A dispatching method and system for unmanned transport vehicles in underground mines

Abstract

The present application discloses a method and system for dispatching unmanned transport vehicles in underground mines, which relates to the field of unmanned driving, and includes the following steps: obtaining underground mine map data; processing the obtained map data to obtain a point cloud map; constructing an underground mine map; obtaining driving data of each unmanned transport vehicle; using the constructed underground mine map as a directed topological map according to the current state of each unmanned transport vehicle, whether empty or fully loaded, and performing path planning through an A* algorithm to obtain a path for the unmanned transport vehicle to reach each loading point or unloading point from the current position; selecting an optimal path among all the obtained paths according to the total time taken by the unmanned transport vehicle to enter and leave the passing observation area; determining the final destination of the unmanned transport vehicle according to the selected optimal path; in view of the low efficiency of underground unmanned driving dispatching in the prior art, the present application improves the efficiency of underground unmanned vehicle dispatching.

Description

Technical Field

[0001] The present application relates to the field of unmanned driving, and in particular to a scheduling method and system for unmanned transport vehicles in underground mines. Background Art

[0002] In recent years, mining operations have rapidly evolved towards intelligent and unmanned operations, with an increasing number of mining companies introducing driverless transport vehicles (DVs) to replace traditional manual transport systems. Due to their efficiency, safety, and cost-effectiveness, DVs have been widely adopted in complex mining environments, such as open-pit and underground mines. However, the increasing number of DVs and the increasing complexity of their application scenarios have placed higher demands on the intelligent scheduling of DV fleets.

[0003] Traditional mine transport scheduling relies primarily on manual command and simple dispatching rules, resulting in low efficiency, slow response, and prone to errors. To adapt to the development of autonomous driving technology, there is an urgent need for an intelligent dispatching system that can automatically perceive the mine environment, dynamically optimize vehicle scheduling, and coordinate fleet operations. This intelligent dispatching system must possess key capabilities such as data integration, real-time decision-making, and precise control to maximize the effectiveness of autonomous transport vehicles in complex and changing mining environments. Currently, intelligent dispatching technology for autonomous transport vehicles has been researched and applied to some extent in open-pit mines, but its application in underground mining environments is relatively limited. Compared to open-pit mines, underground mines have narrow tunnels, harsh environments, and dense traffic, presenting greater difficulties and challenges for autonomous vehicle scheduling. In particular, in terms of traffic flow distribution and vehicle avoidance, how to rationally plan vehicle routes, prevent vehicle congestion, and improve system transportation efficiency remains a key issue that needs to be addressed.

[0004] Specifically, underground mine tunnels are generally arranged with a single, two-way lane. If two unmanned vehicles are traveling towards each other in a narrow tunnel, without timely and appropriate avoidance measures, they can easily cause a traffic jam, paralyzing the entire transportation system. To avoid this, unmanned vehicles must perform pass-by maneuvers within designated passing chambers. However, frequent pass-by maneuvers significantly reduce transportation efficiency. If the dispatching system blindly sends vehicles to sections where vehicles are passing each other, the risk of traffic jams will increase, affecting the efficiency and stability of the entire transportation system. Summary of the Invention

[0005] In response to the problem of low efficiency in underground unmanned driving scheduling in the existing technology, this application provides a scheduling method and system for underground unmanned transport vehicles. By constructing an underground mine map and A* algorithm path planning, the optimal path is obtained to improve the scheduling efficiency of underground unmanned vehicles.

[0006] The purpose of this application is achieved through the following technical solutions.

[0007] One aspect of the present specification provides a scheduling method for unmanned transport vehicles in underground mines, comprising: obtaining underground mine map data; processing the obtained map data to obtain a point cloud map; constructing an underground mine map based on the point cloud map; obtaining driving data of each unmanned transport vehicle; the driving data including the total time consumed by vehicles entering and leaving a passing observation area; using the constructed underground mine map as a directed topological graph based on whether the current state of each unmanned transport vehicle is empty or fully loaded, and performing path planning using an A* algorithm to obtain a path for the unmanned transport vehicle to reach each loading point or unloading point from the current position; selecting an optimal path from all obtained paths based on the total time consumed by the unmanned transport vehicle to enter and leave the passing observation area; determining a final destination of the unmanned transport vehicle based on the selected optimal path; and sending the determined optimal path to the selected unmanned transport vehicle, controlling the transport vehicle to travel along the optimal path.

[0008] Among them, point cloud map: a digital map containing a large amount of three-dimensional spatial point data obtained by equipment such as a three-dimensional scanner. Each point contains its three-dimensional coordinates and other attribute information. The point cloud map can truly reflect the three-dimensional structural characteristics of the scene. Passing vehicle observation area: a specific road section used by unmanned vehicles to observe whether other vehicles are entering and determine whether they need to give way. It is usually set up in intersections, narrow sections and other areas where vehicles are likely to meet. A* algorithm: a heuristic search algorithm, commonly used in the field of path planning. This algorithm can efficiently find the optimal path from the starting point to the end point by prioritizing the search for the nodes that are most likely to reach the target. Driving data also includes current position, speed, heading angle, task status, driving status and state machine; calculate the total time taken by the unmanned transport vehicle on each path taking into account the time taken to pass by vehicles.

[0009] Furthermore, an optimal path is selected using the following formula: Where n represents the loading point number to which the transport vehicle is finally dispatched; T n T represents the time when the transport vehicle is expected to arrive at the loading point n; sn It indicates the time when the shovel loader at loading point n is expected to start loading the current transport vehicle; N is the total number of loading points available for transport vehicle scheduling.

[0010] Furthermore, the time Tsn at which the scraper at loading point n is expected to start loading the current transport vehicle is calculated using the following formula: Among them, T nowrepresents the current time; Q represents the number of transport vehicles waiting for loading in the waiting area at loading point n when the transport vehicle arrives at loading point n; t1 represents the idle waiting time of the forklift before transport vehicle q arrives at the loading point; t2 represents the driving time from the waiting area to the loading point; t3 represents the loading time of the transport vehicle; t4 represents the time it takes for the transport vehicle to complete loading and leave the current loading point to travel to the main road.

