A safety control system and method for unmanned vehicles in coal mines

By quantifying the impact of dust environment on the perception system in unmanned vehicles in coal mines and dynamically adjusting safety control strategies, the problem of insufficient perception capability in dusty environments was solved, and operational safety and transportation efficiency were improved.

CN123061418APending Publication Date: 2026-11-10ANHUI UNIV OF SCI & TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611066549.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-07-17
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

Existing safety control methods for unmanned underground coal mines do not consider the impact of dust environment on the performance degradation of the perception system, resulting in insufficient perception capability in high dust areas and an underestimation of collision risk, while in low dust areas, unnecessary deceleration or braking may be triggered, affecting transportation efficiency.

Method used

By acquiring the dust concentration value of the vehicle's current environment, and calculating the normalized dust concentration value in combination with the working surface type, the degradation index of LiDAR and visual perception are quantified respectively, the credibility of fused perception is calculated, and the collision risk assessment value is corrected based on this, and the safety control strategy is dynamically adjusted.

Benefits of technology

The threat level of dusty environments to the perception system was quantified, and collision risk assessments were dynamically adjusted, improving the operational safety and transportation efficiency of autonomous vehicles in complex dusty environments.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123061418A_ABST
    Figure CN123061418A_ABST
Patent Text Reader

Abstract

This invention discloses a safety control system and method for unmanned vehicles in coal mines, relating to the field of unmanned driving safety control technology. The system includes: a dust concentration normalization module for acquiring the ambient dust concentration value of the vehicle's current environment and calculating a normalized dust concentration value; a perception degradation calculation module for calculating the lidar perception degradation index and the visual perception degradation index based on the dust concentration normalization value and perception performance indicators; a fusion credibility calculation module for calculating the fusion perception credibility; a collision risk correction module for correcting the original collision risk assessment value to obtain a corrected collision risk value; and a safety control strategy module for determining the vehicle's current safety control strategy based on the fusion perception credibility and the corrected collision risk value. This invention solves the technical problem that existing unmanned driving safety control methods in coal mines do not consider the impact of dust environment on the performance degradation of the perception system.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of unmanned driving safety control technology, specifically to a safety control system and method for unmanned vehicles in coal mines. Background Technology

[0002] The underground environment of coal mines is characterized by low illumination, high dust concentration, narrow tunnels, and severe electromagnetic interference. Unmanned vehicles operating underground rely on multiple sensors, including lidar, millimeter-wave radar, and visual cameras, for environmental perception and obstacle detection. However, dust concentrations vary significantly across different areas underground, with much higher concentrations in tunneling and mining faces than in transport roadways. High dust concentrations severely reduce the detection range and effective point cloud quantity of lidar, while also decreasing the contrast and clarity of visual images, leading to a significant decline in the perception system's obstacle detection capability and positioning accuracy.

[0003] Existing safety control methods for unmanned underground coal mines typically employ fixed safety distance thresholds and collision risk assessment strategies, without considering the impact of dust environments on the performance degradation of the sensing system. In areas with high dust concentrations, insufficient sensing capabilities can easily lead to an underestimation of collision risks, while in areas with low dust concentrations, sufficient sensing capabilities may trigger unnecessary deceleration or braking, affecting transportation efficiency. Summary of the Invention

[0004] This invention provides a safety control system and method for unmanned vehicles in coal mines, which solves the technical problem that existing safety control methods for unmanned vehicles in coal mines do not consider the impact of dust environment on the performance degradation of the sensing system.

[0005] In a first aspect, the present invention provides a safety control system for unmanned vehicles operating underground in coal mines, the system comprising: The dust concentration normalization module is used to obtain the ambient dust concentration value of the current environment of the vehicle, and calculate the normalized dust concentration value based on the current environment. The perception degradation calculation module is used to calculate and obtain the lidar perception degradation index and the visual perception degradation index based on the normalized dust concentration value and the perception performance index of lidar and image recognition, respectively. The fusion credibility calculation module is used to calculate the fusion perception credibility based on the lidar perception degradation index and the visual perception degradation index. The collision risk correction module is used to correct the original collision risk assessment value of the obstacle detection based on the fusion perception confidence level, so as to obtain a corrected collision risk value. The safety control strategy module is used to determine the current safety control strategy of the vehicle based on the fused perception confidence level and the corrected collision risk value.

[0006] Secondly, the present invention also provides a safety control method for unmanned vehicles in coal mines, the method comprising: Obtain the ambient dust concentration value of the vehicle's current environment, and calculate the normalized dust concentration value based on the current environment. Based on the normalized dust concentration value, and combined with the perception performance indicators of lidar and image recognition, the lidar perception degradation index and the visual perception degradation index are calculated and obtained respectively. The credibility of fused perception is calculated based on the LiDAR perception degradation index and the visual perception degradation index. Based on the fusion perception reliability, the original collision risk assessment value of obstacle detection is corrected to obtain the corrected collision risk value; Based on the fused perception confidence level and the corrected collision risk value, the current safety control strategy for the vehicle is determined.

[0007] One or more technical solutions provided in this invention have at least the following technical effects or advantages: First, this invention collects the dust concentration value of the current environment of the vehicle and calculates the normalized dust concentration value in combination with the working face type of the current roadway. Different working face types correspond to different dust concentration benchmark values. Compared with existing methods that do not consider the impact of dust environment on the performance degradation of the sensing system, it can quantify the actual threat level of different dust environments to the sensing system.

[0008] Second, this invention calculates the lidar perception degradation index and the visual perception degradation index separately by using the normalized value of dust concentration. Based on the two degradation indices, it calculates the fusion perception credibility and uses the fusion perception credibility to correct the original collision risk assessment value. Compared with the existing methods that use a fixed collision risk assessment strategy, this invention can dynamically adjust the collision risk assessment results according to the actual degradation degree of the perception system under the current dust environment.

[0009] Third, this invention determines the vehicle's current safety control strategy based on the fusion of perception credibility and the correction of collision risk value, and triggers a higher level of safety response when perception is untrustworthy. Compared with existing methods that lack a linkage mechanism between perception credibility and safety control strategy, this invention can take safety measures in advance when perception degradation is severe, thereby improving the operational safety of unmanned underground vehicles in complex dust environments. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.

[0011] Figure 1 This is a structural diagram of a safety control system for an unmanned vehicle in a coal mine, provided in an embodiment of the present invention. Figure 2 This is a schematic diagram of a safety control method for unmanned vehicles in coal mines provided in an embodiment of the present invention. Figure 3 This is a sample table diagram illustrating the data acquisition of a safety control system for an unmanned vehicle in a coal mine, provided in an embodiment of the present invention. Figure 4 This is a flowchart illustrating a safety control method for unmanned vehicles in coal mines provided by an embodiment of the present invention. The diagram is labeled as follows: Dust concentration normalization module 11, perception degradation calculation module 12, fusion credibility calculation module 13, collision risk correction module 14, and safety control strategy module 15. Detailed Implementation

[0012] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0013] Example 1, as Figure 1 and Figure 2 As shown, this application provides an intelligent adjustment system for water-saving sprinkler irrigation equipment based on dynamic soil moisture sensing. The system includes the following modules: The dust concentration normalization module 11 is used to obtain the ambient dust concentration value of the current environment of the vehicle, and calculate the normalized dust concentration value in combination with the current environment. During operation, unmanned vehicles in coal mines rely heavily on sensing sensors such as lidar and vision cameras, which are highly sensitive to ambient dust concentration. When a vehicle travels to an area with high dust concentration, the number of point clouds in the lidar image decreases significantly due to scattering and absorption by dust particles, leading to a decline in the contrast and clarity of the visual image. However, dust generation intensity varies significantly across different working faces: tunneling and longwall faces generate substantial amounts of dust continuously during mining operations, resulting in a much higher normalized dust concentration than transport roadways. If the absolute value of dust concentration is directly used as the input indicator for sensing degradation, the normalized value will remain at a high level for an extended period in normally high-dust environments such as tunneling faces. This leads to oversensitivity in sensing degradation assessments when no safety response is required, resulting in frequent deceleration or braking and reduced transport efficiency.

