconstruction. Based on the set of height reference collapse areas and the comprehensive evaluation value of spatial offset, the method can specifically strengthen the lower-level obstacle gate orientation reference, remap the spatial relationship of the diving area, compress the reference switching window, and adjust the virtual vertical pressure intensity. This allows the virtual training environment to be reconstructed from a static track into a three-dimensional confrontation scenario with a real pressure gradient. As a result, the operator can gradually establish a stable inter-layer spatial adaptation state through repeated iterative training and effectively transfer the reference switching behavior learned in training to real competitions, greatly improving their ability to stably pass through obstacles and recover spatial orientation when facing diving layer pressure. Attached Figure Description

[0015] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 This is a flowchart illustrating the drone racing training method based on a virtual-real confrontation mechanism according to the present invention. Detailed Implementation

[0016] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0017] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0018] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0019] Example 1: like Figure 1 As shown in the figure, the UAV racing training method based on the virtual-real adversarial mechanism of this invention includes the following: S1: By performing layered analysis on the real multi-layered three-dimensional track structure, we independently construct virtual and real racing spaces for each layer and define its visual reference structure, boundary compression structure and spatial direction benchmark, generating a set of layered spatial references. Furthermore, S1 includes: S11: First, taking a real drone racing track as the object, collect the height of each obstacle gate and the historical flight altitude change trajectory of the drone, analyze the height difference between the center of the obstacle gate and the track reference plane, and divide the track into a high-level racing area, a middle-level transition area and a lower-level obstacle crossing area to obtain the track level division results. In a preferred embodiment, the height difference between the center of the obstacle gate and the track reference surface is obtained as follows: First, based on the height of the bottom of the obstacle gate and half of the height of the obstacle gate itself, the absolute height of the center point of the obstacle gate in space is determined; then, the obtained absolute height of the center point of the obstacle gate is compared with the height of the current track reference ground, and the difference between the two is obtained, which is the height difference between the center of the obstacle gate and the track reference surface.

[0020] Based on the height difference between the center of the obstacle gate and the track reference plane, combined with the historical flight altitude change trajectory of the drone and the altitude difference judgment value, the track is divided into levels: When the height difference between the center of the obstacle gate and the track reference plane is greater than the height difference judgment value, it means that the center of the obstacle gate is significantly higher than the current track reference plane, and it is classified into the high-level racing area. When the height difference between the center of the obstacle gate and the track reference plane is close to the height difference judgment value (or within a floating range near the reference value), it is classified as the middle transition zone.

[0021] When the height difference between the center of the obstacle gate and the track reference plane is less than the height difference judgment value, it means that the center of the obstacle gate is close to the current track reference plane, and it is classified as the lower obstacle crossing area. It should be noted that the high-level racing area, the middle-level transition area, and the lower-level obstacle crossing area each form an independent hierarchical unit, which are then integrated into the output of the track hierarchy division result. The height difference determination value is set according to the actual distance between the layers of the track. In this embodiment, its value range is set to 0.8 meters to 2.5 meters as the judgment benchmark. It should be noted that if the obstacle gates are densely packed and the distance between track layers is small, the height difference judgment value should be lower; if the distance between indoor three-dimensional track layers is large, the height difference judgment value should be higher.

[0022] S12: Based on the results of the track level division, further calibrate the vertical spacing, dive angle range, cut-in and cut-out distances, and turning compression areas between adjacent levels, record the key positions and compression center points of the dive segment, and form the vertical switching calibration results between levels. It should be noted that for the dive section from the high-level racing area to the lower obstacle crossing area, the position of the drone before entering the dive, the lowest point of the dive, and the position of the lower obstacle gate entrance are recorded; for the middle-level transition area, the position of the drone with the most drastic attitude change is recorded as the compression center point. The dive angle range is determined by evaluating the dive angle. Specifically, the magnitude of the dive angle depends on the vertical height difference between two adjacent track levels and the horizontal projection distance of the UAV from the upper entry point to the lower entry point. Generally speaking, the greater the vertical height difference, the steeper the dive angle; the longer the horizontal projection distance, the gentler the dive angle. Both factors together determine the steepness of the dive, which is then used to assess whether the area is prone to altitude reference collapse risk.

[0023] It should be noted that when the dive angle is below 20 degrees, it is closer to a smooth descent, while when it is above 55 degrees, it is easy to form a near-falling layer, which is not suitable as a regular training range. Therefore, in this embodiment, the area between 20 degrees and 55 degrees is taken as the dive angle range. S13: Based on the vertical switching calibration results between layers, the high-level racing area, the middle-level transition area, and the lower-level obstacle crossing area of ​​the real track are mapped to the virtual training space respectively. At the same time, the compression intensity of the inter-layer switching is evaluated to obtain the result of the construction of the layered virtual and real racing space. It should be noted that during mapping, all tracks are no longer placed into a single continuous coordinate space. Instead, an independent spatial reference is established for each level. The independent spatial reference includes the entrance direction, main flight direction, obstacle gate centerline, boundary compression range, and turning compression range for that level. For the diving layer cutting area, the real inter-layer distance, real diving angle and real obstacle gate entrance offset are preserved in the virtual training space, so that the operator can feel the spatial changes in the real competition during training. In a preferred embodiment, the interlayer switching compression intensity is mainly related to the vertical height difference between adjacent layers, the flight speed of the UAV when entering the dive phase, the horizontal projection distance, and the safe distance to the lower barrier gate entrance. Generally speaking, the greater the vertical height difference and the faster the dive speed, the higher the compression intensity. On the other hand, the longer the horizontal projection distance and the larger the entrance safety distance, the more the pressure can be buffered, thus reducing the compression intensity. The interlayer switching compression intensity reflects the amount of pressure the operator experiences when completing a height reference switch in a short period of time; It should also be noted that this step is used to convert the real 3D track into a trainable virtual-real layered space, so that the virtual training environment is no longer overly smooth, but retains the hierarchical abrupt features of the real competition.

[0024] S14: Based on the results of constructing a hierarchical virtual-real racing space, a visual reference structure, a boundary compression structure, and a spatial direction benchmark are defined for each level of space. At the same time, the hierarchical spatial reference intensity is quantified, and the visual reference structure, boundary compression structure, spatial direction benchmark, and hierarchical spatial reference intensity are combined into a hierarchical spatial reference set for output.

[0025] In a preferred embodiment, the visual reference structure includes: gate frame, pillar boundary, ground texture, top beam, and track markings; the boundary compression structure includes: wall distance, pillar lateral distance, lower level entrance narrowing width, and dive exit passage width; the spatial orientation reference includes: the current level's main flight direction, turning direction, dive entry direction, and gate passage direction; In this embodiment, the hierarchical spatial reference strength is used to determine whether a certain level of space can provide the operator with a stable height judgment basis. The hierarchical spatial reference strength is contributed by three aspects: the clarity of visual reference, the strength of boundary compression, and the stability of spatial orientation reference. Each aspect has its own weight, among which visual reference and spatial orientation reference account for a larger proportion, followed by boundary compression. The final reference strength is the weighted sum of these three factors. The higher the value of the hierarchical spatial reference strength, the more stable the height judgment basis provided by the hierarchical space to the operator. S2: Based on the hierarchical spatial reference set, the dive velocity compression region, the view size contraction and occlusion state, and the dynamic height compression region are generated sequentially. The process of switching between upper and lower references is compressed to obtain the vertical compression state set.

