Mine road network self-learning dynamic construction and updating method based on vehicle-mounted positioning

Abstract

The invention discloses a self-learning dynamic construction and updating method for a mine road network based on vehicle-mounted positioning. The method comprises the following steps: S1, collecting vehicle moving trajectory data, and preprocessing the obtained data to obtain structured data; s2, performing statistical analysis on a speed sequence formed by each mine transport vehicle in a preset time window so as to realize automatic identification of a vehicle load state; s3, identifying the slope trend of the mine road, and generating a corresponding slope penalty coefficient; s4, constructing a basic framework of a mine vehicle road network, and respectively establishing a no-load road network diagram and a full-load road network diagram on the road network framework; s5, calculating a dynamic weight; s6, updating the dynamic weight; and S7, online self-learning and versioning updating are carried out, and a versioning management and continuous optimization mechanism of the road network is formed. The method can be used for reducing the system deployment cost and enhancing the adaptive capacity under the change of the mining area road environment.

Description

Technical Field

[0001] This invention relates to the field of intelligent transportation and vehicle scheduling technology in mines, specifically to a method for self-learning dynamic construction and updating of mine road networks based on vehicle positioning. Background Technology

[0002] With the development of smart mines and unmanned transportation systems, vehicle scheduling and route planning in mining areas have become core technical aspects affecting production efficiency and transportation safety. Currently, the industry widely adopts methods such as manually surveying to construct elevation maps, relying on weighing systems to identify load status, or manually maintaining road traffic information.

[0003] A mining car weighing system and method, disclosed in CN102788635A, uses a weighing sensor to identify the vehicle's load status. However, this type of method still has the following prominent problems in practical applications: High cost and complex maintenance: Existing systems typically rely on high-precision surveying equipment or multiple types of sensors (LiDAR, weighing instruments, tilt sensors, etc.), resulting in large hardware investments and high deployment difficulty.

[0004] Data update lag: Mine roads are frequently affected by weather, blasting, and vehicle traffic, while most existing maps use a periodic offline update mechanism, which cannot reflect road condition changes in real time, resulting in a lag in the dispatch system's response to scenarios such as collapses, water accumulation, and temporary closures.

[0005] Load identification relies on fixed hardware or manual input: weighing systems are easily interfered with on bumpy roads, and some mines still rely on drivers to manually enter empty / full load status, which has problems such as poor real-time performance, inaccurate identification, and high maintenance costs.

[0006] The route planning lacks intelligence: existing scheduling systems are mostly based on the shortest distance or shortest time strategy (such as the classic rule base of commercial scheduling systems Modular and Wencomine), which do not incorporate dynamic factors such as road slope, curvature, congestion, and risk events into the cost function, making it difficult to adapt to complex terrain and frequently changing transportation environments.

[0007] Lack of self-learning ability: Although vehicles generate a large amount of trajectory data every day, the existing system lacks a mechanism to use the trajectory to optimize the road network structure or dynamically update the right-of-way, and cannot achieve "data-driven" adaptive learning of road conditions, resulting in long-term fixed scheduling paths and lagging optimization.

[0008] In summary, existing technologies generally rely on high-cost hardware, manual updates, or static rules, making it difficult to meet the needs of unmanned mining trucks for low-cost deployment, real-time road condition perception, and self-learning scheduling. Therefore, there is an urgent need for a method that can achieve load identification, slope inference, dynamic updating of road weights, and self-learning construction of the road network based solely on vehicle positioning data, in order to reduce the difficulty of system deployment and improve the level of intelligent scheduling. Summary of the Invention

[0009] To overcome the above technical problems, the purpose of this invention is to provide a self-learning dynamic construction and update method for mine road networks based on vehicle positioning. This method can complete vehicle load status identification and road slope trend inference solely based on vehicle positioning trajectory data, significantly reducing system deployment costs and improving dynamic adaptability to changes in the mine road environment.

[0010] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A self-learning dynamic construction and update method for mine road networks based on vehicle positioning includes the following steps; S1: Collect vehicle trajectory data and preprocess the acquired data to obtain structured data; S2: Based on the structured trajectory data obtained in step S1, generate a speed sequence for each mining transport vehicle within a preset time window. Statistical analysis is performed to achieve automatic identification of vehicle load status; S3: Without relying on road terrain elevation data, based on the vehicle load status identification results obtained in step S2, the slope trend of the mining area road is identified by utilizing the difference in vehicle speed between empty and fully loaded states, and a corresponding slope penalty coefficient is generated. S4: Based on the structured trajectory data obtained in step S1, as well as the vehicle load status and slope penalty coefficient, construct the basic framework of the mining vehicle road network, and establish an empty road network map and a full-load road network map on the road network framework to support the path planning requirements under different load conditions. S5: Calculate dynamic weights based on the unloaded and fully loaded road network maps; S6: Update the dynamic weights; S7: Based on dynamically updated weights, online self-learning and version updates are implemented. During system operation, mining vehicle operation data is summarized and analyzed according to a preset update cycle. The road network structure and road segment parameters in the mining area are updated online through self-learning, forming a version management and continuous optimization mechanism for the road network.

[0011] Specifically, S1 is: Continuous operating trajectory data is obtained from the on-board positioning device of the mining transport vehicle. The operating trajectory data includes the following fields: Timestamp t, longitude Lon, latitude Lat, instantaneous velocity v, direction angle θ; The trajectory data is preprocessed, and the specific preprocessing steps are as follows: (1) Trajectory denoising and smoothing: First, abnormal trajectory points in the running trajectory data are identified and removed. These abnormal trajectory points include those caused by positioning drift or positioning error. After removing the abnormal trajectory points, the remaining trajectory data is smoothed using Kalman filtering or moving average methods to improve trajectory continuity and speed data stability, thereby providing a reliable data foundation for subsequent speed normalization and road segmentation. (2) Speed ​​normalization processing: Based on the completion of trajectory denoising and smoothing, the smoothed speed data is normalized according to the vehicle type and corresponding operating conditions to make the speed data of different vehicles or different operating conditions comparable, providing a unified data scale for subsequent load state identification based on speed features. (3) Road segment division: Based on the speed normalization process, the continuous running trajectory of the vehicle is segmented according to the changes in the vehicle's driving direction, parking behavior, and loading or unloading position. The running trajectory is divided into multiple independent road segments so that vehicle state analysis and road parameter calculation can be performed on a segment-by-segment basis in subsequent steps. Through the preprocessing steps executed sequentially above, the final output is structured data: .