[0014] In one embodiment, the dust concentration normalization module 11 is also used to collect the dust concentration value of the current environment of the vehicle. Get the working face type of the current roadway; The dust concentration benchmark value corresponding to the working surface type is obtained by querying the preset mapping table between working surface type and dust concentration benchmark value; Calculate the ratio of the dust concentration value to the dust concentration reference value to obtain the normalized dust concentration value; The working face type includes at least one of the following: tunneling face, longwall face, and haulage roadway.

[0015] Specifically, the dust concentration value of the vehicle's current environment is collected. This dust concentration value is collected in real time by a dust concentration sensor mounted on the autonomous vehicle. The dust concentration sensor is installed at the front air intake grille of the vehicle. Depending on its specific model and factory specifications, the dust concentration sensor continuously collects the dust concentration value of the surrounding environment at a corresponding sampling frequency. For example, the dust concentration sensor is a laser scattering type dust sensor, and according to the factory specifications of this model, its sampling frequency is once every 2 seconds, and the unit of dust concentration is milligrams per cubic meter.

[0016] The system retrieves the working face type corresponding to the vehicle's current location in the roadway. The working face type is determined by matching the vehicle's current location coordinates with the underground roadway electronic map. The underground roadway electronic map pre-stores the roadway number, start and end coordinates, and working face type for each roadway. Working face types include at least one of the following: tunneling face, longwall face, and haulage roadway. Tunneling faces are work areas where roadway excavation is underway, generating the highest dust intensity; longwall faces are work areas where coal mining is being carried out, generating the next highest dust intensity; haulage roadways are primarily used for transporting coal and materials, generating the lowest dust intensity. The vehicle obtains its current location coordinates through its own positioning system and then queries the underground roadway electronic map for the roadway number and corresponding working face type at those coordinates.

[0017] The dust concentration benchmark value corresponding to a specific working face type is retrieved from a pre-defined mapping table of working face types and dust concentration benchmark values. This mapping table is an indexed table based on working face type, with each working face type corresponding to a specific dust concentration benchmark value. The dust concentration benchmark value is determined by obtaining dust concentration monitoring data for multiple historical monitoring periods under normal operating conditions and without any abnormal dust events for that working face type, and taking the arithmetic mean of all monitoring data as the dust concentration benchmark value. The arithmetic mean reflects the typical dust level of that working face type under normal operating conditions. The mapping table is established using the above method before initial use and updated with the latest monitoring data after each complete monitoring period, dynamically adjusting the benchmark value according to changes in the actual dust level of the working face. For example, from the pre-defined mapping table of working face types and dust concentration benchmark values, it is found that the vehicle is currently in a transport roadway, and the corresponding dust concentration benchmark value for the transport roadway is 2.6 mg / m³. 3 .

[0018] Calculate the ratio of the dust concentration value to the baseline dust concentration value to obtain the normalized dust concentration value. Normalized dust concentration value = Dust concentration value ÷ Baseline dust concentration value. The normalized dust concentration value reflects the degree of deviation of the current dust concentration from the normal dust level of the working face. The closer the normalized value is to 1, the more comparable the current dust concentration is to the normal level of the working face; a normalized value greater than 1 indicates that the current dust concentration is higher than the normal level, and the sensing system may be subject to additional degradation; a normalized value less than 1 indicates that the current dust concentration is lower than the normal level.

[0019] For example, the current dust concentration collected by the vehicle is 4.8 mg / m³. 3 The current roadway is a transport roadway, with a baseline value of 2.6 mg / m³. 3 The normalized dust concentration value is approximately 1.85 (4.8 ÷ 2.6 ≈ 1.85). This value is greater than 1, indicating that the current dust concentration is about 1.85 times the normal level in the transport roadway, which is high. The sensing system may be affected by some degree of degradation. If the vehicle is currently at the tunneling face, the baseline value is 12.5 mg / m³. 3 The same dust concentration value of 4.8 mg / m³ 3 The calculated normalized value is 0.38, which is less than 1, and is considered low under the normal high dust conditions at the tunneling face.

[0020] The following technical effects have been achieved through this module: By introducing a dust concentration benchmark value corresponding to the type of working face, the absolute value of dust concentration is converted into a normalized value relative to the normal dust level of that working face, so that the dust concentration assessment can be adapted to the actual dust generation characteristics of different working faces.

[0021] The perception degradation calculation module 12 is used to calculate and obtain the lidar perception degradation index and the visual perception degradation index based on the normalized dust concentration value and the perception performance index of lidar and image recognition, respectively. Unmanned vehicles in coal mines rely on lidar and vision cameras for environmental perception and obstacle detection. However, dust affects these two sensors differently: lidar detects targets by emitting laser beams and receiving reflected echoes; dust particles scatter and absorb the laser, reducing the number of effective echo points and shortening the detection distance. Vision cameras distinguish obstacles from the background through image contrast; dust reduces the overall contrast and clarity of the image. The degradation degree of each sensor needs to be quantified separately. Furthermore, higher dust concentrations result in more severe actual perception degradation for the same initial degradation level. Therefore, a dust concentration normalization value is needed to correct the original degradation degree to reflect the nonlinear accelerated degradation effect of dust on sensor performance.

[0022] In one implementation, the perception degradation calculation module 12 is also used to obtain the total number of original point cloud points within the preset effective detection distance range of the lidar in the current frame point cloud data, as well as the number of effective obstacle point cloud points after ground point cloud filtering and voxel filtering. Calculate the ratio of the effective obstacle point cloud points to the total number of points in the original point cloud to obtain the current effective point cloud ratio; The baseline effective point cloud ratio of the lidar within the preset effective detection range under dust-free baseline conditions is obtained. The original degradation degree of the lidar is calculated based on the baseline effective point cloud ratio, wherein the original degradation degree of the lidar is negatively correlated with the current effective point cloud ratio. The original degradation degree of the lidar is corrected by using the normalized value of the dust concentration to obtain the lidar perception degradation index; Obtain the image contrast value of the visual camera in the current frame image, and obtain the baseline contrast value of the visual camera under dust-free baseline conditions; The original visual degradation degree is calculated based on the image contrast value and the reference contrast value, wherein the original visual degradation degree and the image contrast value are negatively correlated. The original visual degradation degree is corrected by using the normalized value of the dust concentration to obtain the visual perception degradation index.

[0023] Specifically, first, the LiDAR perception degradation index is calculated: Obtain the total number of original point cloud points within the preset effective detection range of the LiDAR in the current frame point cloud data. The preset effective detection range is determined comprehensively based on the safe braking distance of the unmanned vehicle in the underground tunnel and the effective detection capability of the LiDAR. The method for determining the preset effective detection range is as follows: take the emergency braking distance of the vehicle at the current driving speed, and multiply the emergency braking distance by a preset distance expansion coefficient to obtain the upper limit of the preset effective detection range. The preset distance expansion coefficient is set to 2, so that the detection range covers the area in front of the vehicle that is twice the braking distance, providing sufficient time margin for collision risk assessment. For example, if the vehicle's current driving speed is 5 meters per second and the emergency braking distance is 14 meters, the upper limit of the preset effective detection range = 14 × 2 = 28 meters, and the effective detection range is from 0.5 meters to 28 meters.