[0011] Furthermore, the estimated time T of the transport vehicle arriving at the loading point n is calculated by the following formula: n :

[0012]

[0013] Where n represents the number of the loading point, there are N loading points in total; T n T represents the time when the transport vehicle is expected to arrive at the loading point n; now represents the current time; P represents the number of paths for the transport vehicle to reach the loading point n; L represents the number of road sections in the path p for the transport vehicle to reach the loading point n; s l represents the length of the lth segment of the path p for the transport vehicle to reach the loading point n; v l represents the speed of the transport vehicle on the lth road section of the path p to the loading point n; M represents the number of vehicle-passing observation zones on the path p of the transport vehicle to the loading point n; s m v represents the distance of the main path section within the mth vehicle passing observation area on the expected driving trajectory in the path p of the transport vehicle to the loading point n; m represents the driving speed of the transport vehicle in the mth overtaking observation zone on the main path section of the expected driving trajectory in the path p of the loading point n; C represents the number of transport vehicles whose arrival time in the overtaking observation zone overlaps with the time of the current transport vehicle and whose driving priority is higher than the current transport vehicle, and the driving priority is based on the first-come, first-served principle; s m It represents the time taken to pass the other vehicle, which is equal to the total time taken for the other vehicle to enter and leave.

[0014] Furthermore, according to the selected optimal path, the final destination of the unmanned transport vehicle is determined, including: calculating the estimated time T when the w-th transport vehicle in the waiting area of ​​the loading point n is expected to finish loading w,n,e , the calculation formula is as follows:

[0015]

[0016] Among them, T w,n,s T represents the time when the w-th transport vehicle in the waiting area of ​​loading point n arrives at the waiting area; load It represents the total time taken by the transport vehicle to go from the waiting area to the loading point, load the vehicle, and leave the loading point; the calculated load loader is estimated to be the time when the current transport vehicle starts loading T snand the time T at which the transport vehicle is expected to finish loading w,n,e After comparison, the loading point with the longest idle time is selected as the final destination of the transport vehicle, and the destination of the optimal path is updated to the selected final destination.

[0017] Furthermore, constructing an underground mine map includes: performing a three-dimensional scan of the underground mine to obtain map data containing spatial location information and distance information of the underground mine; performing feature extraction on the obtained map data to obtain a point cloud map of the underground mine; performing semantic annotation on the obtained point cloud map; and converting the semantically annotated point cloud map into a three-dimensional raster map data format to obtain a map of the underground mine.

[0018] Among them, spatial position information: represents the three-dimensional spatial coordinate data (such as x, y, z coordinates) of each element in the map (such as road sections, landmarks, etc.), which is used to describe the position of the element in three-dimensional space. Distance information: represents the distance data between each element in the map, such as the distance between the endpoints of two road sections. Combined with spatial position information, distance information is used for applications such as path planning and positioning. Semantic annotation: In map data processing, it is the process of annotating certain point sets as having specific semantic attributes (such as lane lines, traffic signs, buildings, etc.) by identifying specific patterns in raw data such as point clouds. Semantic annotation makes the map readable at the feature level.

[0019] Furthermore, semantic annotation includes: marking a set of lane line feature points in the point cloud map to form lane line information, which is used to represent the driving path of the unmanned transport vehicle; marking a set of observation area feature points in the point cloud map to form observation area information, which is used to represent the area where the unmanned transport vehicle observes passing vehicles; marking a set of passing area feature points in the point cloud map to form passing area information, which is used to represent the area where the unmanned transport vehicle passes vehicles; marking a set of stop line feature points in the point cloud map to form stop line information, which is used to represent the position where the unmanned transport vehicle stops and waits; marking a set of loading point feature points in the point cloud map to form loading point information, which is used to represent the position where the unmanned transport vehicle loads goods; marking a set of unloading point feature points in the point cloud map to form unloading point information, which is used to represent the position where the unmanned transport vehicle unloads goods; marking a set of waiting area feature points in the point cloud map to form waiting area information, which is used to represent the area where the unmanned transport vehicle waits to load or unload.

[0020] Furthermore, the driving data of each unmanned transport vehicle is obtained, including: obtaining the status data of the unmanned transport vehicle through the wireless gateway; the status book contains the current position, speed, heading angle, mission status, driving status and state machine information of the unmanned transport vehicle; according to the obtained status data, the driving data of each unmanned transport vehicle is obtained: according to the preset statistical frequency f, the average speed of each unmanned transport vehicle in each road section within the preset time period is counted to obtain the speed data; according to the preset statistical frequency f, combined with the change time interval of the unmanned transport vehicle state machine, the total time consumed by the unmanned transport vehicle in passing the vehicle in the passing observation area is counted to obtain the passing time consumption data; according to the preset statistical frequency f, the actual arrival time and the expected arrival time of the unmanned transport vehicle at each stop line on the preset driving trajectory are counted, Obtain stop line queue data; determine whether the current position of the unmanned transport vehicle is within the range of the waiting area of ​​the loading point or unloading point on the preset driving trajectory according to the preset statistical frequency f; if so, add the corresponding unmanned transport vehicle to the waiting area queue of the loading point or unloading point, and record the corresponding arrival time of the waiting area to obtain the queue data of the loading point or unloading point waiting area; according to the time interval of the change of the state machine of the unmanned transport vehicle, count the time interval of the unmanned transport vehicle entering the loading point or unloading point, completing the loading or unloading operation, and leaving the loading point or unloading point to obtain the time data of entering or loading or unloading or leaving; use the obtained vehicle speed data, time data of passing the vehicle, stop line queue data, queue data of the waiting area of ​​the loading point or unloading point, and time data of entering or loading or unloading or leaving as driving data.

[0021] State machine information: A finite state machine model is used to describe the autonomous vehicle's behavioral logic and state transition conditions. For example, possible autonomous vehicle states include driving, obstacle avoidance, and parking. Transitions between states depend on the vehicle's perception data and the current task. A directed topological graph: A data structure that uses graph theory to represent a real-world map. Nodes in the graph typically represent road segments, and directed edges indicate the connectivity between road segments and the direction of travel. This topological graph facilitates operations such as route search.

[0022] Furthermore, according to whether the current state of each unmanned transport vehicle is empty or fully loaded, the path of the unmanned transport vehicle from the current position to each loading point or unloading point is obtained, including: taking the set of feature points representing the road sections in the constructed underground mine map data as the vertices of the topological graph, taking the set of feature points of the connectivity relationship between the road sections as the directed edges of the topological graph, and constructing a directed topological graph representing the topological structure of the underground mine map by connecting all vertices and directed edges; obtaining data representing the current mission state of the unmanned transport vehicle in the driving data, and judging whether the current cargo state of each unmanned transport vehicle is empty or fully loaded according to the mission state data; when the current cargo state of the unmanned transport vehicle is empty, in the constructed directed topological graph, The vertex corresponding to the current position of the unmanned transport vehicle is set as the starting point of the path, and the vertex corresponding to each loading point in the underground mine map is set as the end point of the path in turn. The A* algorithm is used to search for the shortest path from the starting point of the path to each end point, and the shortest path for the current unmanned transport vehicle to reach each loading point from the current position is obtained; when the current cargo status of the unmanned transport vehicle is fully loaded, in the constructed directed topological graph, the vertex corresponding to the current position of the unmanned transport vehicle is set as the starting point of the path, and the vertex corresponding to each unloading point in the underground mine map is set as the end point of the path in turn, and the A* algorithm is used to search for the shortest path from the starting point of the path to each end point, and the shortest path for the current unmanned transport vehicle to reach each unloading point from the current position is obtained.