[0026] Furthermore, S2 includes: S21: Based on the hierarchical spatial reference set, extract the vertical height difference between adjacent levels, the position of the dive entrance and exit, the distance of the lower obstacle gate entrance, and the dive flight speed of the UAV. Generate a dive speed compression region in the virtual training space, and evaluate the dive speed compression intensity of the region to form the dive speed compression region result. In a preferred embodiment, the dive speed compression region is determined based on whether the UAV forms a high-speed dive state when switching from a high level to a lower level. For example, when the UAV enters the dive phase with a high flight speed, a short horizontal switching distance, and a large height difference between levels, it means that the operator needs to complete the spatial height reference switching in a short time, and a corresponding spatial compression region is generated in the virtual training space. The aforementioned dive speed compression zone is not simply about reducing the size of the track, but rather about enhancing the approach speed of the obstacle gates, the boundary contraction speed, and the approach intensity of the lower entrance from the drone's perspective, so that the operator can feel the oppressive state of a high-speed collapse in a real competition. For example, in this embodiment, the dive velocity compression intensity is related to the vertical height difference between layers, the flight speed of the UAV when entering the dive phase, the average acceleration during the dive, the horizontal projection distance, and the lower-level entrance safety distance. Generally speaking, the greater the vertical height difference, the faster the flight speed, and the greater the average acceleration, the higher the dive velocity compression intensity. The longer the horizontal projection distance and the greater the entrance safety distance, the lower the dive velocity compression intensity. This dive velocity compression intensity is used to determine the magnitude of the spatial compression area generated in the virtual environment to simulate the sensation of rapid collapse.

[0027] S22: Based on the results of the dive velocity compression region, the obstacle gate size change, boundary contraction change and hierarchical switching occlusion state in the first view of the UAV are generated synchronously, and the synchronous results of view size contraction and hierarchical occlusion are obtained through segmented control. For example, in this embodiment, the specific process of generating the obstacle gate size change is as follows: when the UAV rapidly approaches the lower obstacle gate, the system increases the magnification rate of the obstacle gate in the field of view according to the diving speed compression intensity, so that the lower obstacle gate is quickly transformed from a distant reference object into a close target that must be crossed at the current distance; The magnification rate of the barrier gate in the field of view depends on the actual opening size of the barrier gate, the equivalent focal length distance of the virtual view set according to the training display device, the remaining distance from the drone to the center of the barrier gate, and the entrance safety distance. Generally speaking, the larger the actual opening and the larger the equivalent focal length distance, the higher the magnification rate will be; the shorter the remaining distance, the higher the magnification rate (i.e., the closer it is, the greater the magnification rate); the entrance safety distance plays a buffering role, and the larger the distance, the more gradually the magnification rate increases. This magnification rate makes the operator feel the pressure of the barrier gate rapidly approaching from a distant reference point. In this embodiment, the specific process of generating boundary contraction changes is as follows: while generating the size change of the barrier gate, a visual effect of the boundary converging towards the center is generated based on the positional relationship of the left and right boundaries, the top beam, the lower entrance wall and the middle transition column; The specific process of generating the hierarchical switching occlusion state is as follows: When the high-level platform, transition beam or middle-level obstacle structure occludes the lower-level entrance, it forms a short-term occlusion on the lower-level obstacle door, so that the operator cannot obtain a complete lower-level space reference in advance during the dive. It should be noted that in this embodiment, the segmented control is specifically a segmented threshold control method, as follows: when the remaining dive distance is greater than ten meters, weak contraction is the main method; when the remaining dive distance is between three and ten meters, the size change of the obstacle gate is enhanced; when the remaining dive distance is less than three meters, the focus is on generating boundary compression and entrance obstruction.

[0028] S23: Based on the results of the dive velocity compression region, the results of the view size shrinkage and the hierarchical occlusion synchronization, the dive velocity compression intensity is extracted, and combined with the dynamic height compression intensity, a dynamic height compression region is established. The range and intensity of the compression region are adjusted to obtain the dynamic height compression region results. The dynamic height compression zone includes: the dive entrance compression zone, the middle layer shielding compression zone, and the lower layer obstacle compression zone; It should be noted that, in this embodiment, the dive entrance compression zone is used to represent the location where the high-level spatial reference begins to fail; the mid-level obstruction compression zone is used to represent the difficulty in spatial judgment when the transition structure obstructs the lower-level entrance; and the lower-level obstacle compression zone is used to represent the urgent state of altitude judgment when the UAV is about to enter the lower-level obstacle gate. In a preferred embodiment, the dynamic height compression intensity is based on the dive velocity compression intensity, further incorporating factors such as inter-level height difference, current height deviation, level occlusion intensity (ranging from 0 to 1, where 0 indicates no occlusion and 1 indicates complete occlusion), remaining distance, and entrance safety distance. Generally, the greater the height deviation, the more severe the occlusion, and the shorter the remaining distance, the higher the dynamic height compression intensity; while a larger entrance safety distance can alleviate the pressure. The dynamic height compression intensity comprehensively reflects the spatial positioning pressure experienced by the operator at the current moment due to inaccurate height perception and insufficient time to correct the height. The system then adjusts the range and intensity of the compression zone based on the drone's remaining dive distance, current altitude difference, lower barrier gate opening height, and lateral offset. Specifically, the intensity of the compression zone automatically increases when the drone's altitude deviates more from the center of the lower barrier gate, the remaining dive distance is shorter, and the lateral offset is greater. When the drone is close to the center of the lower barrier gate and the remaining dive distance is longer, the intensity of the compression zone gradually decreases.

[0029] S24: Based on the dynamic height compression region results, the switching process between the upper-level spatial reference and the lower-level spatial reference is compressed, and the hierarchical reference takeover strength is quantified to obtain the hierarchical reference switching compression results. In a preferred embodiment, the compression process is as follows: when the UAV has not yet entered the dive entrance, the upper-level spatial orientation reference remains dominant; when the UAV enters the dive entrance compression zone, the upper-level reference intensity begins to decay; when the UAV passes through the middle-level obstruction compression zone, the upper-level reference intensity decreases rapidly, while the lower-level reference intensity has not yet been fully established; when the UAV approaches the lower-level obstacle compression zone, the lower-level spatial reference intensity increases rapidly and takes over the current orientation judgment. The hierarchical reference takeover strength describes the dynamic process of switching from a higher-level reference to a lower-level reference; it depends on the initial strength of the higher-level reference, the target strength of the lower-level reference, the duration after entering the subduction zone, the time constant of the decay of the higher-level reference, and the time constant of the takeover of the lower-level reference; the longer the duration, the higher-level reference will decay exponentially, while the lower-level reference will rise exponentially; in general, the takeover strength transitions from the initial value to the target value over time, and there will be a gap of insufficient dual references in the middle, which is the core source of height reference collapse; It should be noted that the above-mentioned level switching process is not a linear replacement, but rather there is a short window of insufficient dual references, which is the core source of high reference collapse.

