Strip mine transportation scheduling method based on big data processing

Abstract

The invention relates to the technical field of transport vehicle scheduling, in particular to a strip mine transport scheduling method based on big data processing, and the method comprises the steps: collecting multi-source data; environment feature extraction; intelligent region division; performing path local correction; high interference path connection; performing double-track fusion comparison; and executing dynamic scheduling. According to the invention, through tight coupling of transportation position and speed data (indicating a vehicle motion state and a network load), equipment state characteristics (reflecting a vehicle health level) and environmental perception indexes (dust visibility and a point cloud distortion gradient), comprehensive evaluation of three dimensions of traffic capacity, risk exposure and operation efficiency is immediately formed at an edge node; the problems of low transportation efficiency, frequent path conflicts and increased equipment operation risk caused by disconnection of a scheduling scheme and an actual transportation environment in an interference scene due to lack of real-time environment perception and path self-adaption capabilities are effectively solved.

Description

Technical Field

[0001] This invention relates to the field of transportation vehicle scheduling technology, and in particular to an open-pit mine transportation scheduling method based on big data processing. Background Technology

[0002] With the accelerated development of smart mines, open-pit mine transportation faces multiple challenges, including high operational intensity, complex environmental interference, and high vehicle density. Particularly after blasting operations, dust pollution, reduced visibility, and uneven obstacle distribution severely impact the driving safety and route planning of transport vehicles. Furthermore, during continuous high-load operation, the equipment status of transport vehicles is prone to fluctuations. If these fluctuations are not detected and managed in a timely manner, they can easily lead to scheduling chaos, transportation delays, and even safety hazards. Against this backdrop, effectively identifying vehicle status and environmental interference in complex environments, and coordinating multiple vehicle routes to improve overall scheduling efficiency, has become a key challenge for open-pit mine transportation systems.

[0003] Patent document CN113947496A discloses a method for ore blending and scheduling in an unmanned open-pit mine transportation system. The method includes: acquiring information on blast piles, unloading points, and ore blending targets; obtaining an ore blending scheme by solving a pre-constructed mathematical model based on the blast pile information, unloading point information, and ore blending target information; determining the proportion of blast pile loading vehicles according to the ore blending scheme; and matching unmanned mining trucks in the unmanned open-pit mine transportation system with the proportion of blast pile loading vehicles and pre-set information on shovel loading equipment at the blast piles to generate an ore blending and scheduling plan.

[0004] Therefore, the ore allocation and scheduling method of the open-pit mine unmanned transportation system has the following problems: the method only schedules based on the static information of the blast pile and unloading point obtained in advance, and lacks the perception of real-time environmental changes during the transportation process; the generation of scheduling schemes depends on solving pre-built models and lacks dynamic path correction and obstacle avoidance capabilities; under the influence of a large amount of blasting dust in the open-pit mine after blasting operations, the scheduling does not consider interference factors and is prone to scheduling anomalies. Summary of the Invention

[0005] To address this, the present invention provides a big data-based open-pit mine transportation scheduling method. This method overcomes the problems in existing technologies where the lack of real-time environmental perception and path adaptation capabilities leads to a disconnect between the scheduling scheme and the actual transportation environment in interference scenarios, resulting in low transportation efficiency, frequent path conflicts, and increased equipment operation risks. This is achieved by integrating multi-source environmental perception data with a dynamic trajectory optimization mechanism.

[0006] To achieve the above objectives, the present invention provides one aspect: a method for scheduling open-pit mine transportation based on big data processing, comprising:

[0007] Real-time data collection includes the transportation location, speed, driving angle, equipment status data, images within the collection area, and lidar point cloud data of each vehicle awaiting dispatch in the open-pit mine after blasting operations.

[0008] Real-time extraction of dust visibility in the image and point cloud distortion gradient value in the lidar point cloud data;

[0009] The occlusion tolerance threshold is determined based on the dust visibility and the point cloud distortion gradient value, and several schedulable areas and several high-interference areas are determined based on the occlusion tolerance threshold and the point cloud distortion gradient value.

[0010] For each of the schedulable areas, the entry point and exit point are determined according to the transport speed and the transport location, and a local correction path is generated according to the entry point, exit point, the driving angle and the preset driving correction coefficient.

[0011] Within each of the high-interference areas, a connection path is determined based on the exit point of the previous schedulable area and the entry point of the next schedulable area.

[0012] By fusing the local correction path and the connection path, a first scheduling trajectory is obtained;

[0013] The transportation location, transportation speed, and equipment status data are input into a preset big data scheduling model to obtain a second scheduling trajectory.

[0014] The occlusion tolerance threshold is adjusted based on the path deviation between the first scheduling trajectory and the second scheduling trajectory, and the preset driving correction coefficient is adjusted based on the average value of the point cloud distortion gradient in the locally corrected path segment, according to the adjusted occlusion tolerance threshold.

[0015] The vehicles to be dispatched are dispatched based on the first dispatch trajectory obtained after adjusting the preset driving correction coefficient.

[0016] Furthermore, the extraction process of the point cloud distortion gradient values ​​includes:

[0017] The horizontal gradient is obtained by performing gradient convolution on the lidar point cloud data.

[0018] The vertical gradient is obtained by performing gradient convolution on the lidar point cloud data in the vertical direction.

[0019] Based on the fusion result of the gradient magnitudes of the horizontal and vertical gradients, the point cloud distortion gradient value is extracted.

[0020] Furthermore, the process of determining the occlusion tolerance threshold based on the dust visibility and the point cloud distortion gradient value includes:

[0021] The occlusion index is calculated based on the maximum value among all the point cloud distortion gradient values, the preset gradient weight, the dust visibility, and the preset visibility weight.

[0022] The occlusion tolerance threshold is determined based on the occlusion index and the preset occlusion tolerance threshold mapping table.

[0023] Furthermore, the process of determining several schedulable regions and several high-interference regions based on the occlusion tolerance threshold and the point cloud distortion gradient value includes:

[0024] Based on the comparison between the point cloud distortion gradient value and the occlusion tolerance threshold, several schedulable regions and several high-interference regions are determined.

[0025] Furthermore, the process of determining the entry and exit points based on the transport speed and the transport location includes:

[0026] Calculate the Euclidean distance between all boundary points on the boundary between the transport location and the schedulable area, and select the point with the smallest distance as the entry point;

[0027] The exit point is predicted based on the entry point, the predicted duration, and the vector of the transport speed along the direction of the transport speed.

[0028] The predicted duration is calculated from the diagonal length of the schedulable area, the vector of the transport speed, and a preset scaling factor.

[0029] Furthermore, the process of generating a local correction path based on the entry point, exit point, driving angle, and preset driving correction coefficient includes:

[0030] The path length is determined based on the straight-line distance between the entry point and the exit point, and the tangent vector direction is constructed by combining the driving angle and the preset driving correction coefficient.

[0031] Calculate two control points for generating the Bézier curve based on the direction of the tangent vector and the normal vector perpendicular to it.

[0032] Based on the entry point, the two control points, and the exit point, a Bezier local path is constructed to generate the local modified path.