[0023] Preferably, when constructing a directed topological graph, the connectivity relationship between road sections is determined based on the vehicle speed data in the driving data, and the connectivity relationship between two road sections whose average vehicle speed is greater than a preset speed threshold is represented as a directed edge, and there is no directed edge between two road sections whose average vehicle speed is less than the preset speed threshold.

[0024] Preferably, when using the A* algorithm for path search, the vehicle speed data in the driving data is used as the weight of the directed edge, and paths with higher average speeds are searched first, so that the shortest path found meets the optimal driving conditions under actual road conditions.

[0025] Preferably, before using the A* algorithm to search for a path, the directed edges corresponding to the road sections where the passing time exceeds a preset threshold are removed from the directed topology graph based on the passing time data in the driving data, so as to avoid searching for paths with too high passing time.

[0026] Preferably, after using the A* algorithm to obtain the shortest path for the unmanned transport vehicle to reach each loading point or unloading point from the current position, the queue waiting time of different paths passing through the same stop line is compared based on the stop line queue data in the driving data, and the path with the shortest queue waiting time is selected as the recommended path.

[0027] Preferably, after using the A* algorithm to obtain the shortest path from the current position of the unmanned transport vehicle to each loading point or unloading point, the loading and unloading time of different paths to the same destination loading point or unloading point is compared based on the time data of entering, loading, unloading or leaving in the driving data, and the path with the shortest loading and unloading time is selected as the recommended path.

[0028] Another aspect of the present specification also provides a scheduling system for an unmanned transport vehicle for an underground mine, comprising: at least one processing unit; configured to execute instructions to implement a scheduling method for an unmanned transport vehicle for an underground mine.

[0029] Compared with the existing technology, the advantages of this application are:

[0030] By 3D scanning the underground mine to generate a point cloud map, semantically annotating it, and rasterizing it, a mine map model was constructed that accurately represents the complex underground environment. By capturing detailed geometric and semantic information about key areas such as lane markings, observation areas, and loading and unloading points, this provides a reliable map foundation for subsequent path planning, helping to improve the relevance and feasibility of scheduling decisions.

[0031] The mine site point cloud map is abstracted into a directed topological graph with nodes representing road segments and edges representing connectivity relationships. Path planning is performed using the A shortest path search algorithm. Compared to planning directly on the original map, this method simplifies the search space through graph abstraction and rapidly finds the optimal path using the A algorithm's heuristic search. This significantly improves the efficiency and real-time nature of path planning, enabling the dispatch system to quickly respond to dynamically changing vehicle demand.

[0032] A comprehensive vehicle driving data collection and analysis mechanism has been introduced, including statistics on average vehicle speeds on different road sections, time spent passing each other in passing observation areas, and queues at stop lines and loading and unloading points. This detailed measurement of driving data in different areas not only provides data support for path cost calculations, but also provides a more comprehensive optimization basis for scheduling decisions, helping to improve the overall efficiency of vehicle scheduling.

[0033] A path cost calculation model was proposed that comprehensively considers multiple factors, including missed vehicles, queuing, and loading and unloading operations. Unlike simply pursuing the shortest path, this model quantitatively calculates the estimated time to destination and compares different path options. This allows for a more comprehensive balance of path options, selecting the dispatch path with the best overall efficiency under dynamic conditions, thereby improving the dispatch system's adaptability to complex mine environments.

[0034] Based on the optimal path, the estimated loading times at different loading points are further compared with the estimated loading completion times of queued vehicles, and the loading point with the shortest queue time is selected as the final destination. This optimization strategy fully utilizes the service capacity differences of loading points, minimizes vehicle waiting time, and further improves scheduling efficiency from the perspective of regional collaboration. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] This specification will be further described in the form of exemplary embodiments, which will be described in detail with reference to the accompanying drawings. These embodiments are not limiting, and in these embodiments, the same numbers represent the same structures, wherein:

[0036] Figure 1 is an exemplary flow chart of a method for dispatching unmanned transport vehicles for underground mines according to some embodiments of this specification;

[0037] Figure 2 is an exemplary module diagram of a scheduling system for unmanned transport vehicles in underground mines according to some embodiments of this specification;

[0038] Figure 3 is an exemplary flow chart for constructing an underground mine map according to some embodiments of this specification;

[0039] Figure 4 is an exemplary flow chart for obtaining driving data of an unmanned transport vehicle according to some embodiments of this specification;

[0040] Figure 5 This is an exemplary flowchart of obtaining the shortest path according to some embodiments of this specification. DETAILED DESCRIPTION

[0041] The methods and systems provided in the embodiments of this specification are described in detail below with reference to the accompanying drawings.

[0042] Figure 1 is an exemplary flow chart of a method for dispatching an unmanned transport vehicle for an underground mine according to some embodiments of this specification. Figure 2This is an exemplary module diagram of a scheduling system for unmanned transport vehicles in underground mines according to some embodiments of the present specification; a scheduling method for unmanned transport vehicles in underground mines, comprising: obtaining underground mine map data; processing the obtained map data to obtain a point cloud map; constructing an underground mine map based on the point cloud map; obtaining driving data of each unmanned transport vehicle; the driving data includes the total time taken for vehicles to enter and leave a passing observation area; according to the current state of each unmanned transport vehicle being empty or fully loaded, the constructed underground mine map is used as a directed topological graph, and path planning is performed using the A* algorithm to obtain a path for the unmanned transport vehicle to reach each loading point or unloading point from the current position; among all the obtained paths, an optimal path is selected based on the total time taken for the unmanned transport vehicle to enter and leave the passing observation area; based on the selected optimal path, a final destination of the unmanned transport vehicle is determined; the determined optimal path is sent to the selected unmanned transport vehicle, and the transport vehicle is controlled to travel along the optimal path.

[0043] Figure 3 This is an exemplary flow chart for constructing an underground mine map according to some embodiments of this specification. The unmanned transport vehicle scheduling method of this application first needs to obtain high-precision three-dimensional map data of the underground mine. This step can be achieved through professional three-dimensional scanning equipment, such as a three-dimensional laser scanner, an RGB-D camera, a panoramic camera, etc. Specifically, a vehicle-mounted or handheld three-dimensional laser scanner can be used to perform all-round scanning and acquisition in an underground mine environment. The scanner emits a laser beam, and by measuring the time it takes for the laser beam to be reflected from the surface of an object, it calculates the three-dimensional coordinates and reflection intensity of the scanning point, thereby obtaining a large amount of point cloud data. Each point in the point cloud data contains its three-dimensional spatial position coordinates (x, y, z) and reflection intensity attributes, which truly records the geometric information of objects such as underground tunnels, chambers, and equipment.