[0030] S25: Based on the results of the dive velocity compression area, evaluate the overall vertical compression intensity to quantify the total compression degree caused by the current vertical slicing scene to the operator's spatial judgment. Combine the results of view size shrinkage and layer occlusion synchronization, dynamic height compression area results and layer reference switching compression results for unified encapsulation to form a set of vertical compression states, and output them in stages according to the dive entry stage, middle layer occlusion stage and lower layer approach stage. The set of vertical compression states includes at least: dive velocity compression intensity, magnification rate of the barrier gate in the field of view, hierarchical occlusion intensity, dynamic height compression intensity, upper-level reference attenuation state, lower-level reference takeover state, and reference window duration. In a preferred embodiment, the overall vertical compression intensity is determined by the dynamic height compression intensity, the hierarchical reference tube intensity, the dive velocity compression intensity, and the hierarchical occlusion intensity. Specifically, the higher the dynamic height compression intensity, the less sufficient the hierarchical reference tube (i.e., the lower the tube intensity), and the more obvious the dive compression caused by occlusion, the greater the overall compression intensity. It represents the total degree of compression caused by the current vertical slicing scenario on the operator's spatial judgment.

[0031] S3: Based on the set of vertical compression states, the spatial positioning status of the operator during the subduction and shearing process is continuously collected to identify the phenomenon of misinheritance of spatial orientation in high-rise space, assess the intensity of height reference collapse, and perform aggregate analysis on the spatial positioning imbalance areas in different subduction stages to obtain the spatial offset assessment results and the set of height reference collapse areas.

[0032] Furthermore, S3 includes: S31: Based on the set of vertical compression states, the spatial positioning status of the UAV is continuously collected in the dive entry stage, the middle layer obstruction stage, and the lower layer approach stage, and its dive-cutting layer spatial offset density is evaluated to determine whether the operator has experienced spatial positioning instability in the dive-cutting layer stage. Finally, the data is sorted in chronological order to obtain the dive-cutting layer spatial offset acquisition results. It should be noted that the spatial positioning status of the UAV includes: the altitude deviation between the current altitude of the UAV and the center altitude of the lower barrier gate, the lateral deviation between the current lateral position of the UAV and the center line of the lower barrier gate, the directional deviation between the current flight direction of the UAV and the direction of passage of the lower barrier gate, and the advance or lag time between the time when the operator starts turning and the standard turning time. For example, in this embodiment, the sampling frequency is set to 60 to 180 times per second; it should be noted that the higher the racing speed, the higher the sampling frequency; the allowable height deviation is set to 0.1 meters to 0.3 meters; the allowable lateral deviation is set to 0.1 meters to 0.5 meters; and the allowable turning time deviation is set to 0.05 seconds to 0.2 seconds. In a preferred embodiment, the dive-and-cut spatial offset density is mainly obtained based on altitude deviation, lateral deviation, layer obstruction strength, overall vertical pressure strength, reference flight speed, remaining distance, and entry safety distance. Generally speaking, the greater the altitude and lateral deviations, the stronger the obstruction, and the greater the overall pressure, the higher the offset density; while the longer the remaining distance and the greater the entry safety distance, the lower the offset density. This density is used to determine whether the operator has experienced spatial positioning instability during the dive-and-cut phase. It should be noted that this step converts the operator's flight status from a single trajectory offset to a spatial offset acquisition result composed of altitude, lateral, orientation, and time sequence.

[0033] S32: Based on the spatial offset acquisition results of the dive-cut layer, compare the consistency between the operator's current flight direction and the main flight direction of the upper-level track and the passage direction of the lower-level obstacle gate to determine whether there is a high-level spatial direction mis-inheritance phenomenon, and evaluate the intensity of high-level spatial direction mis-inheritance to quantify the severity of mis-inheritance and obtain the high-level spatial direction mis-inheritance identification result. For example, in this embodiment, the determination of misinheritance of high-level space direction is first made based on the direction deviation combined with the hierarchical reference takeover strength, so as to avoid misjudging normal transition flight maneuvers as misinheritance, and then the strength of misinheritance of high-level space direction is evaluated. It should be noted that when the drone is already in the lower-level approach phase, but the operator's flight direction is still significantly close to the upper-level main flight direction, and the turning lead is still changing according to the rhythm of the upper-level track, it is determined that there is a phenomenon of incorrect inheritance of the upper-level spatial direction; when the lower-level reference has not yet been established, it is only recorded as a transitional observation state.

[0034] Among them, when the deviation between the operator's current flight direction and the main flight direction at higher altitudes is less than the preset error inheritance judgment threshold (the preferred range is a direction difference of 10 degrees to 35 degrees, with a larger threshold in the initial training stage and a smaller threshold in the advanced racing stage), it indicates that the direction is significantly close to the direction at higher altitudes. The preferred range for the preset error inheritance judgment threshold is a directional difference of 10 degrees to 35 degrees. It should be noted that a larger threshold is used in the initial training stage and a smaller threshold is used in the advanced racing stage. The intensity of high-level spatial orientation misinheritance depends on the deviation of the operator's current flight direction from the lower-level transit direction, the deviation from the upper-level main flight direction, the layer reference takeover intensity, the dive-cut spatial offset density, the allowable orientation deviation, and the orientation smoothing compensation constant. Generally speaking, the greater the orientation deviation towards the upper level (i.e., the smaller the deviation from the upper-level direction and the larger the deviation from the lower-level direction), the higher the misinheritance intensity. At the same time, the higher the reference takeover intensity and the greater the offset density, the more the misinheritance characteristics will be amplified. The intensity of high-level spatial orientation misinheritance is used to distinguish between normal transitions and true high-level orientation misinheritance. It should be noted that this step is used to distinguish between normal dive transition and incorrect inheritance of high-level direction, so as to avoid simply attributing all dive deviations to control errors.