[0033] Furthermore, the process of determining the connection path based on the exit point of the previous schedulable area and the entry point of the next schedulable area includes:

[0034] Construct a temporary path connecting the exit point and the entry point;

[0035] A maximum curvature constraint is applied to the temporary path to determine the connection path.

[0036] Furthermore, the process of adjusting the occlusion tolerance threshold based on the path deviation between the first scheduling trajectory and the second scheduling trajectory includes:

[0037] Calculate the spatial deviation between the first scheduling trajectory and the second scheduling trajectory at a preset sampling time, and adjust the occlusion tolerance threshold based on the comparison result between the spatial deviation and the preset deviation threshold.

[0038] Furthermore, the process of adjusting the preset driving correction coefficient based on the average value of the point cloud distortion gradient in the local correction path segment, based on the adjusted occlusion tolerance threshold, includes:

[0039] The preset driving correction coefficient is adjusted based on the comparison between the adjusted occlusion tolerance threshold and the average value of the point cloud distortion gradient in all the regenerated local correction path segments.

[0040] On the other hand, the present invention also provides an open-pit mine transportation scheduling system based on big data processing, comprising:

[0041] The data acquisition module is used to collect data in real time on the transportation location, transportation speed, driving angle, equipment status, images within the acquisition area, and multi-source environmental perception data of each vehicle to be dispatched in the open-pit mine after blasting operations.

[0042] An extraction module, which is connected to the acquisition module, is used to extract dust visibility in the image and point cloud distortion gradient values ​​in the multi-source environmental perception data in real time.

[0043] A region determination module, which is connected to the extraction module, is used to determine an occlusion tolerance threshold based on the dust visibility and the point cloud distortion gradient value, and to determine several schedulable regions and several high-interference regions based on the occlusion tolerance threshold and the point cloud distortion gradient value.

[0044] The first generation module is connected to the acquisition module and the area determination module respectively, and is used to determine the entry point and exit point for each schedulable area according to the transportation speed and the transportation location, and generate a local correction path according to the entry point, exit point, driving angle and preset driving correction coefficient.

[0045] The second generation module is connected to the acquisition module and the area determination module respectively, and is used to determine the connection path in each of the high interference areas based on the exit point of the previous schedulable area and the entry point of the next schedulable area.

[0046] A fusion module, which is connected to the first generation module and the second generation module respectively, is used to fuse the local correction path and the connection path to obtain a first scheduling trajectory;

[0047] A big data processing module, which is connected to the acquisition module, is used to input the transportation location, transportation speed and equipment status data into a preset big data scheduling model to obtain a second scheduling trajectory;

[0048] An adjustment module, which is connected to the fusion module, the big data processing module, the first generation module, and the extraction module respectively, is used to adjust the occlusion tolerance threshold according to the path deviation between the first scheduling trajectory and the second scheduling trajectory, and adjust the preset driving correction coefficient based on the average value of the point cloud distortion gradient value in the local correction path segment based on the adjusted occlusion tolerance threshold.

[0049] The scheduling module, which is connected to the fusion module, is used to schedule each of the vehicles to be scheduled based on the first scheduling trajectory obtained after adjusting the preset driving correction coefficient.

[0050] Compared with existing technologies, the advantages of this invention lie in its ability to instantly form a comprehensive assessment of three dimensions—"traffic capacity," "risk exposure," and "operational efficiency"—at edge nodes by tightly coupling transportation location and speed data (indicating vehicle motion status and network load), equipment status characteristics (reflecting vehicle health levels), and environmental perception indicators (dust visibility, point cloud distortion gradient). The Actor-Critic strategy network uses this assessment to generate a preliminary scheduling plan, establishing a basic trajectory using transportation location and speed, and then prioritizing the avoidance of high-failure-risk vehicles based on equipment health tags. The opportunity constraint module dynamically adjusts the occlusion tolerance threshold and path correction coefficient based on estimated transportation time and minimum stacking requirements, ensuring precise local path curvature and obstacle avoidance requirements. The system matches dust concentration and point cloud disturbance intensity. Ultimately, the first scheduling trajectory is continuously updated in the system closed loop based on the deviation from the second predicted trajectory, achieving adaptive and refined adjustment of the correction coefficient. Through multi-level parameter input and calibration, a "perception-prediction-correction-execution-feedback" closed loop is formed. This not only shortens the operation cycle and balances the load at loading and unloading points, but also significantly reduces fuel consumption and downtime. It ensures that scheduling decisions maintain high accuracy and robustness under dynamic conditions such as blasting dust, road congestion, and equipment degradation. This effectively solves the problem of scheduling schemes being out of sync with the actual transportation environment in interference scenarios due to a lack of real-time environmental perception and path adaptation capabilities, resulting in low transportation efficiency, frequent path conflicts, and increased equipment operation risks.

[0051] Furthermore, by applying the Sobel operator to the LiDAR point cloud data in the horizontal and vertical directions respectively, the resulting horizontal and vertical gradients can effectively characterize the degree of distance change in the local spatial structure. The distortion gradient value obtained by fusing the amplitudes of the two can accurately reflect the degree of spatial abrupt change at the edge of complex terrain or obstacles. This fully utilizes the edge sensitivity characteristics of the gradient operator, so that the extraction results contain both directional information and have good boundary localization capabilities. This helps to enhance the perception and resolution of abnormal terrain changes in complex environments such as those with dust interference, thereby providing more reliable and robust input feature support for subsequent occlusion area identification and path correction.

[0052] Furthermore, by fusing dust visibility and point cloud distortion gradient values ​​to construct an occlusion index, and combining this with preset weight parameters to achieve synergistic fusion of the two types of perception information, the degree of occlusion impact on open-pit mine transportation paths can be accurately quantified. Specifically, the point cloud distortion gradient value reflects the degree of disturbance to the environmental geometry, while dust visibility characterizes image transparency and light scattering intensity; both contribute to path reachability determination. Normalization eliminates dimensional differences, and weighted fusion forms a unified-scale occlusion index. Combining this with mapping relationships yields an occlusion tolerance threshold, thereby enabling dynamic perception and reasonable response to occlusion risks in complex environments. This improves the accuracy of path correction and the robustness of the scheduling system, aligning with the physical linkage and functional synergy logic among parameters in the scheduling perception decision-making process.

[0053] Furthermore, by comparing the point cloud distortion gradient with an occlusion tolerance threshold and combining it with a weighted fusion result of dust visibility, accurate classification of each subdivided spatial region of the mining area road is achieved: when the gradient value does not exceed the tolerance threshold, the system identifies the region as a scheduling candidate area with good visibility; when the gradient value exceeds the threshold, it is quickly marked as a high-interference area and triggers path detour or speed limit strategies. This gradient-threshold-based method naturally aligns with the sensitivity of lidar to spatial geometric changes and the response of image transmittance to atmospheric particle concentration. It can dynamically adjust scheduling paths and vehicle allocation under complex dust and variable terrain conditions, significantly improving driving safety, operational efficiency, and system robustness.