[0044] To improve scanning efficiency and accuracy, specialized point cloud data acquisition software can be used to set and optimize scanning parameters such as scanning range, angle, and resolution. During the actual scanning process, it is important to select the appropriate scanning position and angle to obtain comprehensive, unobstructed data. In addition to vehicle-mounted or handheld scanning, fixed 3D sensors can be installed at key locations in underground tunnels to continuously collect map data. This method allows for real-time monitoring of environmental changes, providing data support for dynamic map updates.

[0045] The raw point cloud data obtained from scanning is typically large and noisy, making it unsuitable for subsequent map construction and path planning. Preprocessing is required. This process includes: point cloud filtering to remove discrete points and noise; point cloud registration to align data collected at different locations and angles; and point cloud segmentation to distinguish points belonging to different objects or regions. Preprocessed point cloud data is of higher quality and easier to use in subsequent applications.

[0046] Preferably, to address the problem of high noise and discrete points in the raw point cloud data obtained from underground mine scans, this application proposes a discrete point removal method based on a combination of statistical filtering and radius filtering. The specific steps are as follows: Statistically filter the raw point cloud data. Calculate the average distance between each point and its neighboring points. If the average distance between a point and its neighboring points exceeds a preset threshold, mark it as a discrete point. The threshold can be adaptively adjusted based on the density and noise level of the point cloud. The discrete points marked in the previous step are further verified using radius filtering. A spherical neighborhood with a preset radius r is constructed with each discrete point as the center, and the number of points within the neighborhood is counted. If the number is below the preset threshold, the point is confirmed as a discrete point and removed; otherwise, it is retained. The radius r and the number threshold can be set according to actual needs. To avoid filtering problems caused by multiple filtering, after each discrete point is removed, density analysis is performed on the remaining point cloud to calculate the average density of the point cloud. If the average density is significantly higher than the initial value, the filtering process is terminated and the filtered point cloud data is output. Perform noise analysis on the filtered point cloud data and calculate the signal-to-noise ratio (SNR) of the point cloud. If the SNR meets the expected requirements, the discrete point removal effect is considered ideal. Otherwise, adjust the filtering parameters and repeat steps 1-3 until the desired point cloud quality is achieved. This discrete point removal method comprehensively considers the overall distribution characteristics and local neighborhood characteristics of the point cloud. Through adaptive threshold control and iterative optimization, it can remove the majority of discrete points while maximally preserving the original point cloud's geometric structure, thereby improving the quality and usability of the point cloud data.

[0047] Preferably, in an underground mine environment, due to factors such as narrow space, multiple obstructions, and limited scanning angles, scanning at multiple positions and angles is usually required to obtain complete scene point cloud data. However, there are differences in posture such as rotation and translation between the point cloud data obtained from different perspectives, and point cloud registration is required to unify them into the same coordinate system. In view of the particularity of point cloud registration in underground mines, this application proposes a registration method based on point-line feature extraction and a robust ICP algorithm. The specific steps are as follows: Feature extraction is performed on the point cloud data obtained from multi-perspective scanning. The PCA algorithm is used to estimate the surface normal vector of each point cloud and extract stable point features. At the same time, the RANSAC algorithm is used to perform linear fitting on each point cloud and extract significant line features. A point-line feature descriptor is constructed. For each extracted point feature, the angle between its normal vector direction and the preset direction (such as the positive direction of the z-axis) is calculated, and the angle is discretized into multiple intervals of 0-360 degrees, each interval corresponding to a bin. The frequency of point features falling into each bin is counted to obtain a directional histogram descriptor of the point feature. Similarly, for each extracted line feature, the angle between its orientation and the preset orientation is calculated. This angle is discretized into multiple bins to obtain a directional histogram descriptor for the line feature. The point-line feature descriptor reflects the overall directional distribution of the point cloud and is rotation-invariant. Point clouds are matched using the point-line feature descriptors. The similarity of the point-line feature descriptors between different point clouds is calculated. By setting a similarity threshold, matching feature pairs are identified and an initial point cloud transformation relationship is constructed. This initial transformation relationship is optimized. A robust ICP algorithm is used to iteratively optimize the transformation matrix between the point clouds to maximize the degree of coincidence after registration. The robust ICP algorithm incorporates an M-estimator, which assigns a smaller weight to incorrect corresponding points, improving the algorithm's robustness and convergence speed. The registration results are evaluated. Metrics such as coincidence and mean squared error are calculated for the registered point clouds. If these metrics meet the preset requirements, the registered point cloud data is output; otherwise, the registration parameters are adjusted and the process is repeated until the desired registration result is achieved. This point cloud registration method fully utilizes the rich point and line feature information in underground tunnel scenes. Through robust feature extraction and matching strategies, it achieves high-precision and efficient multi-view point cloud registration, providing a high-quality point cloud data foundation for subsequent map construction. Finally, the preprocessed point cloud data is transmitted to the scheduling system's processing unit via a wired or wireless network. The scheduling system's processing unit is typically equipped with high-performance computers and large-capacity storage devices, capable of rapidly receiving and storing large amounts of map point cloud data.

[0048] The processing unit performs feature extraction on the received map data to generate a point cloud map of the underground mine. After receiving the pre-processed point cloud data from the previous step, the processing unit needs to further extract feature information from the data to facilitate subsequent map construction and semantic annotation. The goal of feature extraction is to identify key points or patches in the point cloud data that represent the scene structure and attributes, thereby reducing data redundancy and improving processing efficiency.

[0049] For underground mining environments, the following feature extraction methods can be used: Plane feature extraction: Use the RANSAC algorithm or region growing algorithm to extract significant plane features from point cloud data, such as tunnel walls, chamber roofs, etc. The extracted plane features can be used for structured representation and segmentation of the scene. Line feature extraction: Use the Hough transform or RANSAC algorithm to extract significant line features from point cloud data, such as tunnel edges, tracks, etc. The extracted line features can be used for skeleton representation and topological relationship establishment of the scene. Key point feature extraction: Use algorithms such as ISS, SIFT, SHOT, etc. to extract stable key point features from point cloud data, such as corner points, deformation points, etc. The extracted key point features can be used for scene registration, recognition and semantic annotation. After feature extraction, the original point cloud data is converted into a point cloud map composed of high-level semantic features such as planes, lines, key points, etc., laying the foundation for subsequent processing.

[0050] The resulting point cloud map is semantically annotated. Semantic annotation involves identifying different objects or areas within the point cloud map and assigning them specific semantic attributes, such as lanes, observation areas, and loading and unloading points. The inclusion of semantic information ensures that the map contains not only geometric structure information but also functional attribute information, facilitating positioning, navigation, and decision-making for autonomous vehicles.