[0035] S33: Based on the high-altitude spatial orientation error inheritance identification results, the three conditions are comprehensively judged to generate the height reference collapse judgment result, and the height reference collapse intensity is given according to the dive stage. For example, in this embodiment, the three conditions include: the operator retains high-level directional movements during the period when the lower-level reference should take over; the drone's altitude deviation and lateral deviation increase simultaneously; and the steering lead is significantly delayed or premature relative to the lower-level obstacle gate passage requirement. When all three conditions above occur simultaneously, it indicates that the operator's flight is not simply unstable (i.e., simply an operational error), but rather that the altitude reference has collapsed during the inter-level switching process, and is therefore determined to be altitude reference collapse. Then, the height-referenced collapse intensity is given for each of the dive stages (dive entry stage, mid-level occlusion stage, and lower-level approach stage). The height-referenced collapse intensity is determined by the direction error inheritance intensity, turning time deviation, height deviation, lateral deviation, and their respective allowable deviation values, combined with the level-referenced take-off intensity. Generally speaking, the more severe the direction error inheritance, the greater the turning time deviation, and the greater the height and lateral deviation, the higher the collapse intensity. The larger the allowable deviation, the greater the deviation that can be tolerated, and the relatively lower the collapse intensity. In addition, if the deviation is still significant when the reference take-off intensity is higher, it indicates that the operator has missed the normal take-off window, and the collapse risk is more prominent. It should be noted that the height reference collapse intensity accounts for the highest proportion during the lower approach stage, because this stage directly affects the safety of crossing the barrier gate.

[0036] S34: Based on the height reference collapse determination results, the locations where the height reference collapse intensity continuously exceeds the preset threshold are aggregated into candidate imbalance regions according to time, space and stage consistency. Then, the real imbalance regions are filtered out by the spatial positioning imbalance aggregation intensity, and the imbalance region type is marked to obtain the spatial positioning imbalance region aggregation result.

[0037] In a preferred embodiment, the spatial positioning imbalance aggregation intensity depends on the height reference collapse intensity, the duration of the same imbalance region, the spatial continuity of the same imbalance region, and the hierarchical occlusion intensity; these four factors together determine the final spatial positioning imbalance aggregation intensity value. Specifically, the spatial positioning imbalance cohesion intensity and the height reference collapse intensity change in the same direction; the stronger the height reference collapse, the higher the cohesion intensity, and the two have a direct positive relationship. Regarding duration, a benchmark value called the "minimum effective duration" needs to be introduced, which is typically between 0.1 and 0.5 seconds. The larger the ratio of the actual measured imbalance duration to this minimum effective duration, the higher the cohesion intensity; that is, the longer the duration, the greater the cohesion intensity. Similarly, a minimum effective spatial length exists as a benchmark for spatial continuity. The benchmark value ranges from 0.2 meters to 1 meter. The larger the ratio of the actual measured continuous spatial length to this minimum effective spatial length, the higher the aggregation strength. In other words, the longer the imbalanced area extends in space, the greater the aggregation strength. The hierarchical shading strength is a quantity between 0 and 1, where 0 represents no shading and 1 represents complete shading. The closer the shading strength is to one, the higher the aggregation strength. Specifically, the aggregation strength increases proportionally with the increase of the shading strength, that is, the aggregation strength is equal to the sum of the aforementioned factors multiplied by (one plus the shading strength). Areas with sufficient duration of imbalance (i.e., 0.1 to 0.5 seconds, short-term single-point anomalies are not considered stable areas), sufficient continuous spatial distance (i.e., 0.2 to 1 meter, to avoid mistaking instantaneous attitude jitter for stable collapse areas), and significant occlusion effects are identified as true spatial positioning imbalance areas. Based on the track level, dive phase, and gate approach distance of the spatial positioning imbalance area, the imbalance area type is marked as follows: when the imbalance area mainly appears in the dive entrance phase, it is marked as a high-level reference exit lag area; when it mainly appears in the mid-level occlusion phase, it is marked as a reference window occlusion enhancement area; when it mainly appears in the lower level approach phase, it is marked as a lower level gate takeover failure area. It should be noted that this step is used to aggregate scattered collapse decision points into region objects that can be used for training correction.

[0038] Furthermore, S3 also includes:

[0039] S35: Based on the aggregation results of spatial positioning imbalance areas, by evaluating the comprehensive evaluation value of spatial offset, generate spatial offset evaluation results for hierarchical spatial benchmark reconstruction and a set of height reference collapse areas, and bind each collapse area to the corresponding hierarchical reference takeover status and the lower-level barrier gate spatial direction benchmark.

[0040] The spatial offset assessment results include at least: average height deviation, maximum height deviation, average lateral deviation, maximum lateral deviation, orientation error inheritance intensity, turning time deviation, collapse intensity level, and imbalance region type; The height-referenced collapse region set includes at least: the dive entrance collapse region, the middle-layer shielding collapse region, and the lower-layer approach collapse region, and records their start position, end position, duration, region intensity, and corresponding obstacle gate number respectively. For example, in this embodiment, the comprehensive evaluation value of spatial offset is determined by four factors: spatial positioning imbalance aggregation intensity, high-level spatial orientation misinheritance intensity, height reference collapse intensity, and subduction shearing spatial offset density; the comprehensive evaluation value of spatial offset is the average level of these four factors; specifically, each factor contributes equally to the evaluation value, and each of the four factors has the same weight. The values ​​of these factors are added together and then divided by four to obtain the final comprehensive evaluation value. It should be noted that when the spatial positioning imbalance aggregation intensity of a certain region is lower than the preset training threshold, it is only used as an observation region; when the spatial positioning imbalance aggregation intensity exceeds the training threshold, it is used as the target region for hierarchical spatial benchmark reconstruction. The preferred range for the training threshold is between 1.2 and 2.0. It should be noted that when in the initial training stage, a higher training threshold is used to avoid frequent corrections, while when in the advanced training stage, a lower threshold is used to improve recognition sensitivity.