[0054] Furthermore, by comparing the vehicle's current position with the Euclidean distances of each point on the boundary of the schedulable area, the entry point with the shortest distance is automatically selected. Then, based on the current position velocity vector and the predicted duration calculated based on the length of the area's diagonal and a preset scaling factor, the exit point is accurately projected along the velocity direction. This path node determination method, which combines the shortest distance with uniform motion prediction, takes into account both spatial geometric features and vehicle dynamics characteristics, while ensuring the reasonable arrangement of entry and exit points. This enables high-precision planning and rapid response of the driving path in a dynamic environment, significantly improving the reliability and efficiency of scheduling decisions.

[0055] Furthermore, the path length is determined by the straight-line distance between the entrance and exit points, and a correction tangent vector is constructed by combining the current driving angle and the preset driving correction coefficient. This allows the local path to flexibly deviate to avoid obstacles while maintaining the overall directional consistency. By using the control points generated by the direction of the tangent vector and its normal vector, a smooth and curvature-continuous local correction path can be formed between the entrance and exit. This avoids obstruction caused by simply traveling straight along the heading and also prevents sharp turns caused by over-correction. It achieves an effective balance between path smoothness and target visibility, which helps to improve tracking stability and transportation safety.

[0056] Furthermore, through discretization sampling and curvature detection, a refined evaluation of temporary straight paths is achieved. When the curvature of the sampling point exceeds the limit, Dubins curve reconstruction is introduced, which can quickly generate feasible paths while ensuring the minimum turning radius limit. It combines the simplicity of straight segments with the controllable curvature characteristics of Dubins curves, enabling a smooth transition between the "shortest straight line" and "restricted turns" in the connecting path, thereby improving the real-time performance and safety redundancy of path planning.

[0057] Furthermore, by splicing the local correction path and the connecting path according to the actual entry order, the temporal consistency between the scheduling trajectory and the environmental occlusion area can be maintained, avoiding logical misalignment in the path generation process. The continuity detection and curvature smoothing of the sub-trajectories ensure that there are no abrupt corners or curvature changes at the splicing points, so that the final trajectory is both smooth and continuous as a whole, and can quickly respond to the avoidance needs of each interference area, thereby significantly improving scheduling safety.

[0058] Furthermore, by comparing the average intensity of the point cloud distortion gradients in all corrected path segments with the latest calculated occlusion tolerance threshold, and automatically reducing the path curvature correction coefficient based on the relative deviation between the two and the preset adjustment force, the driving curve can be tightened in real time when environmental interference increases, increasing the avoidance buffer; when interference decreases, the curvature is maintained or moderately relaxed to ensure smooth driving. This dynamic adjustment not only balances the need for rapid avoidance in obstacle-dense areas with the continuity requirements of unobstructed road sections, but also enables various parameters (gradient mean, tolerance threshold, and adjustment force) to mutually correct and compensate each other during path generation, significantly improving the flexibility, stability, and execution efficiency of the scheduling system under complex terrain and variable occlusion conditions.

[0059] Furthermore, through a modular process of "perception—extraction—region determination—path generation—fusion—prediction—adaptive adjustment—scheduling execution," multi-source parameters such as transportation location, speed, equipment status, dust visibility, and point cloud distortion gradient are coupled layer by layer: First, schedulable and high-interference areas are automatically divided based on environmental and geometric disturbances. Then, a smooth obstacle avoidance path is generated by combining the entrance / exit points, velocity vectors, and correction coefficients. Subsequently, a big data model is used to provide a second scheduling trajectory prediction from different dimensions. After comparing the deviations of the two trajectories, the occlusion threshold and curvature correction coefficient are dynamically updated, ultimately forming a scheduling command that balances safety, efficiency, and reliability. Each parameter serves as an input and correction to the other within the model: location and speed determine path nodes, dust and gradients affect tolerance settings, tolerance in turn drives curvature adjustment, and curvature changes are fed back into region division, forming an adaptive closed loop. Therefore, the system can achieve high-precision and high-robustness transportation scheduling in complex mining environments such as blasting dust, high-temperature vibration, and sudden changes in road conditions. Attached Figure Description

[0060] Figure 1 This is a flowchart of the open-pit mine transportation scheduling method based on big data processing in this embodiment;

[0061] Figure 2 This embodiment defines the logic diagram for determining schedulable regions and high-interference regions.

[0062] Figure 3 This is a logic diagram for adjusting the preset driving correction coefficient in this embodiment;

[0063] Figure 4 This is a schematic diagram of the open-pit mine transportation scheduling system based on big data processing in this embodiment. Detailed Implementation

[0064] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0065] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0066] On the one hand, please refer to Figure 1 As shown, it is a flowchart of the open-pit mine transportation scheduling method based on big data processing in this embodiment;

[0067] This embodiment provides a method for open-pit mine transportation scheduling based on big data processing, including:

[0068] Real-time data collection of transportation location, speed, driving angle, equipment status, images at each collection node, and lidar point cloud data of vehicles waiting to be dispatched in the open-pit mine after blasting operations.

[0069] Real-time extraction of dust visibility in the image and point cloud distortion gradient value in the lidar point cloud data;

[0070] The occlusion tolerance threshold is determined based on the dust visibility and the point cloud distortion gradient value, and several schedulable areas and several high-interference areas are determined based on the occlusion tolerance threshold and the point cloud distortion gradient value.

[0071] For each of the schedulable areas, the entry point and exit point are determined according to the transport speed and the transport location, and a local correction path is generated according to the entry point, exit point, the driving angle and the preset driving correction coefficient.

[0072] Within each of the high-interference areas, a connection path is determined based on the exit point of the previous schedulable area and the entry point of the next schedulable area.

[0073] By fusing the local correction path and the connection path, a first scheduling trajectory is obtained;

[0074] The transportation location, transportation speed, and equipment status data are input into a preset big data scheduling model to obtain a second scheduling trajectory.

[0075] The occlusion tolerance threshold is adjusted based on the path deviation between the first scheduling trajectory and the second scheduling trajectory, and the preset driving correction coefficient is adjusted based on the average value of the point cloud distortion gradient in the locally corrected path segment, according to the adjusted occlusion tolerance threshold.

[0076] The vehicles to be dispatched are dispatched based on the first dispatch trajectory obtained after adjusting the preset driving correction coefficient.

[0077] In this embodiment, the system completes data acquisition through the collaborative work of vehicle-mounted and ground-based multi-source sensing units: each mining truck to be dispatched is equipped with a GPS / IMU module on its roof and a CAN-bus data collector on its chassis. The former outputs three-dimensional position coordinates, instantaneous vehicle speed, and heading angle at a frequency of 10Hz, while the latter captures equipment status data such as engine speed, hydraulic pressure, body vibration amplitude, fuel flow, and operation cycle duration at a frequency of 5Hz. At the same time, high-resolution visible light / near-infrared industrial cameras are deployed along the road and at each loading and unloading platform to capture road images at a rate of 10Hz. Solid-state LiDAR is installed next to the cameras to send the 10Hz point cloud stream (including the X / Y / Z coordinates, reflection intensity, and echo count of each laser echo) to the edge computing gateway via Ethernet. After the edge gateway timestamps and preprocesses all the above raw data, it reports it to the mine dispatch center via a 5G+NB-IoT dual-mode link with encryption, ensuring end-to-end low-latency and high-reliability sensing and control under harsh working conditions such as blasting dust and high-temperature vibration.