[0024] Within the effective detection range, the total number of points in the original point cloud data of the current frame is counted and recorded as the total number of points in the original point cloud. Ground point cloud filtering and voxel filtering are then performed on the original point cloud data. Ground point cloud filtering uses a random sampling consistency plane fitting method to remove point cloud points belonging to the roadway surface, retaining non-ground point cloud points. Voxel filtering uses a spatial downsampling method, dividing the filtered point cloud space into a cubic grid according to a preset voxel size, retaining one point cloud point in each grid, to eliminate density inhomogeneity between near-distance high-density point clouds and far-distance sparse point clouds.

[0025] The method for determining the preset voxel size is as follows: Take the larger of the horizontal and vertical angular resolutions of the lidar, denoted as the angular resolution; take the median distance of the preset effective detection range, denoted as the reference distance. Voxel size = reference distance × tan(angular resolution) × preset voxel coefficient. The preset voxel coefficient is set to 2, ensuring that each voxel grid contains at least two sampling points of adjacent laser scan lines at the reference distance, thus maintaining the integrity of the point cloud at the maximum distance within the effective detection range. For example, if the lidar's horizontal angular resolution is 0.2 degrees and its vertical angular resolution is 0.1 degrees, and the larger angular resolution of 0.2 degrees is used, the preset effective detection range is 0.5 meters to 28 meters, and the reference distance is 14.25 meters, then the voxel size = 14.25 × tan(0.2°) × 2 = 14.25 × 0.00349 × 2 ≈ 0.099 meters, rounded up to 0.1 meters. After ground point cloud filtering and voxel filtering, the number of remaining effective obstacle point cloud points is counted and recorded as the effective obstacle point cloud point number.

[0026] Calculate the ratio of the number of effective obstacle point clouds to the total number of points in the original point cloud to obtain the current effective point cloud ratio. Current effective point cloud ratio = Number of effective obstacle point clouds ÷ Total number of points in the original point cloud. The current effective point cloud ratio reflects the proportion of original points in the lidar point cloud that belong to effective obstacle echo signals in the current dusty environment.

[0027] The baseline effective point cloud ratio of the lidar is obtained within a preset effective detection range under dust-free baseline conditions. The dust-free baseline conditions refer to the lidar's detection data of standard obstacles under well-ventilated underground tunnels with extremely low dust concentrations. The baseline effective point cloud ratio is obtained as follows: Under dust-free baseline conditions, multiple frames of point cloud data are collected, using the same preset effective detection range as the current frame. This preset effective detection range is the effective detection range determined when calculating the lidar's perception degradation index for the current frame. Simultaneously, the same ground point cloud filtering and voxel filtering processing procedures as the current frame are applied, and the arithmetic mean of the current effective point cloud ratios for each frame is taken as the baseline effective point cloud ratio. The baseline effective point cloud ratio reflects the lidar's optimal perception performance level under ideal conditions. For example, the baseline effective point cloud ratio is 0.82.

[0028] The original degradation degree of the LiDAR is calculated based on the baseline effective point cloud ratio. The original degradation degree of the LiDAR is calculated as: 1 - (Current effective point cloud ratio ÷ Baseline effective point cloud ratio). When the current effective point cloud ratio equals the baseline effective point cloud ratio, the original degradation degree of the LiDAR is zero, indicating no degradation in sensing performance. When the current effective point cloud ratio is significantly lower than the baseline effective point cloud ratio, the original degradation degree of the LiDAR is close to 1, indicating severe degradation in sensing performance. The original degradation degree of the LiDAR is negatively correlated with the current effective point cloud ratio; the closer the current effective point cloud ratio is to the baseline value, the smaller the original degradation degree. For example, if the total number of points in the original point cloud of the current frame is 12,500, and the number of effective obstacle point cloud points after filtering is 7,850, then the current effective point cloud ratio is 7850 ÷ 12,500 = 0.628. The baseline effective point cloud ratio is 0.82. The original degradation degree of the LiDAR is: 1 - (0.628 ÷ 0.82) = 1 - 0.766 ≈ 0.234.

[0029] The original degradation degree of the LiDAR is corrected using a normalized dust concentration value to obtain the LiDAR perception degradation index. The LiDAR perception degradation index = original LiDAR degradation degree × normalized dust concentration value. The normalized dust concentration value is output by the dust concentration normalization module, which amplifies and corrects the original degradation degree. Higher dust concentrations result in a larger perception degradation index for the same original degradation degree. For example, if the normalized dust concentration value is 1.85 and the original LiDAR degradation degree is 0.234, the LiDAR perception degradation index = 0.234 × 1.85 ≈ 0.433.

[0030] Then, the visual perception degradation index is calculated: Obtain the image contrast value of the visual camera in the current frame. The image contrast value is calculated using the root mean square contrast method. The calculation method is as follows: subtract the average gray value of the entire image from the gray value of all pixels in the current frame to obtain the gray value deviation of each pixel. Take the square root of the arithmetic mean of the squares of the gray value deviations of all pixels as the image contrast value. The root mean square contrast reflects the magnitude of gray value changes in the image. The higher the image contrast value, the more significant the difference in brightness between different areas in the image, and the clearer the outline and texture of obstacles.

[0031] Obtain the baseline contrast value of the vision camera under dust-free baseline conditions. The baseline contrast value is obtained as follows: Under dust-free baseline conditions, the vision camera captures images of a standard scene. Using the same root mean square contrast calculation method as the current frame, multiple frames are acquired, and the arithmetic mean of the contrast values ​​of each frame is taken as the baseline contrast value. For example, the calculated baseline contrast value is 0.65.

[0032] The original visual degradation is calculated based on the image contrast value and the reference contrast value. Original visual degradation = 1 - (current image contrast value ÷ reference contrast value). When the current image contrast value equals the reference contrast value, the original visual degradation is zero, indicating no degradation in perceptual performance; when the current image contrast value is significantly lower than the reference contrast value, the original visual degradation is close to 1, indicating severe degradation in perceptual performance. The original visual degradation is negatively correlated with the image contrast value; the closer the image contrast value is to the reference value, the smaller the original degradation. For example, if the image contrast value of the current frame is 0.38 and the reference contrast value is 0.65, the original visual degradation = 1 - (0.38 ÷ 0.65) = 1 - 0.585 ≈ 0.415.

[0033] The original visual degradation degree was corrected using a normalized dust concentration value to obtain the visual perception degradation index. Visual perception degradation index = original visual degradation degree × normalized dust concentration value. For example, if the normalized dust concentration value is 1.85 and the original visual degradation degree is 0.415, the visual perception degradation index = 0.415 × 1.85 ≈ 0.768.

[0034] For example, Figure 3 This is an example table of data collection for a safety control system for an unmanned vehicle in a coal mine, provided by an embodiment of the present invention. It includes raw data such as dust concentration, lidar point cloud, and visual images under different working face types.

[0035] The following technical effects have been achieved through this module: The LiDAR perception degradation index and the visual perception degradation index were calculated separately. The performance indicators of the two sensors were used to quantify the different mechanisms by which dust affects them. The LiDAR was based on the effective point cloud ratio, and the visual camera was based on the image contrast ratio. This can more accurately characterize the actual degree of degradation of different sensors in the current dusty environment.