[0051] Semantic annotation of underground mining environments can be divided into the following key steps: Lane line annotation: By extracting line features from the point cloud map, the edge lines on both sides of the lane lines are identified. Then, through clustering and fitting the edge lines, a three-dimensional spatial curve representation of the lane lines is obtained. The lane line annotation results can be used to define the driving path of the unmanned vehicle. Observation zone annotation: By analyzing the geometric topological relationships of the lane lines in the point cloud map, special areas such as lane intersections and traffic conflict zones are identified. These areas are then marked as observation zones based on their spatial extent and location attributes. The observation zone annotation results can be used to control traffic regulations for the unmanned vehicle. Passing zone annotation: By analyzing parameters such as lane width and length in the point cloud map, suitable passing locations are identified and marked as passing zones. Passing zone annotation requires comprehensive consideration of factors such as vehicle size and dynamic characteristics to avoid collisions or deadlocks. Stop line annotation: By analyzing the endpoint positions of lane lines in the point cloud map, locations where the unmanned vehicle needs to stop and wait are identified and marked as stop lines. Stop line annotation needs to be coordinated with information such as observation zones and passing zones to ensure compliance and safety of vehicle behavior. Loading and unloading point labeling: By analyzing the key point features in the point cloud map, special functional areas such as material storage areas and unloading ports are identified and marked as loading or unloading points. Loading and unloading point labeling needs to consider attribute information such as spatial location and capacity. Waiting area labeling: By analyzing the distribution and traffic volume of loading and unloading points, appropriate waiting areas are planned near the loading and unloading points and marked. Waiting area labeling requires reserving sufficient space to accommodate multiple unmanned vehicles waiting at the same time. Combining the above steps, each key element in the point cloud map is given clear semantic attributes, forming a complete semantic map of the underground mine. The semantic map reflects not only the geometric structure of the scene, but also the functional attributes of the scene. It is the basis for unmanned vehicles to perceive the environment and navigate autonomously.

[0052] Convert the semantically annotated point cloud map into a 3D raster map format to obtain an underground mine map. Considering that unmanned vehicles require rapid query and analysis of map information when making navigation decisions, it is necessary to convert the semantically annotated point cloud map into a more organized and efficient data structure. A 3D raster map is a commonly used map representation format. It divides the entire space into a grid of equal-sized cubes (also known as voxels), with each voxel represented by an integer value representing its semantic category. The specific steps for converting a point cloud map into a 3D raster map are as follows: Determine the extent and resolution of the raster map. Based on the spatial dimensions of the point cloud map, determine the starting and ending coordinates of the 3D raster map in the x, y, and z directions. Based on the required navigation accuracy of the vehicle, determine the side length of the raster map and calculate the number of voxels in the raster map in the three directions. Initialize an empty 3D raster array. Based on the map extent and resolution determined in the previous step, allocate a 3D integer array and initialize all elements to 0 to indicate unknown semantic categories. Iterate over all feature points in the point cloud map. For each feature point, the voxel index to which it belongs is calculated based on its x, y, and z coordinates, and the voxel's semantic category is updated to the feature point's semantic label. If a voxel contains multiple feature points of different categories, the category with the largest number of features is used as the voxel's semantic category. Voxels without assigned semantic categories are interpolated. Given the sparsity of point clouds, some voxels may not be occupied by any feature points. For these blank voxels, algorithms such as KNN and trilinear interpolation can be used to reasonably infer and assign values ​​based on the semantic information of neighboring voxels. After these conversion steps, a three-dimensional semantic grid map is constructed. This map stores scene information in the form of a regular array of cubes. The integer value of each voxel represents the semantic attribute of that spatial location, such as 1 for a lane, 2 for an observation zone, and 3 for a passing zone. Autonomous vehicles can conveniently query, analyze, and make decisions based on the map at the voxel level, and quickly plan driving routes based on the map.

[0053] Figure 4 This is an exemplary flow chart for obtaining driving data from an unmanned transport vehicle (AUTO) according to some embodiments of this specification. Simultaneously, the onboard equipment on the AUTO uploads vehicle status data, including location, speed, heading angle, and mission status, to the dispatch system via a wireless communication gateway. To obtain driving data from each AUTO, each AUTO is equipped with onboard equipment, including a positioning module, speed sensor, heading sensor, and mission status recording module. These devices collect real-time status information, including the vehicle's location, speed, heading angle, and mission execution progress.

[0054] The onboard devices establish a data link with the dispatch system via a wireless communication gateway. Common wireless communication technologies used in underground mines include WiFi, ZigBee, and RFID. The onboard devices package and transmit the collected vehicle status data to the dispatch system according to a pre-set data format and upload frequency. The dispatch system's communication module is responsible for receiving data packets from each onboard device and parsing the status information for subsequent driving data statistics and task scheduling.

[0055] Based on the received vehicle status data, the dispatch system further collects and compiles vehicle driving data, including metrics such as speed, passing time, queuing conditions, and loading and unloading time, to assess vehicle operational efficiency and task completion. The specific driving data acquisition method is as follows: Vehicle speed data: The dispatch system calculates the average speed of each vehicle on each road segment over a period of time (e.g., one minute) at a preset statistical frequency f (e.g., every 5 seconds). The road segment division can refer to the topological structure of the map. The average speed is calculated as: v = s / t, where s is the distance traveled by the vehicle on the road segment during the time period, and t is the length of the time period. The travel distance can be obtained from position change or odometer readings. Passing time data: The dispatch system tracks each vehicle's state machine changes, recording the time it enters and leaves the passing observation area. The difference between these two times is used as the passing time for the vehicle. The time taken for multiple passing times is accumulated and divided by the length of the statistical time period to obtain the average passing time. The passing time reflects the efficiency of vehicles when meeting on narrow roads. Stop line queue data: The dispatch system records the time each vehicle passes through each stop line along its trajectory. If a vehicle's actual arrival time at a stop line is later than its estimated arrival time, it is considered to be waiting in a queue at that stop line. The dispatch system counts the number of vehicles queued at the stop line over a period of time to obtain queue length data. Loading and unloading point queue data: The dispatch system tracks each vehicle's position changes and determines whether it enters the waiting area at a loading or unloading point. Once a vehicle enters the waiting area, it is added to the corresponding queue and the entry time is recorded. The dispatch system counts the number of vehicles queued in the waiting area over a period of time to obtain queue length data. When a vehicle completes loading or unloading and leaves the waiting area, it is removed from the queue. Loading and unloading time data: The dispatch system tracks each vehicle's status changes at loading and unloading points, recording the time the vehicle enters, the time it starts loading / unloading, the time it completes loading / unloading, and the time it leaves. The time between the start and completion of loading / unloading is the loading / unloading time, and the total time between entry and departure is the loading / unloading point dwell time. These data reflect the efficiency of loading and unloading operations.