[0041] S4: Based on the spatial offset assessment results and the set of height reference collapse areas, the layered spatial correction results are obtained by strengthening the lower barrier gate direction reference, remapping the spatial relationship of the diving area, adjusting the reference switching rhythm, and correcting the virtual vertical compression intensity. Furthermore, S4 includes: S41: Based on the spatial offset assessment results and the set of height reference collapse regions, extract the directional error inheritance intensity, spatial offset comprehensive situation, height reference collapse intensity, and lower-level barrier gate passage direction in the lower-level near-collapse region; comprehensively determine the reinforcement magnitude of the lower-level barrier gate directional reference, obtain the lower-level barrier gate directional reference reinforcement coefficient, and actually perform reinforcement processing in the virtual environment to obtain the lower-level barrier gate directional reference reinforcement result; For example, in this embodiment, the lower-level barrier gate directional reference reinforcement coefficient depends on the spatial offset comprehensive evaluation value, the upper-level spatial directional error inheritance intensity, the height reference collapse intensity, and the hierarchical reference takeover intensity. Among them, the spatial offset comprehensive evaluation value, the directional error inheritance intensity, and the height reference collapse intensity have a positive impact on the reinforcement coefficient. Generally speaking, the larger these three factors are, the larger the directional reference reinforcement coefficient is. However, the reinforcement coefficient is not simply added together, but rather a combination of these three factors is taken first (the positive contributions of the three factors work together), and then divided by a quantity involving the hierarchical reference takeover intensity. Specifically, the hierarchical reference takeover intensity is a factor with a value between 0 and 1, which plays a regulatory role. For example, when the hierarchical reference takeover is already relatively sufficient (i.e., the takeover intensity is relatively large), the reinforcement coefficient will be appropriately reduced to avoid excessive prompting. It should be noted that when it is detected that the operator is still maintaining the orientation of the high-level space during the approach phase of the lower level, the orientation reference of the lower level obstacle gate is enhanced. The enhanced objects include the center line of the lower level obstacle gate, the passing direction of the lower level obstacle gate, the boundary of the lower level entrance, and the main flight direction of the lower level track. The enhancement process specifically includes: improving the clarity of the lower obstacle gate frame in the virtual racing environment, enhancing the lower entrance centerline prompt, reducing the residual reference intensity of the upper level, and making the lower obstacle gate passable in the operator's view in advance; It should also be noted that, in this embodiment, the enhancement process is actually performed in the virtual environment in order to improve the perceptible strength of the lower spatial reference, prevent the operator from continuing to use the direction of the upper track to judge the location of the lower entrance, and enable the operator to actively abandon the reference of the upper direction at the end of the dive. S42: Based on the enhanced results of the lower barrier gate direction reference, combined with the hierarchical occlusion intensity, the aggregation intensity of the spatial positioning imbalance area, the remaining distance from the UAV to the lower barrier gate, and the entrance safety distance, the remapping intensity of the spatial relationship in the dive area is comprehensively determined; according to the remapping intensity, the dive process is divided into three continuous correction segments: dive entrance, middle layer occlusion, and lower layer approach, and correction processing is performed to obtain the remapping result of the spatial relationship in the dive area; In a preferred embodiment, the remapping intensity of the spatial relationships in the dive region is determined by four factors, specifically including: the lower barrier gate directional reference enhancement coefficient, the hierarchical occlusion intensity, the spatial positioning imbalance aggregation intensity, and two distance-related quantities: the remaining spatial distance from the UAV to the center of the lower barrier gate and the safe distance to the lower barrier gate entrance. The specific relationships are as follows: the larger the directional reference enhancement coefficient, the greater the remapping intensity; the stronger the hierarchical occlusion intensity (the closer its value is to one), the greater the remapping intensity; the higher the spatial positioning imbalance aggregation intensity, the greater the remapping intensity. All three factors positively contribute to the remapping intensity. On the other hand, the ratio of remaining space distance to safe distance has an inhibitory effect: the larger the ratio of remaining distance to safe distance, the smaller the remapping intensity; in other words, the shorter the remaining distance and the closer it is to the safe distance, the greater the remapping intensity.

[0042] The specific corrective processing includes: in the diving entrance section, reducing the directional pull of the high-level platform boundary on the operator, while increasing the visual difference between the walls, columns, beams and barrier gates to help the operator distinguish between the real traversable area and the visually oppressive area; in the middle-level obstruction section, highlighting the outline of the lower-level entrance boundary; in the lower-level approach section, strengthening the consistency between the center line of the barrier gate and the direction of passage, and controlling the magnification rate of the barrier gate in the field of view to avoid sudden expansion leading to secondary misjudgment; S43: Based on the spatial relationship remapping results of the subduction area, and combined with the evaluation of the hierarchical reference switching rhythm adjustment intensity, the exit rhythm of the high-level spatial reference and the takeover rhythm of the lower-level spatial reference are dynamically adjusted to obtain the hierarchical reference switching rhythm adjustment results. In a preferred embodiment, the intensity of the hierarchical reference switching rhythm adjustment is influenced by the following factors: the lower-level barrier gate directional reference enhancement coefficient, the spatial relationship remapping intensity in the dive area, the overall vertical pressure intensity, the reference flight speed, the upper-level reference decay time constant, and the lower-level reference takeover time constant. Among these, the directional reference enhancement coefficient, spatial relationship remapping intensity, and overall vertical pressure intensity have a positive effect on the adjustment intensity; generally, the larger these three factors are, the greater the rhythm adjustment intensity. However, the adjustment intensity is also regulated by two time constants: when the upper-level reference decay time constant is relatively long, the adjustment intensity will decrease to avoid excessively abrupt reference breakage; the lower-level reference takeover time constant indirectly participates through the combined relationship between the reference flight speed and the pressure intensity. Overall, the hierarchical reference switching rhythm adjustment intensity accelerates the hierarchical reference switching rhythm when the vertical pressure is stronger and the spatial remapping is more pronounced; while when the upper-level reference decay time is relatively long, the adjustment intensity decreases.

[0043] It should be noted that the dynamic adjustment is as follows: for the dive entrance section, the high-level reference is not immediately cut off, but the reference intensity in the high-level direction is gradually reduced according to the set decay time; for the mid-level obstruction section, a short reference gap compression process is set to make the operator feel the unstable state of spatial reference in the actual competition; in the lower approach section, the lower reference takeover intensity is increased so that the lower obstacle gate direction reference can take over the operator's spatial judgment in a timely manner. Set the high-level reference decay time to 0.4 to 1 second, and the low-level reference take-off time to 0.5 to 1.7 seconds. It should be noted that for beginners, longer decay and take-off times are used, while for advanced trainees, shorter values ​​are used to enhance the sense of pressure in real racing. It should also be noted that this step aims to allow higher-level references to gradually withdraw and lower-level references to gradually take over, thereby reconstructing the spatial reference transformation relationship in the actual subduction and shearing process. S44: Based on the adjustment results of the hierarchical reference switching rhythm, and combined with the correction intensity of the virtual vertical pressure obtained from the evaluation, the vertical pressure intensity in the virtual racing environment is continuously corrected to obtain the virtual vertical pressure intensity correction result.

[0044] For example, in this embodiment, the correction intensity of the virtual vertical compression is determined based on the comprehensive vertical compression intensity, the intensity of the hierarchical reference switching rhythm adjustment, the intensity of the height reference collapse, and the training stability correction coefficient. Generally, the comprehensive vertical compression intensity, rhythm adjustment intensity, and height reference collapse intensity all have a positive promoting effect on the correction intensity; increasing any one of these factors will increase the virtual vertical compression correction intensity. The training stability correction coefficient, on the other hand, has a suppressive effect. The larger the value of this coefficient (ranging from 0.5 to 1.5, with larger values ​​for primary training and smaller values ​​for advanced training), the smaller the correction intensity. In other words, the training stability correction coefficient limits the correction range to prevent excessively large changes in the training scenario. The continuous correction of the vertical compression intensity in the virtual racing environment, in this embodiment, is performed gradually according to the training rounds, without changing all parameters at once. Specifically: If the lower approach phase in the training environment is too smooth, increase the gate view magnification, boundary contraction intensity, and middle occlusion intensity; if the spatial pressure in the training environment is too strong, causing the operator to be unable to form effective corrective actions, reduce the occlusion duration and boundary pressure amplitude. It should be noted that after each round of training, the intensity of vertical compression is fine-tuned for the next round based on the comprehensive evaluation value of spatial offset, the intensity of height reference collapse, and the rhythm of hierarchical reference switching, so that the virtual environment gradually approaches the spatial imbalance state in real drone racing competitions. Furthermore, S4 also includes: S45: The results of strengthening the lower barrier gate direction reference, remapping the spatial relationship of the diving area, adjusting the level reference switching rhythm, and correcting the virtual vertical pressure intensity are uniformly packaged. At the same time, the comprehensive correction value of the layered space is evaluated to quantify the magnitude of the overall correction and obtain the layered space correction result.