[0012] in, and They represent the first The longitude and latitude coordinates corresponding to each trajectory point This represents the normalized velocity value of the corresponding trajectory point. This represents the timestamp of the corresponding trajectory point. Indicates the index of the trajectory point. This represents the total number of trajectory points.

[0013] S2 specifically includes the following steps: (1) Steps for calculating the velocity probability density function: For the velocity sequence Perform statistical processing to calculate the probability density function of vehicle speed. To characterize the speed distribution features of vehicles under different operating conditions; among them, This represents the normalized speed value of the vehicle at the trajectory point; Show corresponding speed value The probability density of occurrence; (2) Steps for modeling the bimodal velocity distribution: Based on the probability density function A Gaussian mixture model is used to fit the vehicle speed distribution, which is expressed as a weighted mixture model of two Gaussian distributions, as follows: in, and Let represent the mixture weights of two Gaussian distributions, and satisfy . =1, This represents the first Gaussian distribution with a mean of 1 / 2. The variance is , This represents the second Gaussian distribution with a mean of 1 / 2. variance is Among them, the velocity distribution with a smaller mean The speed distribution with a relatively large mean corresponds to the fully loaded operating state of the vehicle. The corresponding vehicle's unloaded operating status; (3) Steps for determining the velocity threshold: By solving for the intersection of the two Gaussian distributions, a speed threshold for distinguishing vehicle load states is determined. The speed threshold satisfies the following relationship: The speed threshold This indicates the dividing speed at which a vehicle transitions from a fully loaded state to an unloaded state. (4) Load condition determination steps: Based on the speed threshold For each trajectory point, the corresponding velocity value The load condition is determined according to the following rules: in: =1 indicates that the vehicle is at the... Each trajectory point is in a fully loaded state; =0 indicates that the vehicle is in the... Each trajectory point is in an unloaded state; Indicates the first The normalized velocity value corresponding to each trajectory point.

[0014] Through the above steps, the load status markers corresponding to each trajectory point during vehicle operation are obtained. These load status markers serve as the input basis for inferring the road slope trend in step S3 and constructing the empty / full-load road network in step S4.

[0015] The system uses a sliding window or time period to recalculate the threshold in order to adapt to the dynamic changes in the mine environment; Output: Payload label for each vehicle positioning sample .

[0016] S3 specifically includes the following steps: (1) Steps for calculating the average speed of a road segment: Using each road segment obtained in step S1 as an analysis unit, the average operating speed of vehicles under empty and fully loaded conditions is calculated as follows: in: This represents the average normalized operating speed of a vehicle when it is unloaded on the corresponding road segment. This represents the average normalized operating speed of a vehicle when it is fully loaded on the corresponding road segment. Indicates the first The normalized velocity values ​​corresponding to each trajectory point; This indicates the first value obtained in step S2. The load state markers corresponding to each trajectory point, among which Indicates no-load status. Indicates a fully loaded state; This indicates the number of empty-load status trajectory points within the corresponding road segment; This indicates the number of fully loaded trajectory points within the corresponding road segment; (2) Steps for calculating the speed difference: Based on the average speed under unloaded and fully loaded conditions, the speed difference for the corresponding road segment is calculated, and its expression is as follows: in, It represents the average speed difference of vehicles under empty and fully loaded conditions on the same road segment, and is used to characterize the degree of influence of the road segment on the operation of vehicles under different load conditions; (3) Steps for determining slope trend: According to the average speed difference The numerical value and sign, combined with the preset speed difference threshold To determine the slope trend of the corresponding road segment, the following rules apply: in, The speed difference threshold is used to suppress false judgments caused by random fluctuations or noise. (4) Steps for generating the slope penalty coefficient: Based on the determined slope trend, the speed difference is mapped to a slope penalty coefficient used for path weight calculation. The calculation method is as follows: in: This represents the gradient penalty coefficient for the corresponding road segment, used to characterize the impact of gradient on vehicle operating costs; This is an empirical coefficient used to adjust the degree of influence of uphill and downhill on path weights.

[0017] Through the above steps, a corresponding slope trend category and slope penalty coefficient are generated for each road segment. The slope penalty coefficient is used as the input parameter for dynamic weight calculation in step S5.

[0018] S4 specifically includes the following steps: (1) Steps for constructing the road network framework: Based on the trajectory points formed by vehicles moving within the mining area, an abstract model of the mining area roads is created to generate a road network skeleton structure, which is represented as a graph structure. ,in: a. This represents a set of road network nodes, which include vehicle loading points, unloading points, and road intersections. b. This represents the set of road network edges, which consists of continuous driving trajectory segments formed by vehicles between adjacent nodes.

[0019] (2) Steps for constructing a bipartite graph under no-load / full-load conditions: In the road network framework Based on the vehicle load state markings obtained in step S2 Construct unloaded road network maps respectively And a full-load road network map ,in: a. Unloaded road network map Used to describe the road operating characteristics of vehicles when they are unloaded, with the objective of time efficiency; b. Fully loaded road network map Used to describe the road operating characteristics of vehicles when fully loaded, with the goal of safety and stability; c. and These represent the sets of road segment edges corresponding to the unloaded and fully loaded states, respectively.