[0078] The process of extracting dust visibility from the image in real time includes: the edge gateway uses a multi-scale dark primary color prior algorithm based on an atmospheric scattering model and multipath light attenuation estimation to perform haze separation and transmittance back calculation on the input image. Combined with the pre-calibrated camera projection parameters and the distance to the ground target, the atmospheric optical attenuation coefficient and visibility distance are calculated in real time. Finally, a "dust visibility" value representing dust concentration and visibility level is generated and used together with other key features for subsequent scheduling decisions.

[0079] The preset driving correction coefficient is a scaling factor used to control the length of the local path Bezier tangent vector. It depends on the distance between the entry and exit points and the expected curvature change requirements. It is usually set between 0.1 and 0.5 to balance path smoothness and obstacle avoidance response speed. In this embodiment, it is set to 0.3, which can quickly bypass light obstacles while maintaining path smoothness and improve overall scheduling efficiency.

[0080] The process of determining the vehicles to be dispatched includes the following three main steps, which are completed sequentially on the edge computing nodes:

[0081] (1) Feature extraction and standardization

[0082] After blasting operations, equipment status data of all transport vehicles in the open-pit mine during transportation was collected. Five key variables were extracted from the equipment status data: engine load percentage L (%), hydraulic system pressure Q (MPa), machine vibration amplitude V (g), single operation cycle duration T (s), and fuel consumption fluctuation rate F (%). To eliminate the influence of dimensions, Z-score normalization was applied to each variable (existing technology, which will not be elaborated further).

[0083] (2) Decision tree model construction process

[0084] Based on the CART algorithm, using the Gini index as the node partitioning criterion:

[0085]

[0086] Where, p k This represents the proportion of samples of category k in the current node; each split is performed using the feature with the maximum information gain until any termination condition is met: node purity ≥ 90%, tree depth > 6 layers, or number of node samples < 10.

[0087] (3) Random Forest Model Construction and Ensemble

[0088] Deploy a random forest model containing N base classification trees:

[0089] Each tree generates a subset by bootstrapping the training sample set;

[0090] At each node split, m feature subsets are randomly selected for Gini partitioning;

[0091] The value of m depends on the balance between the feature diversity required for each base classification tree during splitting and the computational cost; typically, m is set between the square root of the total number of features Z and Z / 2 to balance the classification ability of a single tree with the overall decorrelation effect. In this embodiment, there are five features (L, Q, V, T, F), so setting m to 2 can significantly reduce the correlation between trees while maintaining the judgment power of the base trees, thereby improving the generalization performance and robustness of the random forest.

[0092] Finally, the outputs of all base trees are merged through majority voting.

[0093] After the model inference is completed, the system outputs the equipment status label for each vehicle ("operating normally", "overload warning" or "about to fail"), and selects the set of vehicles marked as "operating normally" or "overload warning" and meeting the scheduling priority as vehicles to be scheduled.

[0094] The pre-defined big data scheduling model in this embodiment refers to a multi-objective open-pit mine transportation scheduling optimization model that integrates the Actor-Critic reinforcement learning mechanism and the chance-constrained programming method. This model takes real-time data on transportation location, speed, and equipment status collected from the edge as input, comprehensively considers multiple objectives such as transportation time, fuel cost, equipment reliability, and scheduling fairness, and dynamically outputs the optimal scheduling scheme that satisfies operational constraints, generating the final second scheduling trajectory.

[0095] The Actor-Critic module is used for real-time policy generation and state value assessment, enabling intelligent scoring and policy updates for different scheduling schemes based on historical learning experience. The Opportunity Constraint module models key constraints such as total task duration, unloading point output, and failure risk under a pre-set confidence level, ensuring the stability and robustness of the generated scheduling trajectory. Internally, the model uses the Chebyshev inequality approximation method to linearize the uncertainty constraints, ensuring solution efficiency and practicality.

[0096] Specifically, within each scheduling cycle, the system obtains the current transportation location, real-time speed, and equipment status parameters of the transport vehicles through edge computing nodes, and constructs a scheduling state vector. This vector is then fed into the Actor network as input, and an initial task allocation suggestion is output. Subsequently, it is modified and optimized in combination with chance constraints, and finally outputs a second scheduling trajectory that satisfies multi-objective optimization and confidence constraints.

[0097] By tightly coupling transport location and speed data (indicating vehicle motion status and network load) with equipment status characteristics (reflecting vehicle health level) and environmental perception indicators (dust visibility, point cloud distortion gradient), a comprehensive assessment of three dimensions—"traffic capacity," "risk exposure," and "operational efficiency"—is instantly generated at the edge nodes. The Actor-Critic strategy network uses this assessment to generate a preliminary scheduling plan, establishing a basic trajectory using transport location and speed, and then prioritizing the avoidance of high-risk vehicles based on equipment health tags. The opportunity constraint module dynamically adjusts the occlusion tolerance threshold and path correction coefficient based on estimated transport time and minimum storage requirements, ensuring that the curvature of local paths and obstacle avoidance requirements precisely match dust concentration and point cloud conditions. Disturbance intensity; ultimately, the first scheduling trajectory is continuously updated in the system closed loop based on the deviation from the second predicted trajectory, realizing adaptive and refined adjustment of the correction coefficient. Through multi-level parameters as inputs and calibrations, a "perception-prediction-correction-execution-feedback" closed loop is formed. This not only shortens the operation cycle and balances the load at loading and unloading points, but also significantly reduces fuel consumption and downtime due to malfunctions. It ensures that scheduling decisions maintain high accuracy and robustness under dynamic conditions such as blasting dust, road congestion, and equipment deterioration. It effectively solves the problem that the scheduling scheme in interference scenarios is out of sync with the actual transportation environment due to the lack of real-time environmental perception and path adaptation capabilities, resulting in low transportation efficiency, frequent path conflicts, and increased equipment operation risks.

[0098] Specifically, the extraction process of the point cloud distortion gradient values ​​includes:

[0099] The horizontal gradient is obtained by performing Sobel convolution on the lidar point cloud data in the horizontal direction.

[0100] The vertical gradient is obtained by performing Sobel convolution on the lidar point cloud data in the vertical direction.

[0101] Based on the gradient magnitude fusion result of the horizontal and vertical gradient maps, the point cloud distortion gradient value is extracted. Where Gx is the horizontal gradient, Gy is the vertical gradient, and G(x,y) is the point cloud distortion gradient value.

[0102] By applying the Sobel operator to convolve LiDAR point cloud data in the horizontal and vertical directions, the resulting horizontal and vertical gradients can effectively characterize the drastic changes in distance within the local spatial structure. Furthermore, the distortion gradient value obtained by fusing the amplitudes of the two gradients can accurately reflect the degree of spatial abrupt changes at the edges of complex terrain or obstacles. This fully utilizes the edge sensitivity characteristics of the gradient operator, enabling the extraction results to contain both directional information and good boundary localization capabilities. This helps to enhance the perception and resolution of abnormal terrain changes in complex environments with dust interference, thereby providing more reliable and robust input feature support for subsequent occlusion area identification and path correction.