[0045] The layered spatial correction results include at least: the lower-level barrier gate direction reference reinforcement coefficient, the remapping intensity of the spatial relationship of the diving area, the adjustment intensity of the layer reference switching rhythm, the correction intensity of the virtual vertical compression, the upper-level reference exit time, the lower-level reference takeover time, the occlusion intensity correction value, the barrier gate proximity relationship correction value, and the layered spatial comprehensive correction value. For example, in this embodiment, the layered spatial comprehensive correction value is determined by four factors, specifically including: the lower-level obstacle gate directional reference enhancement coefficient, the dive area spatial relationship remapping intensity, the layer reference switching rhythm adjustment intensity, and the ratio of virtual vertical compression correction intensity to the reference flight speed; specifically, the directional reference enhancement coefficient, spatial relationship remapping intensity, and rhythm adjustment intensity directly contribute positively to the comprehensive correction value; while the virtual vertical compression correction intensity participates in the positive combination by dividing by the reference flight speed. S5: Based on the results of the hierarchical spatial correction, regenerate the vertical switching racing scenario, comprehensively analyze the operator's various capabilities in continuous dive and layer cutting, continuously enhance the operator's spatial adaptation status in different layers, and correlate and correct the training behavior with the real racing behavior, outputting the UAV three-dimensional racing training results and the real race spatial adaptation ability evaluation results. Furthermore, S5 includes: S51: Based on the hierarchical spatial correction results and combined with the benchmark flight speed, the pressure intensity of the reconstructed scene is determined, and multiple parameters in the virtual training environment are reset to obtain the vertical switching racing scene. For example, in this embodiment, the layered spatial comprehensive correction value is first extracted from the layered spatial correction result, and combined with the reference flight speed to determine the reconstructed scene compression intensity. The reconstructed scene compression intensity is determined by the following factors, specifically including: the layered spatial comprehensive correction value, the reference flight speed, the virtual vertical compression correction intensity, and the training stability correction coefficient. Generally speaking, the larger the layered spatial comprehensive correction value, the greater the reconstructed scene compression intensity; the larger the reference flight speed itself, the greater the compression intensity; the virtual vertical compression correction intensity will further enhance the compression intensity, and the larger its ratio to the reference flight speed, the higher the compression intensity; while the training stability correction coefficient plays a suppressive role: the larger the value of this coefficient, the smaller the reconstructed scene compression intensity. Based on the intensity of the reconstructed scene pressure, multiple parameters in the virtual training environment are reset so that the training scene no longer maintains smooth layering, but forms a diving pressure process that is close to that of a real game. The multiple parameters include: high-level reference exit time, lower-level reference takeover time, magnification rate of the barrier gate in the field of view, boundary pressure intensity, and occlusion duration. It should be noted that during the process of resetting multiple parameters, the boundary pressure intensity is reduced during the initial training and increased during the advanced training, so that the operator can gradually adapt to the switching of height reference in real three-dimensional racing competitions. Finally, these reset parameters are applied to the vertical switching process between the high-level racing area, the mid-level obstruction area, and the low-level obstacle crossing area, forming a racing training scenario that is no longer smooth but has a realistic sense of downward pressure.

[0046] S52: Based on the reconstructed vertical switching racing scenario, continuously detect the operator's spatial positioning stability, obstacle gate crossing stability, and spatial orientation recovery ability during the process of entering the dive entrance, passing through the middle layer of obstruction, and passing through the lower layer obstacle gate, and determine the continuous dive layer cutting stability coefficient. For example, in this embodiment, the spatial positioning stability capability is determined by height deviation, lateral deviation, and directional deviation; the obstacle gate crossing stability capability is determined by crossing center deviation, crossing speed fluctuation, and collision boundary margin; and the spatial orientation recovery capability is determined by the time it takes for the operator to re-align with the lower level direction after the high-level reference is removed. It should be noted that when an operator is able to shorten the direction recovery time, reduce the height deviation, and maintain stable passage through the obstacle gate in multiple rounds of training, it is determined that their real-world racing adaptability has improved. In a preferred embodiment, the continuous dive-cut stability coefficient is determined by three deviations and their corresponding allowable deviations, specifically: altitude deviation (the difference between the UAV's current altitude and the center height of the lower barrier gate), lateral position deviation (the difference between the UAV's current lateral position and the centerline of the lower barrier gate), and turning time deviation (the difference between the actual turning time and the standard turning time). Each deviation is normalized based on its allowable deviation; specifically, the stability coefficient decreases as these normalized deviations increase, for example, the larger the deviation, the lower the stability coefficient. The absolute values ​​of these three deviations are divided by their respective allowable deviations (the allowable altitude deviation ranges from 0.1 meters to 0.3 meters, the allowable lateral deviation from 0.1 meters to 0.5 meters, and the allowable time deviation from 0.05 seconds to 0.2 seconds), and then these three ratios are added together. The stability coefficient is equal to one divided by one plus this sum; that is, the continuous dive-cut stability coefficient is generally negatively correlated with each deviation: the larger the deviation, the smaller the stability coefficient.

[0047] S53: Based on the continuous subduction shearing stability coefficient, combined with the layer reference pipe strength, height reference collapse strength and high-level spatial direction misinheritance strength, the layered racing adaptability value is comprehensively determined, and the spatial adaptability of the operator in different layered racing environments is continuously strengthened to obtain the layered racing adaptability strengthening result. In a preferred embodiment, the stratified racing adaptability value depends on the continuous subduction shear stability coefficient, the stratified reference take-off strength, the height reference collapse strength, and the upper-level spatial orientation misinheritance strength. Generally speaking, the stability coefficient and the stratified reference take-off strength have a positive promoting effect on the adaptability value, specifically: the larger the stability coefficient and the stronger the take-off strength, the higher the adaptability value; while the height reference collapse strength and the upper-level spatial orientation misinheritance strength have an inhibitory effect, specifically: the larger these two strengths are, the lower the adaptability value. It should be noted that in this embodiment, the continuous reinforcement process is executed progressively according to training rounds. Each round only adjusts a small number of reference parameters to avoid the operator developing a training habit that relies on prompts. Continuous reinforcement specifically includes: When the operator deviates significantly during the dive entrance phase, the high-level reference exit time is extended and the reconstruction scene pressure intensity is reduced; when the operator is significantly unbalanced during the mid-level occlusion phase, the occlusion duration is shortened and the lower-level entrance boundary outline is enhanced; when the operator still exhibits directional error inheritance during the lower-level approach phase, the lower-level obstacle gate directional reference enhancement coefficient is increased. It should also be noted that this step adjusts the training intensity in reverse based on the operator's actual performance, so that the training goal is to improve from completing the crossing to stably adapting to real racing and layer cutting.