[0056] Specifically, the statistical data are shown in the following table:

[0057]

[0058]

[0059] Figure 5 This is a method for calculating the shortest path to a loading or unloading point based on the cargo status of an unmanned transport vehicle, according to an exemplary flowchart for obtaining the shortest path shown in some embodiments of this specification. This method utilizes the topological structure of an underground mine map and combines it with the A* search algorithm to achieve rapid vehicle path planning. First, a set of feature points representing road segments in the underground mine map data is used as vertices in the topological graph. Each road segment consists of a series of ordered feature points, which can be used to abstractly represent the road segment. Then, a set of feature points that indicate connectivity between road segments is used as directed edges in the topological graph. If two road segments are connected, a directed edge exists between their endpoint feature points. The direction of the directed edge represents the permitted direction of vehicle travel. Finally, all vertices representing road segments and directed edges representing connectivity are connected to construct a directed topological graph representing the topological structure of the underground mine map. The number of vertices in the topological graph is equal to the number of road segments, and the number of directed edges is less than or equal to the square of the number of road segments.

[0060] The dispatch system obtains data representing the current mission status of each autonomous transport vehicle from driving data. The mission status reflects the type of transport mission the vehicle is currently performing, such as traveling empty to a loading point or fully loaded to a unloading point. Based on this mission status data, the dispatch system determines whether each vehicle is empty or fully loaded. Generally speaking, if the vehicle's mission status is traveling to a loading point, its cargo status is empty; if the mission status is transporting to a unloading point, its cargo status is fully loaded.

[0061] Plan the shortest path for an unmanned transport vehicle to reach the loading point. For an unmanned transport vehicle with no cargo, the scheduling system needs to plan the shortest path to the loading point. The planning steps are as follows: In the underground mine topology map, the vertex corresponding to the vehicle's current position is set as the path starting point. The vertices corresponding to each loading point in the topology map are sequentially set as the path ending points. Use the A* algorithm to search for the shortest path from the path starting point to each path ending point. The A* algorithm is a heuristic search algorithm that uses an evaluation function to assess the path cost from the starting point to the end point, prioritizing the most promising paths. In the underground mine scenario, path cost can be defined using metrics such as segment length, travel time, and energy consumption. The A* algorithm's search process is similar to breadth-first search, but it uses a priority queue to select the next search node, greatly improving search efficiency. The shortest path found is saved as the optimal path for the vehicle to reach each loading point.

[0062] Plan the shortest path for the autonomous transport vehicle to the unloading point. For an autonomous transport vehicle that is fully loaded with cargo, the scheduling system needs to plan the shortest path to the unloading point. The planning steps are similar to the previous case: In the underground mine topology map, the vertex corresponding to the vehicle's current position is set as the path starting point. The vertices corresponding to each unloading point in the topology map are set as the path endpoints. Use the A* algorithm to search for the shortest path from the path starting point to each path endpoint. Save the shortest path found as the optimal path for the vehicle to reach each unloading point.

[0063] In summary, this application leverages the topological structure of the underground mine map to construct a directed topological graph that facilitates path searching. Based on the different cargo states of the unmanned vehicle, the A* algorithm is used to search for the shortest path from the vehicle's current location to the target loading and unloading point on the topological graph. Compared to traditional road network search algorithms, this application fully utilizes the map's topological information, significantly reducing the search space and improving path planning efficiency.

[0064] The path selection module first evaluates the pros and cons of each path from the set of alternative paths that meet the constraints generated in the previous step, and then selects the optimal path. The factors that affect the quality of the path are mainly: Time spent in the missed vehicle observation area: The more missed vehicle observation areas there are on the path, the longer the vehicle spends waiting for the missed vehicle, and the higher the time cost of the path. Estimated loading time: The shorter the waiting time in the queue at the loading point corresponding to the end of the path, the earlier the vehicle can start loading, thereby shortening the entire transportation cycle. The path selection module calculates the estimated arrival time T for each path. n To judge the quality of the path. Estimated arrival time T n The calculation formula is:

[0065]

[0066] Where n represents the number of the loading point, there are N loading points in total; T n T represents the time when the transport vehicle is expected to arrive at the loading point n; now represents the current time; P represents the number of paths for the transport vehicle to reach the loading point n; L represents the number of road sections in the path p for the transport vehicle to reach the loading point n; s l represents the length of the lth segment of the path p for the transport vehicle to reach the loading point n; v l represents the speed of the transport vehicle on the lth road section of the path p to the loading point n; M represents the number of vehicle-passing observation zones on the path p of the transport vehicle to the loading point n; s m v represents the distance of the main path section within the mth vehicle passing observation area on the expected driving trajectory in the path p of the transport vehicle to the loading point n; mrepresents the driving speed of the transport vehicle in the mth overtaking observation zone on the main path section of the expected driving trajectory in the path p of the loading point n; C represents the number of transport vehicles whose arrival time in the overtaking observation zone overlaps with the time of the current transport vehicle and whose driving priority is higher than the current transport vehicle, and the driving priority is based on the first-come, first-served principle; s m It represents the time taken to pass the other vehicle, which is equal to the total time taken for the other vehicle to enter and leave.

[0067] Intuitively speaking, T n The calculation of T takes into account two parts: the travel time of the route and the delay time of the missed vehicle. The travel time is obtained by dividing the length of each section of the route by the speed and then summing the results. The delay time of the missed vehicle is obtained by multiplying the number of missed vehicles in each missed vehicle observation area (C) by the average missed vehicle time (ta) and then summing them. By comparing the T of different routes, n value, the path selection module can select the optimal path with the shortest estimated arrival time.

[0068] After determining the optimal path, the path selection module further compares the queue waiting time of different loading points at the end of the path, and selects a loading point with the highest scheduling efficiency as the final destination.

[0069] The length of the waiting time at the loading point depends on two factors: the number of vehicles in the queue when the transport truck arrives at the loading point, and the loading efficiency of the loading point. The more vehicles in the queue, the lower the loading efficiency, and the longer the waiting time.

[0070] The route selection module calculates the estimated loading start time T for each loading point sn To judge the length of waiting time. sn The calculation formula is:

[0071]

[0072] Among them, T now represents the current time; Q represents the number of transport vehicles waiting for loading in the waiting area at loading point n when the transport vehicle arrives at loading point n; t1 represents the idle waiting time of the forklift before transport vehicle q arrives at the loading point; t2 represents the driving time from the waiting area to the loading point; t3 represents the loading time of the transport vehicle; t4 represents the time it takes for the transport vehicle to complete loading and drive from the current loading point to the main road; T sn The calculation of T takes into full account the various factors that affect the waiting time at the loading point and gives the estimated time for the start of loading. sn The smaller the value, the shorter the waiting time of the vehicle at the loading point and the higher the scheduling efficiency.