[0036] The original degradation degree is amplified and corrected by using the normalized value of dust concentration. The higher the dust concentration, the greater the correction magnitude, so that the sensing degradation index can better reflect the nonlinear accelerated degradation effect of dust on sensor performance, and provide a more reliable degradation assessment input for subsequent calculation of fusion sensing credibility.

[0037] The fusion credibility calculation module 13 is used to calculate the fusion perception credibility based on the lidar perception degradation index and the visual perception degradation index. Unmanned vehicles in coal mines rely on multi-sensor fusion for environmental perception. Different environmental conditions can lead to inconsistent degradation levels between two sensors: for example, in areas with high dust concentration but sufficient illumination in the tunnel, lidar may suffer severe degradation due to dust scattering, while the vision camera maintains good perception capabilities due to sufficient illumination; in areas with moderate dust concentration but heavy water mist, both sensors may experience moderate degradation simultaneously. In this situation, the reliability of the entire perception system does not depend on the best-performing sensor, but rather on the sensor with the most severe degradation.

[0038] In one implementation, the fusion credibility calculation module 13 is further configured to use the larger value of the lidar perception degradation index and the visual perception degradation index as the dominant degradation index, and the smaller value as the auxiliary degradation index. Calculate 1 minus the dominant degradation index, and use the difference as the dominant sensor availability. Also calculate 1 minus the auxiliary degradation index to obtain the auxiliary sensor availability. The reliability of fused perception is obtained by multiplying the availability of the dominant sensor by the availability of the auxiliary sensor.

[0039] Specifically, the larger of the LiDAR perception degradation index and the visual perception degradation index is used as the dominant degradation index, and the smaller value is used as the auxiliary degradation index. Both the LiDAR perception degradation index and the visual perception degradation index are output by the perception degradation calculation module. The values ​​of both degradation indices range from 0 to 1, with larger values ​​indicating more severe performance degradation of the corresponding sensor. The dominant degradation index reflects the limiting effect of a highly degraded sensor on the overall perception capability of the system, while the auxiliary degradation index reflects the usability of a less degraded sensor as a supplementary perception method.

[0040] Calculate 1 minus the dominant degradation index; the difference is taken as the dominant sensor availability. Dominant sensor availability = 1 - dominant degradation index. Dominant sensor availability represents the proportion of sensors that have severely degraded in the current dusty environment that can still function normally. When the dominant degradation index is close to 0, the dominant sensor availability is close to 1, indicating that the performance of a severely degraded sensor is still good; when the dominant degradation index is close to 1, the dominant sensor availability is close to 0, indicating that the severely degraded sensor has essentially failed.

[0041] Calculate 1 minus the auxiliary degradation index; the difference is taken as the auxiliary sensor availability. Auxiliary sensor availability = 1 - Auxiliary degradation index. Auxiliary sensor availability indicates the degree to which a sensor with a lower degree of degradation can be used as a supplementary sensing method.

[0042] The fusion sensing reliability is obtained by multiplying the availability of the primary sensor by the availability of the secondary sensor. Fusion sensing reliability = Primary sensor availability × Secondary sensor availability. Both availability values ​​are between 0 and 1, and the product is less than or equal to either availability value. When one sensor is severely degraded, its availability is close to 0, and the product is also close to 0, reflecting the "weakest link effect," meaning the worst-performing sensor in the sensing system determines the lower limit of the overall sensing capability. When both sensors have low degradation levels, both availability values ​​are close to 1, and the product is also close to 1, indicating that the overall sensing system is highly reliable in the current dusty environment.

[0043] For example, the LiDAR perception degradation index output by the perception degradation calculation module is 0.433, and the visual perception degradation index is 0.768. The larger value, 0.768, is taken as the dominant degradation index; the smaller value, 0.433, is taken as the auxiliary degradation index. The dominant sensor availability = 1 - 0.768 = 0.232, and the auxiliary sensor availability = 1 - 0.433 = 0.567. The fusion perception reliability = 0.232 × 0.567 ≈ 0.132. The fusion perception reliability of 0.132, close to 0, indicates that in the current dusty environment, the overall perception system is highly unreliable due to severe degradation of the visual camera, and subsequent collision risk assessment requires significant correction.

[0044] The following technical effects have been achieved through this module: By using the larger of the two sensor degradation indices as the dominant degradation index and the smaller as the auxiliary degradation index, the fusion perception reliability is calculated by multiplying the availability of the dominant sensor and the availability of the auxiliary sensor. This ensures that the fusion reliability never exceeds the availability of any single sensor. When either sensor is severely degraded, the fusion reliability decreases significantly, enabling subsequent collision risk assessment and safety control strategies to respond promptly to the risk of decreased perception capabilities.

[0045] The collision risk correction module 14 is used to correct the original collision risk assessment value of the obstacle detection based on the fusion perception confidence level, so as to obtain a corrected collision risk value. When autonomous vehicles operate in underground mine tunnels, they need to perform collision risk assessments based on obstacle information detected by sensors to determine whether safety control measures should be triggered. However, the original collision risk assessment value is calculated based on the obstacle distance and relative speed output by the sensors, and its accuracy depends on the reliability of the perception data. When the reliability of fused perception is low, the obstacle distance and relative speed detected by the sensors may have significant deviations. The original collision risk assessment value calculated directly using this data will be unreliable and may underestimate the actual collision risk. Therefore, it is necessary to correct the original collision risk assessment value based on the reliability of fused perception, amplifying the collision risk value when perception is unreliable, and ensuring that the safety control strategy can take more conservative measures in advance when perception degradation is severe.

[0046] In one embodiment, the collision risk correction module 14 is further configured to obtain a perception uncertainty based on the fused perception confidence, wherein the perception uncertainty is inversely proportional to the fused perception confidence; Based on the perceived uncertainty, the original collision risk assessment value is corrected to obtain the corrected collision risk value; The acquisition of the original collision risk assessment value includes: The distance and relative speed of the nearest obstacle detected by the lidar and image recognition are obtained, as well as the current vehicle speed and the preset emergency braking safety distance. Calculate the ratio of the distance value to the emergency braking safe distance to obtain the distance-risk ratio; Based on the distance-risk ratio, a distance-risk factor is calculated, wherein when the distance value is less than the emergency braking safe distance, the distance-risk factor is set to 1; Calculate the ratio of the relative velocity value to the preset collision reference velocity to obtain the velocity risk factor; The distance risk factor and the speed risk factor are weighted and calculated to obtain the original collision risk assessment value.

[0047] Specifically, first, obtain the original collision risk assessment value: The system acquires the distance and relative velocity values ​​of the nearest obstacles detected by LiDAR and image recognition. The LiDAR and vision camera independently detect obstacles in each frame of data, with each sensor outputting a list of detected obstacles. Each obstacle includes attributes such as distance, azimuth, and relative velocity. The obstacle lists from the two sensors are then fused and correlated. Obstacles with similar distance and azimuth values ​​in both lists are considered the same obstacle. The smaller distance value detected by both sensors for the same obstacle is taken as the combined distance value for that obstacle, and the arithmetic mean of the relative velocities detected by both sensors is taken as the combined relative velocity value for that obstacle.

[0048] For all obstacles after fusion and association, the obstacle with the smallest distance value is selected as the nearest obstacle. Its distance and relative velocity values ​​are extracted and used as the obstacle distance and relative velocity values ​​for this calculation. The obstacle distance value is the distance of the obstacle relative to the vehicle's current position, in meters; the obstacle relative velocity value is the relative speed between the obstacle and the vehicle, in meters per second. A positive relative velocity value indicates that the obstacle is approaching the vehicle, and a negative value indicates that the obstacle is moving away from the vehicle.