[0020] (3) Steps for calculating road segment statistical parameters: Regarding the aforementioned unloaded road network map And a full-load road network map Each road segment in The average operating speed and risk parameters under the corresponding load conditions are calculated as follows: in: a. Indicates road segment Under load conditions The average normalized running speed under these conditions; b. Indicates no-load status. Indicates a fully loaded state; c. Indicates the first one that falls on the road segment Normalized velocity corresponding to each trajectory point; d. Indicates road segment Under load conditions The number of corresponding trajectory points.

[0021] (4) Steps for recording road segment attribute parameters: The unloaded road network map And a full-load road network map In the middle, the following attribute parameters are recorded for each road segment edge: a. ; b. Slope penalty coefficient generated in step S3 ; c. Risk parameters obtained from historical operational data statistics ; d. Road segment weight-related parameters used for dynamic weight calculation in subsequent step S5, including road segment distance parameters, slope penalty coefficient, road curvature parameters, risk coefficient, and speed factor parameters.

[0022] Through the above steps, a bipartite graph structure of an unloaded road network map and a fully loaded road network map is formed on the same road network skeleton, providing a basic road network model for dynamic weight calculation and route planning in step S5.

[0023] Specifically, S5 is: For the unloaded road network map constructed in step S4 With full load of road network map Each road segment in The dynamic weight of the road segment is calculated by considering factors such as distance, gradient penalty, curvature, risk, and speed. ,in These represent the empty and full-load states, respectively, and the dynamic weights are used for segment cost evaluation in subsequent path planning. (1) Dynamic weight model The factors are defined as follows: ; in, Indicates road segment The first one arranged in chronological order The coordinates of the trajectory points, dist( () represents the planar distance between two points on the trajectory. This indicates the number of trajectory points corresponding to this road segment; ; in, The average speed difference between empty and fully loaded sections of the same road segment calculated in step S3. ( ) represents the slope penalty mapping function defined in step S3 (i.e., the slope penalty coefficient generated from the uphill / flat / downhill trend); curvature factor The curvature penalty is calculated based on the steering angle of adjacent trajectory segments within a road segment. The curvature factor is defined as follows: in, This is the curvature weighting coefficient. For road section The average curvature index; the average curvature index can be obtained by averaging the absolute values ​​of the turning angles of continuous trajectory segments within the road segment; ; ; in, The road segment risk coefficient is obtained in step S6 by observing risks and updating them through self-learning. ; in, The road segments obtained in step S4 Under load conditions The average normalized running speed is as follows. As the reference velocity constant, To avoid positive constants with a denominator of zero.

[0024] (2) Weighting priority of no-load / full-load To reflect optimization preferences under different load conditions, different factor index weights are applied to the unloaded road network map and the fully loaded road network map, and the dynamic weights are further expressed as follows: in: a. ; indicates the dynamic weight of the road segment under no-load conditions, which is taken under no-load conditions. Larger values ​​are used to increase the influence weight of the speed factor; b. This represents the dynamic weight of a road segment under full load conditions. Under full load conditions, the weight is... and Larger values ​​are used to increase the influence weight of slope factor and risk factor; c. , , , , , , , These are the preset factor weight index parameters.

[0025] Based on the dynamic weight The minimum weight path search is performed in the corresponding empty or full-load road network map to obtain a driving path that meets the requirements of efficiency and safety.

[0026] Specifically, S6 is: Based on the structured trajectory data obtained in step S1 and the road segmentation results in step S4, the empty road network map is analyzed. Each road segment in The characteristics of vehicle operation behavior in this road segment are statistically analyzed to generate the road segment observation risk value. The road segment risk coefficient is iteratively updated using an exponential smoothing self-learning update mechanism. The updated risk coefficient is used for the dynamic weight calculation in step S5. (1) Steps for calculating road section observation risks: For each road segment The vehicle operation behavior within a preset statistical period is statistically analyzed to obtain the observed risk value. The calculation method is as follows: in: Indicates the load condition. Unloaded 1 indicates full load; Indicates road segment Under load conditions The observed risk value; Indicates the number of vehicles on the road segment within the statistical period. Upper and load conditions are The number of times emergency braking occurred; Indicates the number of vehicles on the road segment within the statistical period. Upper and load conditions are The standard deviation of the velocity over time; Indicates the number of vehicles on the road segment within the statistical period. Upper and load conditions are The number of abnormal parking incidents; This indicates a risk correction factor related to weather or road conditions. , , , These are preset weighting coefficients used to adjust the degree of influence of each observation indicator on the observation risk value.

[0027] (2) Risk coefficient self-learning update steps: For each road segment Establish risk coefficient And based on observed risk values Update using exponential smoothing rules: in: Indicates the period of statistical analysis After the end, the section Under load conditions The risk factor for updates is as follows; This indicates the risk coefficient for the previous statistical period; This represents the observed risk value obtained in the current statistical period; Let be the learning rate parameter, satisfying 0 < 0. <1, used to control the fusion ratio of historical risk and current observed risk.

[0028] Updated Stored in the road segment attribute parameters and used as a risk factor in step S5. The calculation input enables self-learning iterative updates of road segment risk parameters.

[0029] (3) Update the results and write them into the database for use in the next cycle weight calculation.

[0030] S7 specifically includes the following steps: (1) Steps for summarizing running data cycles Within a preset time period, structured trajectory data generated by vehicles operating within the mining area are aggregated. This time period can be a shift cycle, a daily cycle, or other preset cycles. The aggregated data serves as input data for the road network update in this cycle.

[0031] (2) Steps for self-learning and updating road segment parameters; Based on the aggregated runtime data, perform the following update operations in sequence: Summarize the data for the day and recalculate the speed threshold, slope coefficient, and risk weight; Based on step S3, update the slope trend determination results and corresponding slope penalty coefficients for each road segment; According to step S4, update the statistical parameters such as the average operating speed of the road segment under empty and full load conditions; According to step S6, update the road segment risk coefficient to reflect the changes in safety risks during the most recent operating cycle; The above updates enable road segment parameters to be continuously corrected as vehicle operation data accumulates.