[0048] After continuous enhancement, the results of layered racing adaptation enhancement are obtained, including at least: layered racing adaptation capability value and adjustment suggestions for each stage, such as: adjustment amount of high-level reference exit time, adjustment amount of reconstruction scene pressure intensity, adjustment amount of occlusion duration, degree of display enhancement of lower-level entrance boundary contour, and adjustment amount of lower-level obstacle gate direction reference enhancement coefficient. S54: Based on the results of hierarchical racing adaptation reinforcement, the spatial reference switching behavior formed during training is correlated and corrected with the real racing behavior, and key data of each round of training is recorded. Based on the recorded key data, the operator's adaptability to the real competition space is comprehensively determined, the operator's degree of adaptability to the real competition is quantified, and finally the results of the drone 3D racing training and the evaluation results of the adaptability to the real competition space are formed.

[0049] It should be noted that, in this embodiment, the key data recorded for each round of training specifically includes the reconstructed scene pressure intensity, continuous dive-and-cut stability coefficient, layered racing adaptability value, lower obstacle gate crossing success rate, direction recovery time, and height deviation convergence status in each round of training. In a preferred embodiment, the operator's real-world space adaptability is jointly determined by the layered racing adaptability value, the continuous dive-and-cut stability coefficient, the success rate of passing through the lower obstacle gate, and the height-referenced collapse intensity. Generally, the layered racing adaptability value, stability coefficient, and success rate of passing through obstacles have a positive promoting effect on the adaptability value. The higher these three factors are, the higher the real-world space adaptability value. The height-referenced collapse intensity has an inhibitory effect: the higher the collapse intensity, the lower the adaptability value. That is, the real-world space adaptability is an evaluation result that can be used for real-world events by combining adaptability, stable passing ability, actual passing success rate, and collapse risk. Then, a judgment is made based on the operator's actual performance in high-pressure scenarios and the adaptability to real-world competition spaces. Specifically, if the operator can maintain stable passage in high-pressure scenarios, a higher assessment result for adaptability to real-world competition spaces is output; if the operator is only stable in low-pressure scenarios, a result indicating that further training is needed is output.

[0050] The final results will be the training results for three-dimensional drone racing and the evaluation results of its adaptability to real competition spaces.

[0051] Example 2: This embodiment provides an electronic device, including: a processor and a memory, wherein the memory stores a computer program that can be called by the processor; The processor executes the aforementioned drone racing training method based on a virtual-real confrontation mechanism by calling the computer program stored in memory.

[0052] The electronic device can vary considerably depending on its configuration or performance. It may include one or more Central Processing Units (CPUs) and one or more memories, wherein the memory stores at least one computer program, which is loaded and executed by the processor to implement the UAV racing training method based on a virtual-real adversarial mechanism provided in the above-described embodiment. The electronic device may also include other components for implementing its functions; for example, it may have wired or wireless network interfaces and input / output interfaces for data input and output. Details will not be elaborated upon in this embodiment.

[0053] Example 3: This embodiment proposes a computer-readable storage medium on which an erasable and rewritable computer program is stored. When the computer program runs on the computer device, it causes the computer device to execute the above-mentioned drone racing training method based on the virtual-real confrontation mechanism.

[0054] For example, computer-readable storage media can be read-only memory (ROM), random access memory (RAM), compact disc read-only memory (CD-ROM), magnetic tape, floppy disk, and optical data storage devices.

[0055] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0056] It should be understood that determining B based on A does not mean determining B solely based on A; it also means determining B based on A and / or other information.

[0057] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0058] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A drone racing training method based on a virtual-real adversarial mechanism, characterized in that, The method includes the following: S1: By performing layered analysis on the real multi-layered three-dimensional track structure, we independently construct virtual and real racing spaces for each layer and define its visual reference structure, boundary compression structure and spatial direction benchmark, generating a set of layered spatial references. S2: Based on the hierarchical spatial reference set, the dive velocity compression region, the view size contraction and occlusion state, and the dynamic height compression region are generated sequentially. The process of switching between high-level and low-level references is compressed to obtain the vertical compression state set. S3: Based on the set of vertical compression states, continuously collect the spatial positioning status of the operator during the subduction and shearing process, identify the phenomenon of misinheritance of spatial orientation in high-rise space, evaluate the intensity of height reference collapse, and perform aggregate analysis on the spatial positioning imbalance area in different subduction stages to obtain the spatial offset evaluation results and the set of height reference collapse areas. S4: Based on the spatial offset assessment results and the set of height reference collapse areas, the layered spatial correction results are obtained by strengthening the lower barrier gate direction reference, remapping the spatial relationship of the diving area, adjusting the reference switching rhythm, and correcting the virtual vertical compression intensity. S5: Based on the hierarchical spatial correction results, regenerate the vertical switching racing scenario, comprehensively analyze the operator's various capabilities in continuous dive-cutting, continuously enhance the operator's spatial adaptation status in different levels, and correlate and correct the training behavior with the real racing behavior, outputting the UAV 3D racing training results and the real race spatial adaptation ability evaluation results.

2. The UAV racing training method based on virtual-real adversarial mechanism according to claim 1, characterized in that, S1 includes: S11: First, taking a real drone racing track as the object, collect the height of each obstacle gate and the historical flight altitude change trajectory of the drone, analyze the height difference between the center of the obstacle gate and the track reference plane, and divide the track into a high-level racing area, a middle-level transition area and a lower-level obstacle crossing area to obtain the track level division results. S12: Based on the results of the track level division, further calibrate the vertical spacing, dive angle range, cut-in and cut-out distances, and turning compression areas between adjacent levels, record the key positions and compression center points of the dive segment, and form the vertical switching calibration results between levels. S13: Based on the vertical switching calibration results between layers, the high-level racing area, the middle-level transition area, and the lower-level obstacle crossing area of ​​the real track are mapped to the virtual training space respectively. At the same time, the compression intensity of the inter-layer switching is evaluated to obtain the result of the construction of the layered virtual and real racing space. S14: Based on the results of constructing a hierarchical virtual-real racing space, a visual reference structure, a boundary compression structure, and a spatial direction benchmark are defined for each level of space. At the same time, the hierarchical spatial reference intensity is quantified, and the visual reference structure, boundary compression structure, spatial direction benchmark, and hierarchical spatial reference intensity are combined into a hierarchical spatial reference set for output.