[0049] Obtain the vehicle's current speed. The speed value is obtained from the vehicle's own speed sensor, in meters per second. Obtain the preset emergency braking safety distance. The emergency braking safety distance is dynamically determined based on the vehicle's current speed and braking performance parameters. The emergency braking safety distance is determined by obtaining the vehicle's maximum braking deceleration, which is the deceleration value the vehicle can achieve when braking with maximum braking force under dry road conditions in an underground mine tunnel. Emergency braking safety distance = vehicle current speed × driver reaction time + vehicle current speed squared ÷ (2 × maximum braking deceleration), where the driver reaction time is taken as 0.5 seconds. For example, if the vehicle's current speed is 5 meters per second and the maximum braking deceleration is 2.5 meters per second squared, the emergency braking safety distance = 5 × 0.5 + 25 ÷ (2 × 2.5) = 2.5 + 5 = 7.5 meters. The higher the speed or the lower the braking deceleration, the greater the emergency braking safety distance.

[0050] Calculate the ratio of obstacle distance to emergency braking safe distance to obtain the distance-risk ratio. Distance-risk ratio = obstacle distance ÷ emergency braking safe distance. A smaller distance-risk ratio indicates that the obstacle is closer or has entered the emergency braking range, and the risk of collision is higher; a larger distance-risk ratio indicates that the obstacle is farther away, and the risk of collision is lower.

[0051] The distance risk factor is calculated based on the distance-risk ratio. Distance risk factor = 1 - distance-risk ratio, with a value of 0 when the result is less than 0. When the distance-risk ratio is greater than or equal to 1, it indicates that the obstacle distance is greater than or equal to the emergency braking safe distance, and the distance risk factor is 0; when the distance-risk ratio is less than 1, it indicates that the obstacle has entered the emergency braking safe distance range, and the distance risk factor is greater than 0; when the obstacle distance is 0, the distance-risk ratio is 0, and the distance risk factor is 1. For example, if the obstacle distance is 4.5 meters and the emergency braking safe distance is 7.5 meters, the distance-risk ratio = 4.5 ÷ 7.5 = 0.60, and the distance risk factor = 1 - 0.60 = 0.40.

[0052] The speed risk factor is obtained by calculating the ratio of the relative speed value to the preset collision reference speed. Speed ​​risk factor = relative speed value ÷ preset collision reference speed. A positive relative speed value indicates that the obstacle is approaching the vehicle, while a negative value indicates that the obstacle is moving away from the vehicle. When the relative speed value is negative, the speed risk factor is 0, indicating that the obstacle is moving away and the collision risk is lowest. When the calculated result is greater than 1, the speed risk factor is 1. The preset collision reference speed is determined by taking the larger value between the vehicle's current speed and the preset lower limit of relative speed. The preset lower limit of relative speed is set to 1 meter per second to avoid the speed risk factor being unreasonably amplified due to a collision reference speed of zero or too small. For example, if the preset collision reference speed is 5 meters per second and the obstacle's relative speed is 3 meters per second, the speed risk factor = 3 ÷ 5 = 0.60.

[0053] The distance risk factor and speed risk factor are weighted and calculated to obtain the original collision risk assessment value. Original collision risk assessment value = Distance risk weight × Distance risk factor + Speed ​​risk weight × Speed ​​risk factor. The distance risk weight and speed risk weight are determined as follows: obtain the correlation coefficients between the distance risk factor and speed risk factor and the collision result in historical collision events, and take the ratio of the absolute value of each correlation coefficient to the sum of the absolute values ​​of the two correlation coefficients as the weight coefficient. For example, if the correlation coefficient between the distance risk factor and the collision result is 0.78, and the correlation coefficient between the speed risk factor and the collision result is 0.52, then the distance risk weight = 0.78 ÷ (0.78 + 0.52) = 0.78 ÷ 1.30 = 0.60, and the speed risk weight = 0.52 ÷ 1.30 = 0.40. The original collision risk assessment value = 0.60 × 0.40 + 0.40 × 0.60 = 0.24 + 0.24 = 0.48.

[0054] Then, the original collision risk assessment value is corrected based on the credibility of the fused perception: Based on the fusion perception confidence level, the perception uncertainty is obtained. Perception uncertainty = 1 - fusion perception confidence level. The fusion perception confidence level is output by the fusion confidence level calculation module, and its value ranges from 0 to 1. The higher the fusion perception confidence level, the lower the perception uncertainty; the two are inversely proportional. For example, if the fusion confidence level calculation module outputs a fusion perception confidence level of 0.132, the perception uncertainty = 1 - 0.132 = 0.868.

[0055] The original collision risk assessment value is corrected based on perceived uncertainty to obtain a corrected collision risk value. Corrected collision risk value = Original collision risk assessment value + Perceived uncertainty × Original collision risk assessment value = Original collision risk assessment value × (1 + Perceived uncertainty). The larger the perceived uncertainty, the larger the correction magnitude, and the greater the amplification of the original collision risk assessment value. When perception is unreliable, even if the original assessment considers the collision risk to be low, the corrected risk value is significantly amplified to trigger a higher level of safety response. For example, if the original collision risk assessment value is 0.48 and the perceived uncertainty is 0.868, the corrected collision risk value = 0.48 × (1 + 0.868) = 0.48 × 1.868 ≈ 0.897. The corrected risk value of 0.897 is significantly higher than the original risk value of 0.48, reflecting a conservative assessment of collision risk when perception is highly unreliable.

[0056] The following technical effects have been achieved through this module: The original collision risk assessment value is obtained by weighted fusion of distance and speed risk factors, integrating collision risk information from two dimensions: obstacle distance and relative speed. The original collision risk assessment value is amplified and corrected by perception uncertainty; the greater the perception uncertainty, the larger the correction. This automatically increases the conservatism of the collision risk assessment when perception degradation is severe, providing subsequent safety control strategies with risk input adjusted for perception reliability, thus improving the reliability of safety decisions for unmanned underground vehicles in complex dust environments.

[0057] The safety control strategy module 15 is used to determine the current safety control strategy of the vehicle based on the fused perception confidence level and the corrected collision risk value.

[0058] When determining a vehicle's safety control strategy, both risk level and perception reliability need to be considered. When perception reliability is high, the corrected collision risk value serves directly as the trigger for safety control, and the vehicle takes appropriate safety measures according to the normal risk level. When perception reliability is low, sensor data itself is unreliable. If the response is still based on the normal risk level, safety measures may be delayed due to perception blind spots. Therefore, it is necessary to further lower the threshold for triggering safety control, so that even a lower corrected risk value can trigger a higher level of safety measures. Simultaneously, the more severe the perception degradation, the lower the permissible vehicle speed should be to allow for more reaction time.

[0059] In one embodiment, the safety control strategy module 15 is further configured to use the perceived uncertainty to correct the preset first risk trigger threshold and second risk trigger threshold, and obtain the audible and visual warning trigger threshold and the automatic deceleration trigger threshold. When the corrected collision risk value is greater than or equal to the sound and light warning trigger threshold, the first-level sound and light warning is triggered, projecting a warning beam into the alleyway in front of the vehicle and sounding a warning sound. When the corrected collision risk value is greater than or equal to the automatic deceleration trigger threshold, the second-level automatic deceleration is triggered to reduce the vehicle's current speed to the target speed. The target driving speed is compared with the minimum safe driving speed. When the target driving speed is lower than the minimum safe driving speed, the minimum safe driving speed is taken as the final driving speed.