[0032] (3) Road network version generation and management steps Based on the updated road segment parameters, a corresponding road network version is generated, and a version identifier is assigned to each road network version. Simultaneously, historical road network versions are retained for backtracking analysis or anomaly comparison. In subsequent route planning, the system prioritizes using the latest generated road network version for route calculation.

[0033] (4) Online application and abnormal trigger update steps During the online operation of the system, if significant abnormalities in vehicle operation behavior or sudden changes in road segment parameters are detected, the system can trigger a local update or early update mechanism for the relevant road segments to ensure that the road network model can reflect changes in the road conditions in the mining area in a timely manner.

[0034] Through the above-mentioned online self-learning and version update steps, the mine vehicle road network can achieve closed-loop optimization of "run-learn-update-run" during continuous operation, thereby improving the mine vehicle dispatching system's adaptability to road changes and operational safety.

[0035] The beneficial effects of this invention are: This invention uses only vehicle positioning trajectory data of mining vehicles as the basic data source in step S1, without the need to build high-precision elevation maps or deploy expensive sensors such as weighing equipment and lidar, thus significantly reducing system deployment and maintenance costs. At the same time, the vehicle positioning method is not limited by specific positioning technologies and is compatible with multiple positioning systems such as GPS and Beidou, making it suitable for application scenarios where the mining environment changes frequently. This invention automatically identifies the vehicle load status based on the bimodal characteristics of vehicle speed distribution in step S2, eliminating the need for manual input of whether the vehicle is empty or fully loaded, avoiding errors caused by manual labeling, improving the real-time performance and accuracy of load status identification, and providing a reliable foundation for subsequent road parameter analysis. In step S3, this invention uses the difference in vehicle speed between empty and fully loaded states to determine the road slope trend. It does not rely on road elevation or slope sensors, and realizes slope trend recognition based on operating data, providing effective road feature information for path planning in complex mining environments. This invention constructs an unloaded road network map and a fully loaded road network map on the same road network skeleton through step S4, and records the running speed, gradient penalty coefficient and risk parameters of each road segment under the corresponding load conditions, so that the road network model can reflect the differences in vehicle operation under different load conditions and improve the pertinence and rationality of the route planning results. In step S5, this invention comprehensively considers road segment distance, slope penalty, road curvature, operational risk, and speed factor to dynamically calculate the road segment weights and adjust the weight priority of each factor according to the vehicle load status, so that the route planning can achieve a balance between operational efficiency and safety, and improve the intelligence level of vehicle dispatching in the mining area. This invention performs statistical analysis on vehicle operation behavior in step S6 and uses a self-learning update mechanism to iteratively update the road segment risk coefficient, so that the risk parameters can be dynamically adjusted as the operation data changes, thereby improving the system's ability to perceive abnormal road conditions and potential safety risks.

[0036] In step S7, the present invention periodically updates the road network structure and road segment parameters and performs version management, enabling the system to continuously optimize the road network model as operational data accumulates, realizing a closed-loop scheduling mechanism of "operation-learning-update-reoperation", and improving the long-term adaptability and stability of the mining area vehicle scheduling system.

[0037] The method of this invention does not depend on a specific mine type or vehicle platform, and can be widely applied to similar application scenarios that require vehicle scheduling and route planning, such as port transportation, slag transportation, and construction machinery. It has strong versatility and promotional value. Attached Figure Description

[0038] Figure 1 This is a schematic diagram of the overall system architecture of the present invention.

[0039] Figure 2 This is a flowchart of the automatic load status identification algorithm.

[0040] Figure 3 This is a flowchart of the algorithm for identifying slopes without elevation.

[0041] Figure 4 This is a flowchart of the dynamic weight calculation and risk self-learning algorithm.

[0042] Figure 5 A flowchart for online self-learning and version updates.

[0043] Figure 6 This is an example of road network node table data.

[0044] Figure 7 This is an example of road network edge table data.

[0045] Figure 8 A schematic diagram of the overall road network overlaid on the terrain of the mining area.

[0046] Figure 9 This is a detailed topological map of a local road network in the mining area. Detailed Implementation

[0047] The present invention will now be described in further detail with reference to the accompanying drawings.

[0048] It should be noted that the flowchart in the attached figure is a functional flow diagram of the method of the present invention. For ease of understanding, some weight factors are represented in a modular form. The calculation of the dynamic weight of the actual road segment shall be based on the mathematical model and formula given in the specification. The combination relationship between the factors can be equivalently adjusted according to the implementation method.

[0049] like Figure 1 As shown, the mine road network self-learning dynamic construction and update system based on vehicle positioning according to the present invention includes the following parts: The module includes a data acquisition module, a load status identification module, a slope identification module without elevation, a dual-map construction module for unloaded and fully loaded conditions, a dynamic weight calculation module, a risk coefficient self-learning module, and an online self-learning update module, among which: The data acquisition module is used to acquire vehicle trajectory data and provide the trajectory data to the load status recognition module and the elevation-free slope recognition module; The load status identification module is used to identify the unloaded and fully loaded status of vehicle operation data, and its identification results serve as the input to the unloaded and fully loaded bipartite graph construction module. The elevation-free slope identification module is used to identify the slope trend of road sections based on the difference between empty and full-load operating speeds, and provides the generated slope parameters to the empty and full-load bipartite map construction module; The empty and full load bipartite map construction module is used to construct empty and full load road network maps on the same road network skeleton, and provide road segment attribute parameters to the dynamic weight calculation module. The dynamic weight calculation module is used to calculate the dynamic weights required for path planning based on road segment attribute parameters; The risk coefficient self-learning module is used to update the road segment risk parameters based on vehicle operation behavior, and the update results are used in dynamic weight calculation. The online self-learning update module is used to periodically update road network parameters, enabling continuous self-learning optimization of the system.