[0103] Specifically, the process of determining the occlusion tolerance threshold based on the dust visibility and the point cloud distortion gradient value includes:

[0104] Calculate the ratio of the maximum value in the point cloud distortion gradient to the preset maximum gradient threshold to obtain the normalized distortion gradient in the interval [0, 1].

[0105] Dust visibility itself is a parameter in the interval [0, 1]. The occlusion index is obtained by weighting and summing the normalized distortion gradient, the preset gradient weight, the dust visibility, and the preset visibility weight. Y = α × U + β × G, where Y is the occlusion index, α is the preset gradient weight, U is the normalized distortion gradient, β is the preset visibility weight, and G is the dust visibility.

[0106] The occlusion tolerance threshold is determined based on the occlusion index and the preset occlusion tolerance threshold mapping table, wherein the preset occlusion tolerance threshold mapping table is as follows:

[0107] Occlusion Index Y Occlusion tolerance threshold T (unit: meters) 0.0≤Y＜0.2 1.0 0.2≤Y＜0.4 0.8 0.4≤Y＜0.6 0.6 0.6≤Y＜0.8 0.4 0.8≤Y＜1.0 0.2

[0108] The preset maximum gradient threshold is a reference upper limit used to normalize the point cloud distortion gradient values, and its unit is gradient magnitude (dimensionless). This threshold depends on the maximum gradient response intensity extracted from the lidar point cloud in a typical unobstructed environment, and is usually set between [15, 35] to cover the gradient variation range from flat roads to complex obstacle areas. In this embodiment, the preset maximum gradient threshold is set to 25, which can effectively distinguish between slight disturbances and significant occlusion, avoiding misjudgment of small fluctuations while maintaining sensitive response, thereby improving the stability and judgment accuracy of the occlusion index.

[0109] The preset visibility weight is a parameter that controls the proportion of dust visibility in the occlusion index calculation. Its value depends on the scheduling system's sensitivity to optical interference and the influence of dust disturbance in the overall occlusion determination. It is usually set between [0.3, 0.7] to balance the contribution of image sharpness and geometric disturbance. In this embodiment, the preset visibility weight is set to 0.4, which can maintain a reasonable assessment of occlusion risk under conditions of large dust concentration fluctuations and enhance the adaptability and stability of scheduling path correction.

[0110] The preset gradient weight is an adjustment parameter that controls the influence of point cloud distortion gradient in the calculation of occlusion index. Its value depends on the system's sensitivity assessment to structural occlusion (such as abrupt changes in obstacle outlines, rapid changes in slope, etc.), and is usually set between [0.3, 0.7] to achieve a dynamic balance with the dust visibility parameter. In scenarios with mild dust interference but significant geometric distortion, this weight should be increased to promptly identify sudden occlusion risks. In this embodiment, the preset gradient weight is set to 0.6, which enhances the system's responsiveness to environmental geometric anomalies, enabling the system to more accurately distinguish schedulable areas from high-interference areas in dynamic and complex terrain, thereby optimizing path planning and vehicle scheduling strategies.

[0111] By fusing dust visibility and point cloud distortion gradient values ​​to construct an occlusion index, and combining this with preset weight parameters to achieve synergistic fusion of the two types of perception information, the degree of occlusion impact on open-pit mine transportation paths can be accurately quantified. Specifically, the point cloud distortion gradient value reflects the degree of disturbance to the environmental geometry, while dust visibility characterizes image transparency and light scattering intensity; both contribute to path accessibility determination. Normalization eliminates dimensional differences, and weighted fusion forms a unified-scale occlusion index. Combining this with mapping relationships yields an occlusion tolerance threshold, thereby enabling dynamic perception and reasonable response to occlusion risks in complex environments. This improves the accuracy of path correction and the robustness of the scheduling system, aligning with the physical linkage and functional synergy logic among parameters in the scheduling perception decision-making process.

[0112] Please continue reading. Figure 2 As shown, this is the logic diagram for determining the schedulable region and the high-interference region in this embodiment;

[0113] The process of determining several schedulable regions and several high-interference regions based on the occlusion tolerance threshold and the point cloud distortion gradient value includes:

[0114] When the point cloud distortion gradient value is less than or equal to the occlusion tolerance threshold, the part corresponding to the lidar point cloud data is determined to be a schedulable area.

[0115] When the point cloud distortion gradient value is greater than the occlusion tolerance threshold, the part corresponding to the lidar point cloud data is determined to be a high interference area.

[0116] By comparing the point cloud distortion gradient with an occlusion tolerance threshold and combining it with a weighted fusion result of dust visibility, the system achieves accurate classification of each subdivided spatial region of the mining area road: when the gradient value does not exceed the tolerance threshold, the system identifies the region as a scheduling candidate area with good visibility; when the gradient value exceeds the threshold, it is quickly marked as a high-interference area and triggers path detour or speed limit strategies. This gradient-threshold-based method naturally aligns with the sensitivity of lidar to spatial geometric changes and the response of image transmittance to atmospheric particle concentration. It can dynamically adjust scheduling paths and vehicle allocation under complex dust and variable terrain conditions, significantly improving driving safety, operational efficiency, and system robustness.

[0117] Specifically, the process of determining the entry and exit points based on the transport speed and the transport location includes:

[0118] Calculate the Euclidean distances between all boundary points on the boundary between the transport location and the schedulable area, and select the point with the smallest distance as the entry point. Pr is the entry point, and P is the transportation location. Pa represents the boundary of the schedulable region, and Pa is the boundary point.

[0119] The exit point is predicted based on the entry point, the predicted duration, and the vector of the transport speed along the direction of the transport speed, Pc = Pr + v × Δt, where v is the vector of the transport speed and Pc is the exit point.

[0120] The prediction duration is calculated from the diagonal length of the schedulable area, the vector of transport speeds, and a preset scaling factor. k is the preset scaling factor, Δt is the prediction duration, and diag(A) is the diagonal length of the schedulable region.

[0121] The preset ratio coefficient is an adjustment factor that maps the diagonal length of the schedulable area to the prediction time. It depends on the ratio of vehicle response delay to area size and is usually set between 0.3 and 1.0. In this embodiment, it is set to 0.6, which can avoid path deviation caused by excessive prediction while taking into account area coverage and path accuracy, thereby improving the real-time performance and reliability of exit point prediction.

[0122] By comparing the vehicle's current position with the Euclidean distances of each point on the boundary of the schedulable area, the system automatically selects the entry point with the shortest distance. Then, based on the current position's velocity vector and the predicted duration calculated using the area's diagonal length and a preset scaling factor, the exit point is precisely projected along the velocity direction. This path node determination method, which combines the shortest distance with uniform motion prediction, takes into account both spatial geometric features and vehicle dynamics, while ensuring the reasonable arrangement of entry and exit points. This enables high-precision planning and rapid response of driving paths in dynamic environments, significantly improving the reliability and efficiency of scheduling decisions.