[0060] Obtaining the minimum safe driving speed and the target driving speed includes: Based on the current environment, a baseline safe driving speed is obtained through mapping. An environmental clarity factor is calculated based on the normalized dust concentration value, wherein the environmental clarity factor is inversely proportional to the normalized dust concentration value. Multiply the baseline safe driving speed by the environmental clarity factor, and the product is taken as the minimum safe driving speed. The perception degradation degree is calculated based on the fusion perception credibility, and the driving speed value is multiplied by the perception degradation degree to obtain the perception constraint speed reduction. Subtract the perceived speed reduction from the driving speed value, and the difference is taken as the target driving speed. When the target driving speed is lower than the preset emergency braking trigger speed, the vehicle emergency braking is triggered.

[0061] Specifically, first, obtain the current vehicle's safety control trigger threshold: The system uses a perception uncertainty correction to adjust for preset first and second risk trigger thresholds, thereby obtaining the audible and visual warning trigger thresholds and the automatic deceleration trigger thresholds. The perception uncertainty is output by the collision risk correction module and ranges from 0 to 1. A higher perception uncertainty indicates lower fusion perception reliability and a less reliable perception system, necessitating a reduction in the safety control trigger thresholds so that even a lower corrected collision risk value can trigger a safety response. The first and second risk trigger thresholds are preset fixed values. The first risk trigger threshold corresponds to the audible and visual warning trigger level, and the second risk trigger threshold corresponds to the automatic deceleration trigger level. The value of the second risk trigger threshold is higher than that of the first risk trigger threshold.

[0062] The method for determining the first and second risk trigger thresholds is as follows: Obtain the original collision risk assessment values ​​calculated for each detection cycle during the vehicle's historical normal operation. Take the 95th percentile of all original collision risk assessment values ​​as the first risk trigger threshold, and the 99th percentile as the second risk trigger threshold. The 95th percentile indicates that the original collision risk assessment value is lower than this value in approximately 95% of the normal operation cycles; collisions exceeding this value are considered to have a relatively high risk level. The 99th percentile indicates that the original collision risk assessment value is lower than this value in approximately 99% of the normal operation cycles; collisions exceeding this value are considered to have an extremely high risk level. For example, the first risk trigger threshold is 0.35, and the second risk trigger threshold is 0.65.

[0063] The audible and visual warning trigger threshold is calculated as follows: First Risk Trigger Threshold × (1 - Perception Uncertainty). When the perception uncertainty is zero, the audible and visual warning trigger threshold equals the first risk trigger threshold. When the perception uncertainty is close to 1, the audible and visual warning trigger threshold is close to zero, indicating that even a very low corrected collision risk value is sufficient to trigger the audible and visual warning when the perception is highly unreliable. The automatic deceleration trigger threshold is calculated as follows: Second Risk Trigger Threshold × (1 - Perception Uncertainty), with the same correction method.

[0064] For example, the first risk trigger threshold is 0.35, the second risk trigger threshold is 0.65, and the perception uncertainty is 0.868. The audible and visual warning trigger threshold = 0.35 × (1 - 0.868) = 0.35 × 0.132 ≈ 0.046. The automatic deceleration trigger threshold = 0.65 × 0.132 ≈ 0.086. The corrected threshold values ​​are much lower than the original threshold values; even when the perception is highly unreliable, a lower risk value can trigger a safety response.

[0065] Then, the corrected collision risk value is compared with the trigger threshold to determine the safety control response level: When the corrected collision risk value is greater than or equal to the trigger threshold for the audible and visual warning, the first-level audible and visual warning is triggered. The audible and visual warning measures include: projecting a bright warning beam of light (flashing yellow) into the roadway ahead of the vehicle, while simultaneously emitting a continuous buzzing warning sound at a frequency of 3 times per second and a sound pressure level of no less than 85 decibels. The purpose of the audible and visual warning is to alert pedestrians and workers in the roadway to the approaching vehicle, and to simultaneously alert the remote monitoring center to the vehicle's operational status.

[0066] When the corrected collision risk value is greater than or equal to the automatic deceleration trigger threshold, the second level of automatic deceleration is triggered. The automatic deceleration measure is to calculate the target speed based on the current degree of perception degradation and smoothly reduce the vehicle's current speed to the target speed.

[0067] Calculate the target driving speed. Obtain the perception degradation degree based on the fusion perception confidence level. Perception degradation degree = 1 - fusion perception confidence level. The lower the fusion perception confidence level, the higher the perception degradation degree, and the greater the speed reduction required. Multiply the driving speed value by the perception degradation degree to obtain the perception constraint speed reduction amount. Perception constraint speed reduction amount = current driving speed × perception degradation degree. Subtract the perception constraint speed reduction amount from the driving speed value; the difference is the target driving speed. Target driving speed = current driving speed - perception constraint speed reduction amount = current driving speed × (1 - perception degradation degree) = current driving speed × fusion perception confidence level. When the fusion perception confidence level is close to 1, the target driving speed is close to the current driving speed, with almost no speed reduction; when the fusion perception confidence level is close to 0, the target driving speed is close to 0.

[0068] For example, the vehicle's current speed is 5 meters per second, the fusion perception confidence is 0.132, the perception degradation is 1-0.132=0.868, and the target speed is 5×0.132=0.66 meters per second.

[0069] Calculate the minimum safe driving speed. Based on the current environment, obtain the baseline safe driving speed through mapping. The baseline safe driving speed is determined according to the working face type of the roadway the vehicle is currently in. The preset mapping table between working face types and baseline safe driving speeds is established by obtaining the maximum walking speed of each working face type under normal operating conditions, and multiplying the maximum walking speed by a preset safety factor as the baseline safe driving speed. The preset safety factor is set to 1.2, ensuring that the minimum vehicle speed is slightly higher than the walking speed of pedestrians, thus balancing transportation efficiency with safety. For example, the baseline safe driving speed for a haulage roadway is 1.5 meters per second, the baseline safe driving speed for a tunneling face is 1.0 meter per second, and the baseline safe driving speed for a longwall face is 1.2 meters per second. The vehicle is currently in a haulage roadway, and the baseline safe driving speed is 1.5 meters per second.

[0070] The environmental clarity factor is calculated based on the normalized dust concentration value. Environmental clarity factor = 1 ÷ (1 + normalized dust concentration value). The environmental clarity factor ranges from 0 to 1. The larger the normalized dust concentration value, the smaller the environmental clarity factor, and the lower the minimum safe driving speed. For example, if the normalized dust concentration value is 1.85, the environmental clarity factor = 1 ÷ (1 + 1.85) = 1 ÷ 2.85 ≈ 0.351. The minimum safe driving speed is obtained by multiplying the baseline safe driving speed by the environmental clarity factor. Minimum safe driving speed = 1.5 × 0.351 ≈ 0.527 meters per second.

[0071] The target driving speed is compared with the minimum safe driving speed. If the target driving speed is lower than the minimum safe driving speed, the minimum safe driving speed is taken as the final driving speed. For example, if the target driving speed is 0.66 meters per second and the minimum safe driving speed is 0.527 meters per second, and the target driving speed is higher than the minimum safe driving speed, the target driving speed of 0.66 meters per second is taken as the final driving speed.

[0072] For example, the collision risk correction module outputs a corrected collision risk value of 0.897. This value is greater than the audible and visual warning trigger threshold of 0.046, triggering the first-level audible and visual warning; at the same time, it is greater than the automatic deceleration trigger threshold of 0.086, triggering the second-level automatic deceleration, reducing the vehicle speed from 5 meters per second to 0.66 meters per second.