[0050] In this embodiment, a transport vehicle from an open-pit mine is used as the application object. All transport vehicles are equipped with onboard positioning devices to continuously collect vehicle trajectory data during operation. The onboard positioning device can be implemented using GPS, BeiDou, or other satellite positioning systems; this invention does not limit its use.

[0051] In this embodiment, based on the aforementioned system structure, each functional module operates collaboratively in the order of steps S1 to S7 to realize a self-learning dynamic construction and update method for mine road networks based on vehicle positioning.

[0052] S1: Data Acquisition and Preprocessing In step S1, trajectory data formed by the continuous operation of the mining transport vehicle is obtained from the vehicle-mounted positioning device. The trajectory data includes at least fields such as timestamp, longitude, latitude, instantaneous speed and direction angle.

[0053] The acquired raw trajectory data is preprocessed, and the preprocessing includes: Abnormal trajectory points with obvious drift or positioning errors are removed, and the trajectory data is smoothed by Kalman filtering or moving average. Speed ​​data is normalized according to vehicle type and operating conditions; By combining vehicle turning behavior, parking behavior, and loading and unloading positions, continuous trajectory points are divided into road segments, thereby dividing the vehicle's running trajectory into several independent road segments.

[0054] After the above processing, structured trajectory data for subsequent analysis is obtained.

[0055] S2: Automatic load status identification: In step S2, statistical analysis is performed on the speed sequence of each vehicle to calculate the speed probability density distribution. The analysis reveals that the vehicle speed distribution typically exhibits a bimodal characteristic, with one peak corresponding to the vehicle's fully loaded operation and the other peak corresponding to its unloaded operation.

[0056] A Gaussian mixture model is used to fit the speed distribution to obtain a corresponding bimodal distribution model, and the intersection of the two distribution curves is used as the speed threshold. Based on the speed threshold, operating samples with speeds below the threshold are determined to be in a fully loaded state, and operating samples with speeds above the threshold are determined to be in an unloaded state, thereby realizing the automatic identification and labeling output of vehicle load status.

[0057] The system can update the speed threshold using a sliding window or periodic recalculation to adapt to the dynamic changes in the mining area's operating environment.

[0058] S3: Slope identification and slope coefficient generation without elevation: In step S3, without relying on road elevation data, the road slope trend is determined by the difference in vehicle speed between unloaded and fully loaded states.

[0059] Using each road segment as an analysis unit, the average operating speed of vehicles in the empty and fully loaded phases of that segment is statistically analyzed, and the speed difference between the two is calculated. Based on the sign and magnitude of the speed difference, the slope trend of the road segment is determined: when the average speed under full load is significantly lower than the average speed under empty load, it is determined to be an uphill segment; when the average speed under full load is higher than the average speed under empty load, it is determined to be a downhill segment; when the two are close, it is determined to be a level slope segment.

[0060] Based on the slope trend determination results, a corresponding slope penalty coefficient is generated to reflect the impact of slope on vehicle operating efficiency and safety.

[0061] S4: Construction of a dual-mode network with both unloaded and fully loaded traffic: In step S4, a road network skeleton structure for the mining area is constructed based on vehicle trajectory points. The road network nodes include loading points, unloading points, and road intersections, and the road network edges correspond to the trajectory segments formed by vehicle travel.

[0062] On the same road network framework, construct both unloaded and fully loaded road network maps, and record the average operating speed, gradient penalty coefficient, risk coefficient, and road segment weight-related parameters for subsequent dynamic weight calculation for each road segment under unloaded and fully loaded conditions.

[0063] Among them, the unloaded road network map focuses on reflecting the operating efficiency of vehicles when they are unloaded, while the fully loaded road network map focuses on reflecting the operating safety and stability of vehicles when they are fully loaded.

[0064] S5: Dynamic weight calculation: In step S5, a comprehensive dynamic weight is calculated for each road segment during the route planning process. The dynamic weight comprehensively considers the segment distance, slope penalty coefficient, road curvature parameter, risk coefficient, and historical average speed factor.

[0065] Based on the current load status of the vehicles, a choice is made between an unloaded road network map and a fully loaded road network map, and the priority of the weight factors of each road segment is adjusted to achieve a dynamic balance between efficiency and safety at different operating stages.

[0066] Based on the dynamic weights, the minimum weight path search algorithm is used to plan the vehicle's driving path.

[0067] S6: Risk coefficient self-learning and updating: In step S6, road segment risks are assessed by statistically analyzing the behavioral characteristic data generated by vehicles during their operation on each road segment. This behavioral characteristic data includes the frequency of sudden braking, the standard deviation of speed fluctuations, the number of abnormal stops, and risk correction information related to weather or road conditions.

[0068] Based on the above statistical results, the observed risk values ​​of the road segments are generated, and the risk coefficients of the road segments are updated using an exponential smoothing method, so that the risk coefficients can reflect the changes in safety risks in the latest operating cycle while retaining historical risk information.

[0069] The updated risk coefficient is written to the database for dynamic weight calculation and path planning in subsequent cycles.

[0070] S7: Online self-learning and version updates: In step S7, the system summarizes and analyzes vehicle operation data on a shift cycle or daily cycle basis, and updates speed thresholds, gradient penalty coefficients, risk coefficients, and road segment weight parameters to generate a new road network version.

[0071] During subsequent operation, the system prioritizes the latest road network version for route planning and retains historical road network versions for backtracking analysis. When a significant change in road conditions is detected, the system can trigger a local update for the relevant road segment.

[0072] Through the aforementioned online self-learning and version update mechanism, the road network model is continuously optimized as operational data accumulates, achieving a closed-loop optimization process of "operation—learning—update—re-operation".