3. The UAV racing training method based on virtual-real confrontation mechanism according to claim 2, characterized in that, S2 include: S21: Based on the hierarchical spatial reference set, extract the vertical height difference between adjacent levels, the position of the dive entrance and exit, the distance of the lower obstacle gate entrance, and the dive flight speed of the UAV. Generate a dive speed compression region in the virtual training space, and evaluate the dive speed compression intensity of the region to form the dive speed compression region result. S22: Based on the results of the dive velocity compression region, the obstacle gate size change, boundary contraction change and hierarchical switching occlusion state in the first view of the UAV are generated synchronously, and the synchronous results of view size contraction and hierarchical occlusion are obtained through segmented control. S23: Based on the results of the dive velocity compression region, the results of the view size shrinkage and the hierarchical occlusion synchronization, the dive velocity compression intensity is extracted, and combined with the dynamic height compression intensity, a dynamic height compression region is established. The range and intensity of the compression region are adjusted to obtain the dynamic height compression region results. The dynamic height compression zone includes: the dive entrance compression zone, the middle layer shielding compression zone, and the lower layer obstacle compression zone; S24: Based on the dynamic height compression region results, the switching process between the upper-level spatial reference and the lower-level spatial reference is compressed, and the hierarchical reference takeover strength is quantified to obtain the hierarchical reference switching compression results. S25: Based on the results of the dive velocity compression region, evaluate the overall vertical compression intensity, and combine the results of the viewpoint size shrinkage and hierarchical occlusion synchronization, the results of the dynamic height compression region, and the results of hierarchical reference switching compression to form a set of vertical compression states, and output them in stages according to the dive entry stage, the middle layer occlusion stage, and the lower layer approach stage.

4. The UAV racing training method based on virtual-real confrontation mechanism according to claim 3, characterized in that, S3 include: S31: Based on the set of vertical compression states, the spatial positioning status of the UAV is continuously collected in the dive entry stage, the middle layer obstruction stage, and the lower layer approach stage, and its dive cut layer spatial offset density is evaluated. Finally, the data is sorted in chronological order to obtain the dive cut layer spatial offset acquisition results. S32: Based on the spatial offset acquisition results of the dive-cut layer, compare the consistency between the operator's current flight direction and the main flight direction of the upper-level track and the passage direction of the lower-level obstacle gate to determine whether there is a high-level spatial direction mis-inheritance phenomenon, and evaluate the intensity of high-level spatial direction mis-inheritance to quantify the severity of mis-inheritance and obtain the high-level spatial direction mis-inheritance identification result. S33: Based on the identification results of high-altitude spatial orientation error inheritance, the three conditions are comprehensively judged to generate the height reference collapse judgment result, and the height reference collapse intensity is given according to the dive stage. S34: Based on the height reference collapse determination results, the locations where the height reference collapse intensity continuously exceeds the preset threshold are aggregated into candidate imbalance regions according to time, space and stage consistency. Then, the real imbalance regions are filtered out by the spatial positioning imbalance aggregation intensity, and the imbalance region type is marked to obtain the spatial positioning imbalance region aggregation result.

5. The UAV racing training method based on virtual-real adversarial mechanism according to claim 4, characterized in that, S3 also includes: S35: Based on the aggregation results of spatial positioning imbalance areas, by evaluating the comprehensive evaluation value of spatial offset, generate spatial offset evaluation results for hierarchical spatial benchmark reconstruction and a set of height reference collapse areas.

6. The UAV racing training method based on virtual-real confrontation mechanism according to claim 5, characterized in that, S4 include: S41: Based on the spatial offset assessment results and the set of height reference collapse regions, extract the directional error inheritance intensity, spatial offset comprehensive situation, height reference collapse intensity, and lower-level barrier gate passage direction in the lower-level near-collapse region; comprehensively determine the reinforcement magnitude of the lower-level barrier gate directional reference, obtain the lower-level barrier gate directional reference reinforcement coefficient, and actually perform reinforcement processing in the virtual environment to obtain the lower-level barrier gate directional reference reinforcement result; S42: Based on the enhanced results of the lower barrier gate direction reference, combined with the hierarchical occlusion intensity, the aggregation intensity of the spatial positioning imbalance area, the remaining distance from the UAV to the lower barrier gate, and the entrance safety distance, the remapping intensity of the spatial relationship in the dive area is comprehensively determined; according to the remapping intensity, the dive process is divided into three continuous correction segments: dive entrance, middle layer occlusion, and lower layer approach, and correction processing is performed to obtain the remapping result of the spatial relationship in the dive area; S43: Based on the spatial relationship remapping results of the subduction area, and combined with the evaluation of the hierarchical reference switching rhythm adjustment intensity, the exit rhythm of the high-level spatial reference and the takeover rhythm of the lower-level spatial reference are dynamically adjusted to obtain the hierarchical reference switching rhythm adjustment results. S44: Based on the adjustment results of the hierarchical reference switching rhythm, and combined with the correction intensity of the virtual vertical pressure obtained from the evaluation, the vertical pressure intensity in the virtual racing environment is continuously corrected to obtain the virtual vertical pressure intensity correction result.

7. The UAV racing training method based on virtual-real confrontation mechanism according to claim 6, characterized in that, S4 also includes: S45: Unify and encapsulate the results of strengthening the lower barrier gate direction reference, remapping the spatial relationship of the diving area, adjusting the level reference switching rhythm, and correcting the virtual vertical pressure intensity. At the same time, evaluate the comprehensive correction value of the layered space and the correction results of the layered space.

8. The UAV racing training method based on virtual-real confrontation mechanism according to claim 7, characterized in that, S5 include: S51: Based on the hierarchical spatial correction results and combined with the benchmark flight speed, the pressure intensity of the reconstructed scene is determined, and multiple parameters in the virtual training environment are reset to obtain the vertical switching racing scene. S52: Based on the reconstructed vertical switching racing scenario, continuously detect the operator's spatial positioning stability, obstacle gate crossing stability, and spatial orientation recovery ability during the process of entering the dive entrance, passing through the middle layer of obstruction, and passing through the lower layer obstacle gate, and determine the continuous dive layer cutting stability coefficient. S53: Based on the continuous subduction shearing stability coefficient, combined with the layer reference pipe strength, height reference collapse strength and high-level spatial direction misinheritance strength, the layered racing adaptability value is comprehensively determined, and the spatial adaptability of the operator in different layered racing environments is continuously strengthened to obtain the layered racing adaptability strengthening result. S54: Based on the results of hierarchical racing adaptation reinforcement, the spatial reference switching behavior formed during training is correlated and corrected with the real racing behavior, and key data of each round of training is recorded. Based on the recorded key data, the operator's adaptability to the real competition space is comprehensively determined, the operator's degree of adaptability to the real competition is quantified, and finally the results of the drone 3D racing training and the evaluation results of the adaptability to the real competition space are formed.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the drone racing training method based on the virtual-real confrontation mechanism as described in any one of claims 1-8.

10. An electronic device, characterized in that, include: Memory, used to store instructions; A processor is configured to execute the instructions, causing the device to perform operations that implement the UAV racing training method based on a virtual-real confrontation mechanism as described in any one of claims 1-8.

Images