[0073] Finally, the path selection module places the optimal path’s endpoint at the loading point T snThe one with the smallest value is selected as the final destination, generating the dispatch decision result of the unmanned vehicle. An optimal path is selected using the following formula:

[0074]

[0075] Where n represents the loading point number to which the transport vehicle is finally dispatched; T n T represents the time when the transport vehicle is expected to arrive at the loading point n; sn = represents the time when the shovel loader at loading point n is expected to start loading the current transport vehicle; N represents the total number of loading points for transport vehicle scheduling to choose from; In summary, the scheduling path selection method proposed in this application cleverly combines path planning with destination selection, and uses the estimated arrival time T n and the expected loading start time T sn Two indicators are used to optimize path selection and loading point selection, striving to achieve the global optimal scheduling decision. Compared with the strategy of simply pursuing the shortest path, this application further selects the loading point with the shortest waiting time based on the optimal path, shortening the waiting time of vehicles and improving the transportation efficiency of the mine.

[0076] After selecting the optimal route, the scheduling system needs to further determine the final destination of the unmanned transport vehicle, that is, the specific loading point. This method selects the destination based on the expected idle time of each loading point, striving to reduce the waiting time of the vehicle and improve loading efficiency. Calculate the time T when the w-th transport vehicle in the waiting area of ​​loading point n is expected to finish loading. w,n,e , the calculation formula is as follows:

[0077]

[0078] Among them, T w,n,s T represents the time when the w-th transport vehicle in the waiting area of ​​loading point n arrives at the waiting area; load It represents the total time taken by the transport vehicle to go from the waiting area to the loading point, load the vehicle, and leave the loading point; the calculated load loader is estimated to be the time when the current transport vehicle starts loading T sn and the time T at which the transport vehicle is expected to finish loading w,n,e Compare and select the loading point with the longest idle time as the final destination of the transport vehicle, and update the destination of the optimal path to the selected final destination; for example, there are 3 vehicles in the waiting area of ​​loading point A, and they arrive at the waiting area at 10:00, 10:02, and 10:05 respectively. load is 5 minutes. Then the T of these 3 cars w,n,e The time is 10:05, 10:07 and 10:10 respectively. The dispatching system calculates the T of each vehicle in the waiting area of ​​each loading point one by one according to the above method. w,n,e , until the T of the last car is obtained w,n,e .

[0079] The dispatching system calculates the T of the last waiting area vehicle at each loading point in the previous step. w,n,e , and the previously calculated time T when the scraper is expected to load the current dispatched vehicle sn Compare. If T sn Later than T w,n,e , it means that when the current dispatched vehicle arrives at the loading point, all queued vehicles have finished loading. At this time, the idle time of the loading point is T sn With T w,n,e If T sn Earlier than T w,n,e , it means that the current dispatched vehicle needs to wait for loading after the existing queued vehicles. At this time, the idle time of the loading point is 0. The dispatching system compares T sn and T w,n,e , we can get the estimated idle time of each loading point. The longer the idle time, the shorter the waiting time of the dispatched vehicle at the loading point.

[0080] The dispatching system ranks all loading points by their idle time and selects the loading point with the longest idle time as the final destination for the autonomous transport vehicle. If multiple loading points have the same idle time and are the longest, the dispatching system can determine the final destination based on other factors such as the loading point's production volume and priority. Finally, the dispatching system updates the optimal route's destination to the selected final destination, completing the destination determination.

[0081] After determining the optimal route and final destination, the dispatch system transmits this decision to the corresponding unmanned transport vehicle via a wireless communication gateway. Upon receiving the dispatch command, the onboard equipment automatically controls the vehicle to follow the optimal route until it reaches the final destination. During travel, the onboard equipment continuously collects data such as vehicle location, speed, and cargo status, and uploads it to the dispatch system in real time via the wireless network. The dispatch system receives the latest status data uploaded by the vehicle and dynamically updates the underground mine map, such as road construction and congestion. Based on the latest data, the dispatch system re-executes route planning and destination selection, generates a new dispatch decision, and transmits it to the vehicle. With the real-time upload of vehicle status data and the dynamic updating of dispatch decisions, the dispatch system and the fleet of unmanned vehicles form a closed-loop real-time dispatch system. This system can quickly adapt to dynamic changes in the mine environment, continuously optimize vehicle scheduling, and ensure efficient ore transportation. By predictively comparing the idle time of different loading points, the loading point with the shortest wait time is selected as the final destination, avoiding long queues at loading points and reducing vehicle idle time. Combined with the previous path optimization strategy, this scheduling method strikes a balance between the shortest path and the shortest waiting time, achieving an overall improvement in the mine vehicle scheduling efficiency.

Claims

1. A method for dispatching an unmanned transport vehicle for an underground mine, comprising: Obtain underground mine map data; Process the acquired map data to obtain a point cloud map; Build underground mine maps based on point cloud maps; Obtain driving data for each unmanned transport vehicle; the driving data includes the total time it takes for vehicles to enter and exit the passing observation area; Based on the current status of each unmanned transport vehicle (empty or fully loaded), the constructed underground mine map is used as a directed topology graph, and the path planning is performed using the A* algorithm to obtain the path of the unmanned transport vehicle from its current position to each loading or unloading point; Among all the obtained paths, the optimal path is selected based on the total time taken by the driverless transport vehicle to enter and leave the passing vehicle observation area; Determine the final destination of the unmanned transport vehicle based on the selected optimal path; Send the determined optimal path to the selected unmanned transport vehicle, and control the transport vehicle to drive along the optimal path; An optimal path is selected using the following formula: ; Where n represents the loading point number to which the transport vehicle is finally dispatched; It represents the time when the transport vehicle is expected to arrive at the loading point n; It indicates the time when the scraper at loading point n is expected to start loading the current transport vehicle; N is the total number of loading points available for transport vehicle scheduling; The time when the loader at loading point n is expected to start loading the current transport vehicle is calculated by the following formula : ; in, represents the current time; Q represents the number of transport vehicles waiting for loading in the waiting area of ​​loading point n when the transport vehicle arrives at loading point n; Indicates transport vehicle The idle waiting time of the forklift before arriving at the loading point; Indicates the driving time from the waiting area to the loading point; Indicates the loading time of the transport vehicle; Indicates the time it takes for the transport vehicle to complete loading and leave the current loading point to reach the main route; The estimated time when the transport vehicle arrives at the loading point n is calculated using the following formula : ; Where n represents the number of the loading point, and there are N loading points in total; It represents the time when the transport vehicle is expected to arrive at the loading point n; represents the current time; P represents the number of paths for the transport vehicle to reach the loading point n; L represents the number of road segments in the path p for the transport vehicle to reach the loading point n; represents the number of the transport vehicle's path p to the loading point n. The length of the road section; represents the first step of the transport vehicle in the path p to the loading point n. The driving speed of each road section; M represents the number of overtaking observation zones in the path p of the transport vehicle to the loading point n; The distance of the main path section within the mth vehicle passing observation area on the expected driving trajectory in the path p of the transport vehicle to the loading point n; represents the driving speed of the transport vehicle on the main path section of the expected driving trajectory within the mth overtaking observation zone in the path p of the loading point n; C represents the number of transport vehicles whose arrival time in the overtaking observation zone overlaps with the time of the current transport vehicle and whose driving priority is higher than the current transport vehicle, and the driving priority is based on the first-come, first-served principle; Indicates the average time it takes to miss a vehicle.