[0073] Feasibility analysis of typical mining area scenarios: Using a typical working condition of an open-pit mine as a background, a feasibility analysis scenario for the method is constructed. Within the mining area, transport vehicles are equipped with onboard positioning terminals to continuously collect trajectory data during vehicle operation. The trajectory data includes timestamps, latitude and longitude, speed, and direction angle. Based on the collected vehicle operation data, the system operates according to the processing flow of steps S1 to S7 described above to verify the adaptability and collaborative working capability of each module under the typical mining area's operational data format.

[0074] Combining the processing flow of steps S1 to S7 described above, and referring to... Figures 2 to 9The illustrated structure and process demonstrate that the method of this invention exhibits good adaptability and engineering feasibility under typical mining area operational data conditions. This method can effectively distinguish the load characteristics of vehicles under different operating conditions and identify road slope trends without relying on road elevation information. Through dynamic updates of road segment weights and risk parameters, it can optimize vehicle transportation route selection to a certain extent, reduce potential operational risks, and improve overall transportation efficiency.

[0075] like Figure 2 As shown, the system takes the vehicle speed sequence as input, constructs a vehicle speed probability distribution model, identifies the bimodal characteristics of the speed distribution, fits the speed distribution using a Gaussian mixture model, calculates the intersection of the two distribution curves as the speed threshold, and determines whether the vehicle is in an empty or fully loaded state based on the speed threshold. At the same time, it combines the location information of the loading and unloading areas to smooth and correct the load determination result, and finally outputs the vehicle load status identification result.

[0076] like Figure 3 As shown, the system takes vehicle trajectory data as input, calculates the average running speed of vehicles in the same road segment under empty and fully loaded conditions, calculates the speed difference between the two, and determines the road segment slope trend based on the relationship between the speed difference and a preset threshold. When the speed difference is greater than the threshold, it is determined to be an uphill trend; when the speed difference is less than the negative threshold, it is determined to be a downhill trend; when the absolute value of the speed difference is less than the threshold, it is determined to be a flat slope trend. Then, the corresponding slope penalty coefficient is generated according to the slope trend, and the slope parameters are output after smoothing.

[0077] like Figure 4 As shown, the system selects the corresponding empty load weight model or full load weight model according to the current load status of the vehicle, and calculates the comprehensive dynamic weight value of the road segment by taking into account various factors such as road segment distance, slope penalty coefficient, road curvature parameter, risk coefficient and historical speed factor, which is used for path planning and right-of-way update.

[0078] like Figure 5 As shown, the system collects behavioral data during vehicle operation, including the number of emergency brakings, speed fluctuations, and abnormal parking behaviors, to calculate the observed risk value of the road segment. It then uses an exponential smoothing method to update the historical risk coefficients. The updated risk coefficients are stored and used for subsequent dynamic weight calculations and path planning.

[0079] like Figure 6 As shown, the road network node data is constructed based on vehicle trajectory points. Each node corresponds to a set of vehicle positioning latitude and longitude coordinates, which are used to describe the spatial location attributes of the road nodes in the mining area. The node data includes node number, longitude, latitude and node attribute information, providing basic spatial data support for the construction of the road network framework in the mining area.

[0080] like Figure 7 As shown, the road network edge data is used to describe the connection relationship between nodes. Each road segment is defined by a start node and an end node, and the corresponding road segment attribute information is recorded. The road segment attribute information includes road segment number, start node number, end node number, and road segment type, etc., which are used to characterize the topological structure relationship of roads in the mining area and provide a data foundation for subsequent right-of-way calculation and path planning.

[0081] like Figure 8 As shown, under the complex terrain conditions of open-pit mines, the roads exhibit a multi-tiered, stepped structure, are winding and have numerous branches, with vehicle trajectories covering loading and unloading areas and multiple transport channels. The method of this invention continuously collects and analyzes vehicle trajectories based on vehicle positioning data. Without relying on high-precision maps or terrain elevation data, it can automatically construct a road network model reflecting the actual road structure of the mine area, providing basic road network support for subsequent route planning and scheduling.

[0082] like Figure 9 As shown, in local areas with dense road intersections and frequent curves, vehicle trajectories exhibit significant directional changes and traffic diversion characteristics. The method of this invention, through comprehensive analysis of vehicle load status, operating speed, and trajectory characteristics, can accurately characterize the traffic relationships and structural features of local road segments. Based on this, it enables the differentiated construction of empty and fully loaded road networks and the dynamic updating of road segment weights, demonstrating the engineering adaptability of this method in complex mining area road environments.

[0083] In summary, the method of this invention has good theoretical feasibility and engineering adaptability in typical mining area operating environments, and can provide technical support at the algorithm and data structure levels for the subsequent deployment of mining area transportation scheduling systems.

Claims

1. A self-learning dynamic construction and update method for mine road networks based on vehicle positioning, characterized in that, Includes the following steps; S1: Collect vehicle trajectory data and preprocess the acquired data to obtain structured data; S2: Based on the structured trajectory data, generate a speed sequence for each mining transport vehicle within a preset time window. Statistical analysis is performed to achieve automatic identification of vehicle load status; S3: Based on the vehicle load status recognition results, the slope trend of the mining area road is identified by utilizing the difference in the vehicle's operating speed under empty and fully loaded conditions, and the corresponding slope penalty coefficient is generated. S4: Based on the structured trajectory data, vehicle load status, and slope penalty coefficient, construct the basic framework of the mining vehicle road network, and establish an empty road network map and a fully loaded road network map on the road network framework respectively. S5: Calculate dynamic weights based on the unloaded and fully loaded road network maps; S6: Update the risk-related parameters in the dynamic weights; S7: Based on the dynamically updated weights, during system operation, the system summarizes and analyzes the operation data of mining vehicles according to the preset update cycle, performs online self-learning updates on the road network structure and road segment parameters in the mining area, forms a versioned management and continuous optimization mechanism for the road network, and calls the latest road network version in subsequent route planning.