[0073] When the acquired target speed is lower than the preset emergency braking trigger speed, the vehicle's emergency braking is triggered. The preset emergency braking trigger speed is determined by multiplying the minimum safe driving speed by a preset emergency braking coefficient. The emergency braking coefficient is determined by obtaining the braking distance required for the vehicle to come to a complete stop in the underground tunnel from the start of braking, denoted as the emergency braking distance. The minimum detection distance of the lidar under dust-free baseline conditions is denoted as the baseline detection distance. The emergency braking coefficient = emergency braking distance ÷ (emergency braking distance + baseline detection distance). When the baseline detection distance is larger than the emergency braking distance, the emergency braking coefficient is smaller, indicating that the vehicle has sufficient reaction distance when perception is normal, and the emergency braking trigger speed can be set lower. When the emergency braking distance is larger, the emergency braking coefficient approaches 1, indicating that the vehicle needs to be driven at a higher speed to trigger braking to ensure safety.

[0074] For example, when the vehicle's current speed is 5 meters per second, the emergency braking distance is 7.5 meters. The minimum detection range of the lidar under dust-free baseline conditions is 8 meters. The emergency braking coefficient = 7.5 ÷ (7.5 + 8) = 7.5 ÷ 15.5 ≈ 0.484, rounded up to 0.5. For example, the minimum safe driving speed is 0.527 meters per second, and the emergency braking trigger speed = 0.527 × 0.5 ≈ 0.264 meters per second. If the target driving speed is 0.2 meters per second, which is lower than 0.264 meters per second, emergency braking is triggered, and the vehicle stops immediately with maximum braking deceleration.

[0075] The following technical effects have been achieved through this module: The safety control trigger threshold is corrected using perception uncertainty; the higher the perception uncertainty, the lower the threshold. Compared with existing methods that use a fixed threshold, this approach can trigger a safety response earlier when perception degradation is severe, allowing for more sufficient reaction time. The target driving speed is calculated using fused perception reliability; the lower the fused perception reliability, the lower the target driving speed. This allows for dynamic adjustment of the driving speed based on the degree of perception degradation, ensuring that the vehicle travels at a safer speed when perception capabilities are insufficient.

[0076] Example 2, as Figure 4 As shown, based on the same inventive concept provided in Embodiment 1, this embodiment of the invention also provides a safety control method for unmanned vehicles in coal mines, the method comprising: Obtain the ambient dust concentration value of the vehicle's current environment, and calculate the normalized dust concentration value based on the current environment. Based on the normalized dust concentration value, and combined with the perception performance indicators of lidar and image recognition, the lidar perception degradation index and the visual perception degradation index are calculated and obtained respectively. The credibility of fused perception is calculated based on the LiDAR perception degradation index and the visual perception degradation index. Based on the fusion perception reliability, the original collision risk assessment value of obstacle detection is corrected to obtain the corrected collision risk value; Based on the fused perception confidence level and the corrected collision risk value, the current safety control strategy for the vehicle is determined.

[0077] In one embodiment, obtaining the ambient dust concentration value of the vehicle's current environment and calculating the normalized dust concentration value based on the current environment further includes: Collect and obtain the dust concentration value of the current environment around the vehicle; Get the working face type of the current roadway; The dust concentration benchmark value corresponding to the working surface type is obtained by querying the preset mapping table between working surface type and dust concentration benchmark value; Calculate the ratio of the dust concentration value to the dust concentration reference value to obtain the normalized dust concentration value; The working face type includes at least one of the following: tunneling face, longwall face, and haulage roadway.

[0078] In one embodiment, calculating the lidar perception degradation index and the visual perception degradation index based on the normalized dust concentration value and the perception performance indicators of lidar and image recognition further includes: Obtain the total number of original point cloud points within the preset effective detection range in the current frame point cloud data of the lidar, as well as the number of effective obstacle point cloud points after ground point cloud filtering and voxel filtering; Calculate the ratio of the effective obstacle point cloud points to the total number of points in the original point cloud to obtain the current effective point cloud ratio; The baseline effective point cloud ratio of the lidar within the preset effective detection range under dust-free baseline conditions is obtained. The original degradation degree of the lidar is calculated based on the baseline effective point cloud ratio, wherein the original degradation degree of the lidar is negatively correlated with the current effective point cloud ratio. The original degradation degree of the lidar is corrected by using the normalized value of the dust concentration to obtain the lidar perception degradation index; Obtain the image contrast value of the visual camera in the current frame image, and obtain the baseline contrast value of the visual camera under dust-free baseline conditions; The original visual degradation degree is calculated based on the image contrast value and the reference contrast value, wherein the original visual degradation degree and the image contrast value are negatively correlated. The original visual degradation degree is corrected by using the normalized value of the dust concentration to obtain the visual perception degradation index.

[0079] In one embodiment, calculating the fused perception reliability based on the lidar perception degradation index and the visual perception degradation index further includes: The larger value between the lidar perception degradation index and the visual perception degradation index is used as the dominant degradation index, and the smaller value is used as the auxiliary degradation index. Calculate 1 minus the dominant degradation index, and use the difference as the dominant sensor availability. Also calculate 1 minus the auxiliary degradation index to obtain the auxiliary sensor availability. The reliability of fused perception is obtained by multiplying the availability of the dominant sensor by the availability of the auxiliary sensor.

[0080] In one embodiment, based on the fused perception reliability, the original collision risk assessment value of the obstacle detection is corrected to obtain a corrected collision risk value, which further includes: Based on the fused perception confidence level, a perception uncertainty is obtained, wherein the perception uncertainty is inversely proportional to the fused perception confidence level. Based on the perceived uncertainty, the original collision risk assessment value is corrected to obtain the corrected collision risk value; The acquisition of the original collision risk assessment value includes: The distance and relative speed of the nearest obstacle detected by the lidar and image recognition are obtained, as well as the current vehicle speed and the preset emergency braking safety distance. Calculate the ratio of the distance value to the emergency braking safe distance to obtain the distance-risk ratio; Based on the distance-risk ratio, a distance-risk factor is calculated, wherein when the distance value is less than the emergency braking safe distance, the distance-risk factor is set to 1; Calculate the ratio of the relative velocity value to the preset collision reference velocity to obtain the velocity risk factor; The distance risk factor and the speed risk factor are weighted and calculated to obtain the original collision risk assessment value.

[0081] In one embodiment, determining the vehicle's current safety control strategy based on the fused perception confidence level and the corrected collision risk value further includes: The perceived uncertainty is used to correct the preset first risk trigger threshold and second risk trigger threshold, and the sound and light warning trigger threshold and automatic deceleration trigger threshold are obtained. When the corrected collision risk value is greater than or equal to the sound and light warning trigger threshold, the first-level sound and light warning is triggered, projecting a warning beam into the alleyway in front of the vehicle and sounding a warning sound. When the corrected collision risk value is greater than or equal to the automatic deceleration trigger threshold, the second-level automatic deceleration is triggered to reduce the vehicle's current speed to the target speed. The target driving speed is compared with the minimum safe driving speed. When the target driving speed is lower than the minimum safe driving speed, the minimum safe driving speed is taken as the final driving speed. Obtaining the minimum safe driving speed and the target driving speed includes: Based on the current environment, a baseline safe driving speed is obtained through mapping. An environmental clarity factor is calculated based on the normalized dust concentration value, wherein the environmental clarity factor is inversely proportional to the normalized dust concentration value. Multiply the baseline safe driving speed by the environmental clarity factor, and the product is taken as the minimum safe driving speed. The perception degradation degree is calculated based on the fusion perception credibility, and the driving speed value is multiplied by the perception degradation degree to obtain the perception constraint speed reduction. Subtract the perceived speed reduction from the driving speed value, and the difference is taken as the target driving speed. When the target driving speed is lower than the preset emergency braking trigger speed, the vehicle emergency braking is triggered.