2. The method for dispatching unmanned transport vehicles for underground mines according to claim 1, characterized in that: Based on the selected optimal path, determine the final destination of the unmanned transport vehicle, including: Calculate the estimated time when the w-th transport vehicle in the waiting area at loading point n will finish loading , the calculation formula is as follows: ; in, represents the time when the w-th transport vehicle in the waiting area at loading point n arrives at the waiting area; It indicates the total time taken by the transport vehicle to travel from the waiting area to the loading point, load the vehicle, and leave the loading point; The calculated time when the scraper is expected to start loading the current transport vehicle and the time when the transport vehicle is expected to finish loading After comparison, the loading point with the longest idle time is selected as the final destination of the transport vehicle, and the destination of the optimal path is updated to the selected final destination.

3. The method for dispatching unmanned transport vehicles for underground mines according to claim 1, characterized in that: Construct underground mine maps, including: Conduct three-dimensional scanning of underground mines to obtain map data containing spatial location information and distance information of underground mines; Perform feature extraction on the obtained map data to obtain a point cloud map of the underground mine; Perform semantic annotation on the obtained point cloud map; The semantically annotated point cloud map is converted into a three-dimensional raster map data format to obtain an underground mine map.

4. The method for dispatching unmanned transport vehicles for underground mines according to claim 3, characterized in that: Semantic annotation includes: Mark out the lane feature point set in the point cloud map to form lane information, which is used to represent the driving path of the unmanned transport vehicle; Mark out the observation area feature point set in the point cloud map to form the observation area information, which is used to represent the area where the unmanned transport vehicle observes the vehicle passing by; Mark out the set of feature points of the passing area in the point cloud map to form passing area information, which is used to represent the area where the unmanned transport vehicle passes; Mark out the stop line feature point set in the point cloud map to form stop line information, which is used to indicate the location where the unmanned transport vehicle stops and waits; Mark out the set of loading point feature points in the point cloud map to form loading point information, which is used to indicate the location where the unmanned transport vehicle loads the goods; Mark out the unloading point feature point set in the point cloud map to form the unloading point information, which is used to indicate the location where the unmanned transport vehicle unloads the goods; Mark out the waiting area feature point set in the point cloud map to form the waiting area information, which is used to represent the area where the unmanned transport vehicle is waiting to be loaded or unloaded.

5. The method for dispatching unmanned transport vehicles for underground mines according to claim 4, characterized in that: Obtain driving data of each unmanned transport vehicle, including: The status data of the unmanned transport vehicle is obtained through the wireless gateway; the status book contains the current position, speed, heading angle, mission status, driving status and state machine information of the unmanned transport vehicle; Based on the acquired status data, the driving data of each unmanned transport vehicle is obtained: According to the preset statistical frequency f, the average speed of each unmanned transport vehicle on each road section within the preset time period is counted to obtain the speed data; According to the preset statistical frequency f, combined with the time interval of the change of the state machine of the unmanned transport vehicle, the total time consumed by the unmanned transport vehicle in the passing vehicle observation area is counted to obtain the passing vehicle time consumption data; According to the preset statistical frequency f, the actual arrival time and estimated arrival time of the unmanned transport vehicle at each stop line on the preset driving trajectory are counted to obtain the stop line queue data; Based on the preset statistical frequency f, determine whether the current position of the unmanned transport vehicle is within the range of the waiting area of ​​the loading point or unloading point on the preset driving trajectory. If so, add the corresponding unmanned transport vehicle to the waiting area queue of the loading point or unloading point, and record the corresponding arrival time of the waiting area to obtain the queue data of the waiting area of ​​the loading point or unloading point; According to the time interval of the change of the state machine of the unmanned transport vehicle, the time interval between the unmanned transport vehicle entering the loading point or unloading point, completing the loading or unloading operation, and leaving the loading point or unloading point is counted to obtain the time data of entering, loading, unloading, or leaving; The obtained vehicle speed data, time consumed in passing other vehicles data, queue data at the stop line, queue data in the waiting area of ​​the loading point or unloading point, and time consumed in entering, loading, unloading or leaving the vehicle are used as driving data.

6. The method for dispatching unmanned transport vehicles for underground mines according to claim 5, characterized in that: Based on the current status of each unmanned transport vehicle, whether it is empty or fully loaded, the path of the unmanned transport vehicle from its current position to each loading point or unloading point is obtained, including: The feature point sets representing road sections in the constructed underground mine map data are used as vertices of the topological graph, and the feature point sets representing the connectivity between road sections are used as directed edges of the topological graph. By connecting all vertices and directed edges, a directed topological graph representing the topological structure of the underground mine map is constructed; Obtain data representing the current mission status of the unmanned transport vehicle from the driving data, and determine whether the current cargo status of each unmanned transport vehicle is empty or fully loaded based on the mission status data; When the current cargo status of the unmanned transport vehicle is empty, in the constructed directed topology graph, the vertex corresponding to the current position of the unmanned transport vehicle is set as the starting point of the path, and the vertex corresponding to each loading point in the underground mine map is set as the end point of the path in turn. The algorithm searches for the shortest path from the starting point of each path to the end point of each path, and obtains the shortest path for the current unmanned transport vehicle to reach each loading point from its current position; When the current cargo state of the unmanned transport vehicle is fully loaded, in the constructed directed topology graph, the vertex corresponding to the current position of the unmanned transport vehicle is set as the starting point of the path, and the vertex corresponding to each unloading point in the underground mine map is set as the end point of the path in turn. The algorithm searches for the shortest path from the starting point of the path to the end point of each path, and obtains the shortest path for the current unmanned transport vehicle to reach each unloading point from its current position.

7. A dispatching system for unmanned transport vehicles in underground mines, characterized in that: include: At least one processing unit; used to execute instructions to implement the scheduling method for underground mining unmanned transport vehicles according to any one of claims 1 to 6.

Images