2. The method for self-learning dynamic construction and updating of mine road networks based on vehicle positioning according to claim 1, characterized in that, Specifically, S1 is: Continuous operating trajectory data is obtained from the on-board positioning device of the mining transport vehicle. The operating trajectory data includes the following fields: Timestamp t, longitude Lon, latitude Lat, instantaneous velocity v, direction angle θ; Preprocess the trajectory data; The preprocessing steps are as follows: (1) Identify and remove abnormal trajectory points in the running trajectory data. The abnormal trajectory points include abnormal points caused by positioning drift or positioning error. After removing the abnormal trajectory points, smooth the remaining trajectory data using Kalman filtering or moving average method. (2) Based on the completion of trajectory denoising and smoothing, the smoothed speed data is normalized according to the vehicle type and corresponding operating conditions. (3) Based on the speed normalization process, the continuous running trajectory of the vehicle is segmented according to the changes in the vehicle's driving direction, parking behavior, and loading or unloading position, and the running trajectory is divided into multiple independent road segments. Through the preprocessing steps executed sequentially above, the final output is structured data: 。 in, and They represent the first The longitude and latitude coordinates corresponding to each trajectory point This represents the normalized velocity value of the corresponding trajectory point. This represents the timestamp of the corresponding trajectory point. Indicates the index of the trajectory point. This represents the total number of trajectory points.

3. The method for self-learning dynamic construction and updating of mine road networks based on vehicle positioning according to claim 2, characterized in that, S2 specifically includes the following steps: (1) For the velocity sequence Perform statistical processing to calculate the probability density function of vehicle speed. To characterize the speed distribution features of vehicles under different operating conditions; among them, This represents the normalized speed value of the vehicle at the trajectory point; Show corresponding speed value The probability density of occurrence; (2) Based on the probability density function A Gaussian mixture model is used to fit the vehicle speed distribution, which is expressed as a weighted mixture model of two Gaussian distributions. One of its expressions is: in, and Let represent the mixture weights of two Gaussian distributions, and satisfy . =1, This represents the first Gaussian distribution with a mean of 1 / 2. The variance is , This represents the second Gaussian distribution with a mean of 1 / 2. variance is Among them, the velocity distribution with a smaller mean The speed distribution with a relatively large mean corresponds to the fully loaded operating state of the vehicle. The corresponding vehicle's unloaded operating status; (3) By solving for the intersection of the two Gaussian distributions, the speed threshold used to distinguish the vehicle load state is determined. The speed threshold satisfies the following relationship: The speed threshold This indicates the dividing speed at which a vehicle transitions from a fully loaded state to an unloaded state. (4) Based on the speed threshold For each trajectory point, the corresponding velocity value The load condition is determined according to the following rules: in: =1 indicates that the vehicle is at the... Each trajectory point is in a fully loaded state; =0 indicates that the vehicle is in the... Each trajectory point is in an unloaded state; Indicates the first The normalized velocity value corresponding to each trajectory point.

4. The method for self-learning dynamic construction and updating of mine road networks based on vehicle positioning according to claim 3, characterized in that, S3 specifically includes the following steps: (1) Taking each road segment obtained in step S1 as an analysis unit, the average running speed of vehicles in both unloaded and fully loaded states is calculated as follows: in: This represents the average normalized operating speed of a vehicle when it is unloaded on the corresponding road segment. This represents the average normalized operating speed of a vehicle when it is fully loaded on the corresponding road segment. Indicates the first The normalized velocity values ​​corresponding to each trajectory point; This indicates the first value obtained in step S2. The load state markers corresponding to each trajectory point, among which Indicates no-load status. Indicates a fully loaded state; This indicates the number of empty-load status trajectory points within the corresponding road segment; This indicates the number of fully loaded trajectory points within the corresponding road segment; (2) Based on the average speed under no-load and average speed under full-load conditions, where the average operating speed is the mean, median, or other statistical measure that can reflect the vehicle's operating level, calculate the speed difference for the corresponding road segment, the expression of which is: in, It represents the average speed difference of vehicles under empty and fully loaded conditions on the same road segment, and is used to characterize the degree of influence of the road segment on the operation of vehicles under different load conditions; (3) Based on the aforementioned average speed difference The numerical value and sign, combined with the preset speed difference threshold To determine the slope trend of the corresponding road segment, the following rules apply: in, The speed difference threshold is used to suppress false judgments caused by random fluctuations or noise. (4) Based on the slope trend obtained from the determination, the speed difference is mapped to a slope penalty coefficient for path weight calculation. The calculation method is as follows: in: This represents the gradient penalty coefficient for the corresponding road segment, used to characterize the impact of gradient on vehicle operating costs; This is an empirical coefficient used to adjust the degree of influence of uphill and downhill on path weights.

5. The method for self-learning dynamic construction and updating of mine road networks based on vehicle positioning according to claim 4, characterized in that, S4 specifically includes the following steps: (1) Based on the trajectory points formed by vehicles running in the mining area, the roads in the mining area are abstractly modeled to generate a road network skeleton structure, which is represented as a graph structure. ,in: a. This represents a set of road network nodes, which include vehicle loading points, unloading points, and road intersections. b. This represents the set of road network edges, which consists of continuous driving trajectory segments formed by vehicles between adjacent nodes. (2) In the road network framework Based on the vehicle load state markings obtained in step S2 Construct unloaded road network maps respectively And a full-load road network map ,in: a. Unloaded road network map Used to describe the road operating characteristics of vehicles when they are unloaded, with an emphasis on time efficiency; b. Fully loaded road network map Used to describe the road operating characteristics of vehicles when fully loaded, with an emphasis on safety and stability; c. and These represent the sets of road segment edges corresponding to the unloaded and fully loaded states, respectively. (3) Regarding the aforementioned unloaded road network map And a full-load road network map Each road segment in The average operating speed and historical risk parameters under the corresponding load conditions are statistically analyzed, and the calculation method is as follows: in: a. Indicates road segment Under load conditions The average normalized running speed under these conditions; b. Indicates no-load status. Indicates a fully loaded state; c. Indicates the first one that falls on the road segment Normalized velocity corresponding to each trajectory point; d. Indicates road segment Under load conditions The number of corresponding trajectory points. (4) In the above unloaded road network map And a full-load road network map In the middle, the following attribute parameters are recorded for each road segment edge: a． ； b. Slope penalty coefficient generated in step S3 ; c. Risk parameters obtained from historical operational data statistics ; d. Road segment weight-related parameters used for dynamic weight calculation in subsequent step S5, including road segment distance parameters, slope penalty coefficient, road curvature parameters, risk coefficient, speed factor parameters, and other relevant parameters used for dynamic weight calculation.