[0123] Specifically, the process of generating a local correction path based on the entry point, exit point, driving angle, and preset driving correction coefficient includes:

[0124] The path length is determined based on the straight-line distance between the entry point and the exit point, and the tangent vector direction is constructed by combining the driving angle and the preset driving correction coefficient.

[0125] The process of determining the path length is as follows:

[0126] L = ||Pc-Pr||, where L is the path length;

[0127] The process of constructing the tangent vector direction by combining the driving angle and the preset driving correction coefficient is as follows:

[0128] Calculate the direction vector from the entry point to the exit point: d = L / ||L||, where d is the direction vector from the entry point to the exit point;

[0129] Convert the driving angle into the corresponding direction unit vector: vθ = (cosθ, sinθ), where θ is the heading angle and vθ is the direction unit vector corresponding to the driving angle;

[0130] The direction unit vector is adjusted according to the preset driving correction coefficient and the direction vector from the entrance point to the exit point to obtain the corrected direction vector, vθ'=(1-R)×d+R×vθ, where vθ' is the corrected direction vector and R is the preset driving correction coefficient;

[0131] Normalize the modified vector to obtain the tangent vector direction, h = vθ' / ||θ'||, where h is the tangent vector direction;

[0132] Based on the direction of the tangent vector and the normal vector perpendicular to it, calculate two control points for generating the Bézier curve: C1 = Pr + s1 × L × vθ' + s2 × L × n, C2 = Pc - s1 × L × vθ' + s2 × L × n, where s1 is a preset first offset coefficient, s2 is a preset second offset coefficient, n is the normal vector perpendicular to the direction of the tangent vector, and C1 and C2 are both control points.

[0133] Based on the entry point, the two control points, and the exit point, a Bezier local path is constructed to generate the local modified path, M = Bezier(Pr, C1, C2, Pc), where M is the local modified path. The Bezier local path is a curve generated by smoothly connecting the entry point, the two control points, and the exit point, which is existing technology and will not be described in detail here.

[0134] The preset first offset coefficient is a proportional parameter that controls the extent of the extension of the Bézier curve in the inlet / outlet direction. It depends on the path curvature requirements and obstacle avoidance distance, and is usually set between 0.2 and 0.5. In this embodiment, it is set to 0.4, which can effectively balance obstacle avoidance offset and heading continuity.

[0135] The preset second offset coefficient is a proportional parameter that controls the curvature of the Bézier curve in the normal direction. It depends on the smoothness of the path and the requirements of the observation angle. It is usually set between 0.1 and 0.3. In this embodiment, it is set to 0.3, which can ensure that the curve curvature is moderate and the observation field of view is continuous.

[0136] The path length is determined by the straight-line distance between the entrance and exit points, and a correction tangent vector is constructed by combining the current driving angle and the preset driving correction coefficient. This allows the local path to flexibly deviate to avoid obstacles while maintaining the overall directional consistency. By using the control points generated by the direction of the tangent vector and its normal vector, a smooth and curvature-continuous local correction path can be formed between the entrance and exit. This avoids obstruction caused by simply traveling straight along the heading and also prevents sharp turns caused by over-correction. It achieves an effective balance between path smoothness and target visibility, which helps to improve tracking stability and transportation safety.

[0137] Specifically, the process of determining the connection path based on the exit point of the previous schedulable region and the entry point of the next schedulable region includes:

[0138] A temporary path is constructed to connect the entry point and the exit point. In this embodiment, a straight line segment is used to directly connect the entry point and the exit point to obtain the temporary path.

[0139] Discretize and sample the temporary path. The sampling is uniformly taken from the interval [0, 1] at a preset fixed step size to obtain a number of sampling points. Each sampling point represents the corresponding spatial position on the path, and a sampling point sequence {P0, P1, ..., Pn} is obtained. Calculate the curvature of each sampling point Pi. ei is the curvature of the i-th sampling point;

[0140] Compare the curvature with the preset maximum curvature threshold. If the curvature at any sampling point exceeds the preset maximum curvature threshold, then based on the Dubins curve (the prior art will not be elaborated further), perform shape reconstruction and adjustment on the path. Under the condition that the curvatures of all sampling points do not exceed the preset maximum curvature threshold, obtain the connected path.

[0141] The preset fixed step size is a parameter during path discretization, which depends on the curvature detection accuracy and computational load balance. It is usually set between 0.01 and 0.05. In this embodiment, it is set to 0.02, which can reduce the computational overhead while ensuring the curvature evaluation accuracy.

[0142] Through discretized sampling and curvature detection, a refined evaluation of the temporary straight path is achieved. When the sampling point curvature exceeds the limit, the Dubins curve reconstruction is introduced, which can quickly generate a feasible path on the premise of ensuring the minimum turning radius limit. It combines the simplicity of the straight segment and the controllable curvature characteristics of the Dubins curve, enabling the connected path to achieve a smooth transition between the "shortest straight line" and the "restricted turn", and improving the real-time performance and safety redundancy ability of path planning.

[0143] Specifically, the process of adjusting the occlusion tolerance threshold according to the path deviation between the first scheduling trajectory and the second scheduling trajectory includes:

[0144] Number all schedulable regions in the order of the actual entry time of the vehicle to be scheduled as the 1st region to the Nth region;

[0145] For the i-th (1 ≤ i ≤ N) region, take the local correction path of the i-th region. If i < N, take the connection path between the i-th and the i + 1-th regions, and dock the end of the local correction path of the i-th region with the start of this connection path to form a continuous sub-trajectory;

[0146] Connect all continuous sub-trajectories in sequence according to the numbering order to obtain the temporary scheduling trajectory.

[0147] Perform path continuity detection and curvature smoothing processing on the temporary scheduling trajectory to obtain the first scheduling trajectory.

[0148] By splicing the local correction path and the connecting path according to the actual entry order, the temporal consistency between the scheduling trajectory and the environmental occlusion area can be maintained, avoiding logical misalignment in the path generation process. The continuity detection and curvature smoothing of the sub-trajectories ensure that there are no abrupt corners or curvature changes at the splicing points, so that the final trajectory is both smooth and continuous as a whole, and can quickly respond to the avoidance needs of each interference area, thereby significantly improving scheduling safety.

[0149] The process of adjusting the occlusion tolerance threshold based on the path deviation between the first scheduling trajectory and the second scheduling trajectory includes:

[0150] Based on a preset sampling time interval, several sampling moments are selected within a preset scheduling time range to form a sampling moment sequence;

[0151] At each sampling time in the sampling time sequence, the spatial location point corresponding to the first scheduling trajectory is obtained to obtain the first trajectory point, and the spatial location point corresponding to the second scheduling trajectory is obtained to obtain the second trajectory point. The Euclidean distance between the first trajectory point and the second trajectory point is calculated to obtain the spatial deviation.

[0152] When the spatial deviation is greater than the preset deviation threshold, the occlusion tolerance threshold is increased according to the relative deviation between the spatial deviation and the preset deviation threshold and the preset deviation coefficient, T'=T×[1+j×(W-W0) / W0], where T' is the increased occlusion tolerance threshold, j is the preset deviation coefficient, W is the spatial deviation, and W0 is the preset deviation threshold.