Claims

1. A safety control system for unmanned vehicles in coal mines, comprising: The dust concentration normalization module is used to obtain the ambient dust concentration value of the current environment of the vehicle, and calculate the normalized dust concentration value based on the current environment. The perception degradation calculation module is used to calculate and obtain the lidar perception degradation index and the visual perception degradation index based on the normalized dust concentration value and the perception performance index of lidar and image recognition, respectively. The fusion credibility calculation module is used to calculate the fusion perception credibility based on the lidar perception degradation index and the visual perception degradation index. The collision risk correction module is used to correct the original collision risk assessment value of the obstacle detection based on the fusion perception confidence level, so as to obtain a corrected collision risk value. The safety control strategy module is used to determine the current safety control strategy of the vehicle based on the fused perception confidence level and the corrected collision risk value.

2. The safety control system for an unmanned vehicle in a coal mine according to claim 1, characterized in that, The step of obtaining the ambient dust concentration value of the vehicle's current environment and calculating the normalized dust concentration value based on the current environment includes: Collect and obtain the dust concentration value of the current environment around the vehicle; Get the working face type of the current roadway; The dust concentration benchmark value corresponding to the working surface type is obtained by querying the preset mapping table between working surface type and dust concentration benchmark value; Calculate the ratio of the dust concentration value to the dust concentration benchmark value to obtain the normalized dust concentration value.

3. The safety control system for an unmanned vehicle in a coal mine according to claim 2, characterized in that, The working face type includes at least one of the following: tunneling face, longwall face, and haulage roadway.

4. The safety control system for unmanned vehicles in coal mines according to claim 1, characterized in that, Based on the normalized dust concentration value, and combined with the perception performance indicators of lidar and image recognition, the lidar perception degradation index and the visual perception degradation index are calculated and obtained respectively, including: Obtain the total number of original point cloud points within the preset effective detection range in the current frame point cloud data of the lidar, as well as the number of effective obstacle point cloud points after ground point cloud filtering and voxel filtering; Calculate the ratio of the effective obstacle point cloud points to the total number of points in the original point cloud to obtain the current effective point cloud ratio; The baseline effective point cloud ratio of the lidar within the preset effective detection range under dust-free baseline conditions is obtained. The original degradation degree of the lidar is calculated based on the baseline effective point cloud ratio, wherein the original degradation degree of the lidar is negatively correlated with the current effective point cloud ratio. The original degradation degree of the lidar is corrected by using the normalized value of the dust concentration to obtain the lidar perception degradation index; Obtain the image contrast value of the visual camera in the current frame image, and obtain the baseline contrast value of the visual camera under dust-free baseline conditions; The original visual degradation degree is calculated based on the image contrast value and the reference contrast value, wherein the original visual degradation degree and the image contrast value are negatively correlated. The original visual degradation degree is corrected by using the normalized value of the dust concentration to obtain the visual perception degradation index.

5. The safety control system for an unmanned vehicle in a coal mine according to claim 1, characterized in that, The calculation of fused perception reliability based on the lidar perception degradation index and the visual perception degradation index includes: The larger value between the lidar perception degradation index and the visual perception degradation index is used as the dominant degradation index, and the smaller value is used as the auxiliary degradation index. Calculate 1 minus the dominant degradation index, and use the difference as the dominant sensor availability. Also calculate 1 minus the auxiliary degradation index to obtain the auxiliary sensor availability. The reliability of fused perception is obtained by multiplying the availability of the dominant sensor by the availability of the auxiliary sensor.

6. The safety control system for an unmanned vehicle in a coal mine according to claim 1, characterized in that, The process of correcting the original collision risk assessment value of the obstacle detection based on the fused perception reliability to obtain a corrected collision risk value includes: Based on the fused perception confidence level, a perception uncertainty is obtained, wherein the perception uncertainty is inversely proportional to the fused perception confidence level. The original collision risk assessment value is corrected based on the perceived uncertainty to obtain a corrected collision risk value.

7. The safety control system for an unmanned vehicle in a coal mine according to claim 1, characterized in that, The acquisition of the original collision risk assessment value includes: The distance and relative speed of the nearest obstacle detected by the lidar and image recognition are obtained, as well as the current driving speed of the vehicle and the preset emergency braking safety distance. Calculate the ratio of the distance value to the emergency braking safe distance to obtain the distance-risk ratio; Based on the distance risk ratio, a distance risk factor is calculated, wherein when the distance value is less than the emergency braking safe distance, the distance risk factor is set to 1; Calculate the ratio of the relative velocity value to the preset collision reference velocity to obtain the velocity risk factor; The distance risk factor and the speed risk factor are weighted and calculated to obtain the original collision risk assessment value.

8. A safety control system for unmanned vehicles in coal mines according to claim 6, characterized in that, The step of determining the vehicle's current safety control strategy based on the fused perception confidence level and the corrected collision risk value includes: The perceived uncertainty is used to correct the preset first risk trigger threshold and second risk trigger threshold, and the sound and light warning trigger threshold and automatic deceleration trigger threshold are obtained. When the corrected collision risk value is greater than or equal to the sound and light warning trigger threshold, the first-level sound and light warning is triggered, a warning beam is projected onto the lane in front of the vehicle and a warning sound is heard. When the corrected collision risk value is greater than or equal to the automatic deceleration trigger threshold, the second-level automatic deceleration is triggered to reduce the vehicle's current speed to the target speed. The target driving speed is compared with the minimum safe driving speed. When the target driving speed is lower than the minimum safe driving speed, the minimum safe driving speed is taken as the final driving speed.

9. A safety control system for unmanned vehicles in coal mines according to claim 8, characterized in that, The acquisition of the minimum safe driving speed and the target driving speed includes: Based on the current environment, a baseline safe driving speed is obtained through mapping. An environmental clarity factor is calculated based on the normalized dust concentration value, wherein the environmental clarity factor is inversely proportional to the normalized dust concentration value. Multiply the baseline safe driving speed by the environmental clarity factor, and the product is taken as the minimum safe driving speed. The perception degradation degree is calculated based on the fusion perception credibility, and the driving speed value is multiplied by the perception degradation degree to obtain the perception constraint speed reduction amount; Subtract the perceived speed reduction from the driving speed value, and the difference is taken as the target driving speed. When the target driving speed is lower than the preset emergency braking trigger speed, the vehicle emergency braking is triggered.

10. A safety control method for unmanned vehicles in coal mines, characterized in that, The method, applied to a safety control system for an unmanned underground vehicle in a coal mine according to any one of claims 1 to 9, comprises: Obtain the ambient dust concentration value of the vehicle's current environment, and calculate the normalized dust concentration value based on the current environment. Based on the normalized dust concentration value, and combined with the perception performance indicators of lidar and image recognition, the lidar perception degradation index and the visual perception degradation index are calculated and obtained respectively. The credibility of fused perception is calculated based on the LiDAR perception degradation index and the visual perception degradation index. Based on the fusion perception reliability, the original collision risk assessment value of obstacle detection is corrected to obtain the corrected collision risk value; Based on the fused perception confidence level and the corrected collision risk value, the current safety control strategy for the vehicle is determined.