6. The method for self-learning dynamic construction and updating of mine road networks based on vehicle positioning according to claim 5, characterized in that, Specifically, S5 is: For the unloaded road network map constructed in step S4 With full load of road network map Each road segment in The dynamic weight of the road segment is calculated by considering factors such as distance, gradient penalty, curvature, risk, and speed. ,in These represent the empty and fully loaded states, respectively. The dynamic weights are used for segment cost evaluation in subsequent path planning. One of the dynamic weight models is as follows: (1) Dynamic weight model The factors are defined as follows: ; in, Indicates road segment The first one arranged in chronological order The coordinates of the trajectory points, dist( () represents the planar distance between two points on the trajectory. This indicates the number of trajectory points corresponding to the road segment, and the distance of the road segment. It can also be obtained through other equivalent spatial distance calculation methods; ； in, The average speed difference between empty and fully loaded sections calculated in step S3, ( ) represents the slope penalty mapping function defined in step S3 (i.e., the slope penalty coefficient generated from the uphill / flat / downhill trend); curvature factor The curvature penalty is calculated based on the steering angle of adjacent trajectory segments within a road segment. The curvature factor is defined as follows: in, This is the curvature weighting coefficient. For road section The average curvature index; the average curvature index can be obtained by averaging the absolute values ​​of the turning angles of continuous trajectory segments within the road segment; ； ； in, The risk coefficient for road sections is updated based on historical operational data; ； in, The road segments obtained in step S4 Under load conditions The average normalized running speed is as follows. As the reference velocity constant, To avoid positive constants with a denominator of zero. (2) Different factor index weights are applied to the unloaded road network map and the fully loaded road network map, and one of the dynamic weights is further expressed as: in: a. ; indicates the dynamic weight of the road segment under no-load conditions, which is taken under no-load conditions. Larger values ​​are used to increase the influence weight of the speed factor; b. This represents the dynamic weight of a road segment under full load conditions. Under full load conditions, the weight is... and Larger values ​​are used to increase the influence weight of slope factor and risk factor; c. , , , , , , , These are the factor weight index parameters, either preset or adjusted based on operational data. Based on the dynamic weight The minimum weight path search is performed in the corresponding empty or full-load road network map to obtain a driving path that meets the requirements of efficiency and safety.

7. The method for self-learning dynamic construction and updating of mine road networks based on vehicle positioning according to claim 1, characterized in that, Specifically, S6 is: Based on the structured trajectory data obtained in step S1 and the road segmentation results in step S4, the empty road network map is analyzed. Each road segment in The system statistically analyzes the vehicle's operational behavior characteristics on this road segment, generates a road segment observation risk value, and updates the road segment risk coefficient based on an exponential smoothing rule. The updated risk coefficient is used for the dynamic weight calculation in step S5, specifically: (1) For each road segment The vehicle operation behavior within a preset statistical period is statistically analyzed to obtain the observed risk value. The calculation method is as follows: in: Indicates the load condition. Unloaded 1 indicates full load; Indicates road segment Under load conditions The observed risk value; Indicates the number of vehicles on the road segment within the statistical period. Upper and load conditions are The number of times emergency braking occurred; Indicates the number of vehicles on the road segment within the statistical period. Upper and load conditions are The standard deviation of the velocity over time; Indicates the number of vehicles on the road segment within the statistical period. Upper and load conditions are The number of abnormal parking incidents; Together with the risk factors in step S5, it represents the risk correction factor related to weather or road conditions; , , , These are preset weighting coefficients used to adjust the degree of influence of each observation indicator on the observation risk value; (2) For each road segment Establish risk coefficient And based on observed risk values Update using exponential smoothing rules: in: Indicates the period of statistical analysis After the end, the section Under load conditions The risk factor for updates is as follows; This represents the risk coefficient for the previous statistical period; This represents the observed risk value obtained in the current statistical period; Let be the learning rate parameter, satisfying 0 < 0. <1, used to control the fusion ratio of historical risk and current observed risk. Updated Stored in the road segment attribute parameters and used as a risk factor in step S5. The calculation input enables self-learning iterative updates of road segment risk parameters; (3) Update the results and write them into the database for use in the next cycle weight calculation.

8. The method for self-learning dynamic construction and updating of mine road networks based on vehicle positioning according to claim 1, characterized in that, S7 specifically includes the following steps: (1) Within a preset time period, the structured trajectory data generated by vehicles operating within the mining area is summarized. The time period can be a shift cycle, a daily cycle, or other preset cycles. The summarized data serves as the input data for the road network update in this cycle. (2) Based on the aggregated runtime data, perform the following update operations in sequence: Summarize the data for the day and recalculate the speed threshold, slope coefficient, and risk weight; Based on step S3, update the slope trend determination results and corresponding slope penalty coefficients for each road segment; According to step S4, update the statistical parameters such as the average operating speed of the road segment under empty and full load conditions; According to step S6, update the road segment risk coefficient to reflect the changes in safety risks in the most recent operating cycle; The above updates enable road segment parameters to be continuously corrected as vehicle operation data accumulates.

Images