[0153] The preset sampling time interval refers to the time interval used to extract trajectory points. It depends on the smoothness of the trajectory change and is usually set between 0.1 seconds and 1 second. In this embodiment, it is set to 0.5 seconds, which can balance the accuracy of trajectory sampling and the calculation efficiency.

[0154] The preset scheduling time range refers to the time period used to analyze trajectory deviation. It depends on the total task duration and the time window for trajectory planning adjustments. It is usually set between 10 seconds and 300 seconds. In this embodiment, it is set to 120 seconds, which can cover critical path segments and effectively support deviation assessment.

[0155] The preset deviation threshold is a reference distance value used to determine whether two trajectories have deviated significantly. It depends on the scheduling task's tolerance for trajectory accuracy and is usually set between 0.5 meters and 3 meters. In this embodiment, it is set to 1 meter, which can accurately identify trajectory deviations that have a real impact on scheduling accuracy.

[0156] The preset deviation coefficient is a proportional parameter that controls the adjustment range of the occlusion tolerance threshold. It depends on the system's requirements for deviation response speed and the sensitivity of occlusion impact. It is usually set between 0.1 and 0.5. In this embodiment, it is set to 0.25, which can moderately adjust the occlusion model and improve the flexibility of path correction.

[0157] By sampling the first and second scheduling trajectories hourly within a preset scheduling time range and quantifying the spatial deviation between each pair of sampling points, the system can dynamically perceive the degree of offset between the two trajectories in actual scheduling. When the spatial deviation exceeds a preset threshold, the occlusion tolerance threshold is automatically adjusted according to the deviation magnitude, thereby improving path adaptability and environmental robustness, effectively avoiding local path failures caused by the accumulation of path deviations, and ensuring a high degree of coordination between scheduling trajectory adjustment and occlusion modeling. This realizes an occlusion perception self-correction mechanism based on feedback from actual scheduling effects, improving the accuracy and stability of path planning.

[0158] Please continue reading. Figure 3 As shown, it is the logic diagram for adjusting the preset driving correction coefficient in this embodiment;

[0159] The process of adjusting the preset driving correction coefficient based on the average value of the point cloud distortion gradient in the locally corrected path segment, based on the adjusted occlusion tolerance threshold, includes:

[0160] Calculate the average value of the point cloud distortion gradient values ​​in all the corrected path segments regenerated based on the increased occlusion tolerance threshold, and obtain the gradient mean.

[0161] When the gradient mean is greater than the increased occlusion tolerance threshold, the preset driving correction coefficient is reduced based on the relative deviation between the gradient mean and the increased occlusion tolerance threshold and the preset driving adjustment coefficient, R'=R×[1-g×(Ts-T') / T'], where g is the preset driving adjustment coefficient, Ts is the gradient mean, and R' is the reduced preset driving correction coefficient.

[0162] The preset driving adjustment coefficient is a proportional parameter that controls the degree of driving direction correction. It depends on the path adjustment sensitivity and is usually set between 0.1 and 0.5. In this embodiment, it is set to 0.3, which can improve the correction efficiency while maintaining path smoothness.

[0163] By comparing the average intensity of the point cloud distortion gradients in the corrected path segment with the latest calculated occlusion tolerance threshold, and automatically reducing the path curvature correction coefficient based on the relative deviation between the two and the preset adjustment force, the system can tighten the driving curve in real time when environmental interference increases, increasing the avoidance buffer; and maintain or moderately relax the curvature when interference decreases, ensuring smooth driving. This dynamic adjustment not only balances the need for rapid avoidance in obstacle-dense areas with the continuity requirements of unobstructed road sections, but also enables various parameters (average gradient, tolerance threshold, and adjustment force) to mutually correct and compensate each other during path generation, significantly improving the flexibility, stability, and execution efficiency of the scheduling system under complex terrain and variable occlusion conditions.

[0164] On the other hand, please continue reading Figure 4 As shown, it is a schematic diagram of the open-pit mine transportation scheduling system based on big data processing in this embodiment;

[0165] This embodiment also provides an open-pit mine transportation scheduling system based on big data processing, including:

[0166] The data acquisition module is used to collect data in real time on the transportation location, transportation speed, driving angle, equipment status data, images at each acquisition node, and multi-source environmental perception data of each vehicle to be dispatched in the open-pit mine after blasting operations.

[0167] An extraction module, which is connected to the acquisition module, is used to extract dust visibility in the image and point cloud distortion gradient values ​​in the multi-source environmental perception data in real time.

[0168] A region determination module, which is connected to the extraction module, is used to determine an occlusion tolerance threshold based on the dust visibility and the point cloud distortion gradient value, and to determine several schedulable regions and several high-interference regions based on the occlusion tolerance threshold and the point cloud distortion gradient value.

[0169] The first generation module is connected to the acquisition module and the area determination module respectively, and is used to determine the entry point and exit point for each schedulable area according to the transportation speed and the transportation location, and generate a local correction path according to the entry point, exit point, driving angle and preset driving correction coefficient.

[0170] The second generation module is connected to the acquisition module and the area determination module respectively, and is used to determine the connection path in each of the high interference areas based on the exit point of the previous schedulable area and the entry point of the next schedulable area.

[0171] A fusion module, which is connected to the first generation module and the second generation module respectively, is used to fuse the local correction path and the connection path to obtain a first scheduling trajectory;

[0172] A big data processing module, which is connected to the acquisition module, is used to input the transportation location, transportation speed and equipment status data into a preset big data scheduling model to obtain a second scheduling trajectory;

[0173] An adjustment module, which is connected to the fusion module, the big data processing module, the first generation module, and the extraction module respectively, is used to adjust the occlusion tolerance threshold according to the path deviation between the first scheduling trajectory and the second scheduling trajectory, and adjust the preset driving correction coefficient based on the average value of the point cloud distortion gradient value in the local correction path segment based on the adjusted occlusion tolerance threshold.

[0174] The scheduling module, which is connected to the fusion module, is used to schedule each of the vehicles to be scheduled based on the first scheduling trajectory obtained after adjusting the preset driving correction coefficient.

[0175] Through a modular process of "perception—extraction—region determination—path generation—fusion—prediction—adaptive adjustment—scheduling execution," the system couples multiple parameters, including transport location, speed, equipment status, dust visibility, and point cloud distortion gradient, layer by layer. First, it automatically divides schedulable and high-interference regions based on environmental and geometric disturbances. Then, it generates smooth obstacle avoidance paths by combining entry / exit points, velocity vectors, and correction coefficients. Subsequently, it uses a big data model to provide second-dimensional scheduling trajectory predictions from different dimensions. After comparing the deviations between the two trajectories, it dynamically updates the occlusion threshold and curvature correction coefficients, ultimately forming a scheduling command that balances safety, efficiency, and reliability. Within the model, each parameter serves as both input and correction: location and speed determine path nodes, dust and gradients affect tolerance settings, tolerance in turn drives curvature adjustment, and curvature changes are fed back into region division, forming an adaptive closed loop. Therefore, the system can achieve high-precision and robust transport scheduling in complex mining environments such as blasting dust, high-temperature vibration, and sudden changes in road conditions.

[0176] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for scheduling transportation in open-pit mines based on big data processing, characterized in that, include: Real-time data collection includes the transportation location, speed, driving angle, equipment status data, images within the collection area, and lidar point cloud data of each vehicle awaiting dispatch in the open-pit mine after blasting operations. Real-time extraction of dust visibility in the image and point cloud distortion gradient value in the lidar point cloud data; The occlusion tolerance threshold is determined based on the dust visibility and the point cloud distortion gradient value, and several schedulable areas and several high-interference areas are determined based on the occlusion tolerance threshold and the point cloud distortion gradient value. For each of the schedulable areas, the entry point and exit point are determined according to the transport speed and the transport location, and a local correction path is generated according to the entry point, exit point, the driving angle and the preset driving correction coefficient. Within each of the high-interference areas, a connection path is determined based on the exit point of the previous schedulable area and the entry point of the next schedulable area. By fusing the local correction path and the connection path, a first scheduling trajectory is obtained; The transportation location, transportation speed, and equipment status data are input into a preset big data scheduling model to obtain a second scheduling trajectory. The occlusion tolerance threshold is adjusted based on the path deviation between the first scheduling trajectory and the second scheduling trajectory, and the preset driving correction coefficient is adjusted based on the average value of the point cloud distortion gradient in the locally corrected path segment, according to the adjusted occlusion tolerance threshold. The vehicles to be dispatched are dispatched based on the first dispatch trajectory obtained after adjusting the preset driving correction coefficient.

2. The open-pit mine transportation scheduling method based on big data processing according to claim 1, characterized in that, The process of extracting the point cloud distortion gradient values ​​includes: The horizontal gradient is obtained by performing gradient convolution on the lidar point cloud data. The vertical gradient is obtained by performing gradient convolution on the lidar point cloud data in the vertical direction. Based on the fusion result of the gradient magnitudes of the horizontal and vertical gradients, the point cloud distortion gradient value is extracted.

3. The open-pit mine transportation scheduling method based on big data processing according to claim 2, characterized in that, The process of determining the occlusion tolerance threshold based on the dust visibility and the point cloud distortion gradient value includes: The occlusion index is calculated based on the maximum value among all the point cloud distortion gradient values, the preset gradient weight, the dust visibility, and the preset visibility weight. The occlusion tolerance threshold is determined based on the occlusion index and the preset occlusion tolerance threshold mapping table.

4. The open-pit mine transportation scheduling method based on big data processing according to claim 3, characterized in that, The process of determining several schedulable regions and several high-interference regions based on the occlusion tolerance threshold and the point cloud distortion gradient value includes: Based on the comparison between the point cloud distortion gradient value and the occlusion tolerance threshold, several schedulable regions and several high-interference regions are determined.

5. The open-pit mine transportation scheduling method based on big data processing according to claim 4, characterized in that, The process of determining the entry and exit points based on the transport speed and the transport location includes: Calculate the Euclidean distance between all boundary points on the boundary between the transport location and the schedulable area, and select the point with the smallest distance as the entry point; The exit point is predicted based on the entry point, the predicted duration, and the vector of the transport speed along the direction of the transport speed. The predicted duration is calculated from the diagonal length of the schedulable area, the vector of the transport speed, and a preset scaling factor.

6. The open-pit mine transportation scheduling method based on big data processing according to claim 5, characterized in that, The process of generating a local correction path based on the entry point, exit point, driving angle, and preset driving correction coefficient includes: The path length is determined based on the straight-line distance between the entry point and the exit point, and the tangent vector direction is constructed by combining the driving angle and the preset driving correction coefficient. Calculate two control points for generating the Bézier curve based on the direction of the tangent vector and the normal vector perpendicular to it. Based on the entry point, the two control points, and the exit point, a Bezier local path is constructed to generate the local modified path.

7. The open-pit mine transportation scheduling method based on big data processing according to claim 6, characterized in that, The process of determining the connection path based on the exit point of the previously schedulable region and the entry point of the next schedulable region includes: Construct a temporary path connecting the exit point and the entry point; A maximum curvature constraint is applied to the temporary path to determine the connection path.

8. The open-pit mine transportation scheduling method based on big data processing according to claim 7, characterized in that, The process of adjusting the occlusion tolerance threshold based on the path deviation between the first scheduling trajectory and the second scheduling trajectory includes: Calculate the spatial deviation between the first scheduling trajectory and the second scheduling trajectory at a preset sampling time, and adjust the occlusion tolerance threshold based on the comparison result between the spatial deviation and the preset deviation threshold.

9. The open-pit mine transportation scheduling method based on big data processing according to claim 8, characterized in that, The process of adjusting the preset driving correction coefficient based on the average value of the point cloud distortion gradient in the locally corrected path segment, based on the adjusted occlusion tolerance threshold, includes: The preset driving correction coefficient is adjusted based on the comparison between the adjusted occlusion tolerance threshold and the average value of the point cloud distortion gradient in all the regenerated local correction path segments.

10. An open-pit mine transportation scheduling system based on big data processing, constructed based on the open-pit mine transportation scheduling method based on big data processing according to any one of claims 1-9, characterized in that, include: The data acquisition module is used to collect data in real time on the transportation location, transportation speed, driving angle, equipment status, images within the acquisition area, and multi-source environmental perception data of each vehicle to be dispatched in the open-pit mine after blasting operations. An extraction module, which is connected to the acquisition module, is used to extract dust visibility in the image and point cloud distortion gradient values ​​in the multi-source environmental perception data in real time. A region determination module, which is connected to the extraction module, is used to determine an occlusion tolerance threshold based on the dust visibility and the point cloud distortion gradient value, and to determine several schedulable regions and several high-interference regions based on the occlusion tolerance threshold and the point cloud distortion gradient value. The first generation module is connected to the acquisition module and the area determination module respectively, and is used to determine the entry point and exit point for each schedulable area according to the transportation speed and the transportation location, and generate a local correction path according to the entry point, exit point, driving angle and preset driving correction coefficient. The second generation module is connected to the acquisition module and the area determination module respectively, and is used to determine the connection path in each of the high interference areas based on the exit point of the previous schedulable area and the entry point of the next schedulable area. A fusion module, which is connected to the first generation module and the second generation module respectively, is used to fuse the local correction path and the connection path to obtain a first scheduling trajectory; A big data processing module, which is connected to the acquisition module, is used to input the transportation location, transportation speed and equipment status data into a preset big data scheduling model to obtain a second scheduling trajectory; An adjustment module, which is connected to the fusion module, the big data processing module, the first generation module, and the extraction module respectively, is used to adjust the occlusion tolerance threshold according to the path deviation between the first scheduling trajectory and the second scheduling trajectory, and adjust the preset driving correction coefficient based on the average value of the point cloud distortion gradient value in the local correction path segment based on the adjusted occlusion tolerance threshold. The scheduling module, which is connected to the fusion module, is used to schedule each of the vehicles to be scheduled based on the first scheduling trajectory obtained after adjusting the preset driving correction coefficient.

Images