Well mining unmanned cloud control platform global path planning method based on V2X

Abstract

The invention discloses a V2X-based global path planning method for a mine unmanned driving cloud control platform, and relates to unmanned driving. The V2X-based global path planning method comprises the following steps: constructing a traffic semantic map data structure # imgabs0 #; according to task issuing or operation plan adjustment, generating path request data R, and performing time constraint, resource constraint and path optimality constraint verification on the path request data R; according to the R and # imgabs1 #, adopting a heuristic search algorithm based on a graph theory to carry out optimal path search on the road topological structure, and generating a global path P containing a node sequence and driving parameters; and acquiring obstacle data detected by the vehicle end through a local sensor, performing obstacle avoidance correction through a local path optimization algorithm according to the obstacle data and the global path P, and generating a local path meeting the safety distance constraint and the path deviation constraint. According to the method, on the premise of meeting multi-dimensional coupling constraints such as time-space, priority-resource, safety-efficiency and the like, the optimal driving path dynamically adapting to the complex environment of the well industry and mining is generated.

Description

Technical Field

[0001] The present application relates to the field of unmanned driving, and in particular to a global path planning method for an unmanned driving cloud control platform in underground mines based on V2X. Background Art

[0002] With the rapid advancement of intelligent mine construction, unmanned transportation systems in underground mines, as a key component of intelligent mining, are crucial for improving mining efficiency, ensuring operational safety, and reducing labor costs. However, the characteristics of underground mining environments, such as narrow tunnels, complex road conditions, limited communications, and dynamically changing operational tasks, place extremely high demands on unmanned path planning technology. Especially in scenarios involving multi-vehicle collaborative operations, path planning must simultaneously meet multiple constraints, including time constraints, spatial restrictions, task priorities, and safety requirements. These constraints are complexly coupled, posing significant challenges to path optimization decisions.

[0003] The existing unmanned transportation system for underground mines has the following technical defects in path planning:

[0004] First, the ability to handle multi-dimensional constraint coupling is insufficient. Existing systems often rely on static preset paths or simple planning algorithms based on local perception, making them difficult to handle complex coupling issues such as time-space constraint conflicts, priority-resource constraint conflicts, and safety-efficiency constraint balances. For example, when the shortest path is blocked by a landslide, the system cannot quickly find an alternative path while meeting time constraints. When multiple high-priority tasks simultaneously request limited path resources, there is a lack of effective coordination and allocation mechanisms.

[0005] Second, there's a lack of a global-local collaborative optimization mechanism. Traditional systems either rely too heavily on centralized cloud-based planning while ignoring the need for real-time obstacle avoidance on the vehicle itself, or rely solely on local vehicle perception and lack a global optimization perspective. This single-level planning approach fails to effectively balance global optimality with local safety requirements, resulting in poor path execution in complex and dynamic environments.

[0006] Third, multi-source data fusion and environmental modeling capabilities are weak. Although some systems have introduced V2X communication technology, they fail to fully utilize the multi-source heterogeneous data collected by devices such as OBUs and RSUs for deep fusion. This leads to insufficient understanding of traffic semantics and difficulty in building accurate, real-time, and unified environmental models. This directly affects the accuracy and reliability of path planning decisions.

[0007] Fourth, dynamic response and real-time replanning capabilities are limited. Underground environments change frequently, including temporary tunnel closures, equipment failures, and urgent task insertions. Existing systems lack a rapid response mechanism and are unable to replan and dynamically adjust routes based on real-time status changes, which can easily lead to decreased transportation efficiency and even safety accidents.

[0008] Fifth, task scheduling and resource allocation strategies are simplistic. Existing technologies fail to fully consider the priorities and resource competition among different tasks. They lack a comprehensive scheduling mechanism based on task attributes, vehicle status, and road network capacity. This results in critical tasks not being prioritized and overall transportation efficiency being low.

[0009] In summary, the existing unmanned driving path planning technology in underground mines has obvious shortcomings in dealing with multi-dimensional constraint coupling, achieving global-local collaborative optimization, and integrating multi-source data to build environmental models. It is difficult to meet the requirements of complex underground environments for efficient, safe, and intelligent transportation. There is an urgent need for a new path planning method that can comprehensively consider multi-dimensional constraint relationships and achieve cloud-side-vehicle collaborative optimization. Summary of the Invention

[0010] In response to the decision conflicts caused by the multi-dimensional constraint coupling in unmanned driving path planning in underground mining environments, this application provides a global path planning method for an underground mining unmanned driving cloud control platform based on V2X. Through hierarchical path optimization driven by multi-source data fusion, global planning on the cloud side and local adjustment on the vehicle side, an optimal driving path that dynamically adapts to the complex environment of underground mining is generated under the premise of satisfying multi-dimensional coupling constraints such as time-space, priority-resources, safety-efficiency, etc.

[0011] The present application provides a global path planning method for an unmanned driving cloud control platform for underground mines based on V2X, including: S1, collecting multi-source data, which includes: vehicle status data collected by OBU equipment, road traffic data collected by RSU equipment, traffic signal status data and underground mine map data; S2, fusing the multi-source data through a Bayesian fusion algorithm to construct a traffic semantic graph data structure. ; S3, according to the task issued or the work plan adjustment, generate the path request data R, and perform time constraint, resource constraint and path optimality constraint verification on the path request data R; wherein the path request data R includes: starting position , end position , start time , latest completion time , task priority Q, access restriction L, safety policy parameter S and path preference E; S4, based on the verified path request data R, and traffic semantic graph data structure , a heuristic search algorithm based on graph theory is used to search for the optimal path of the road topology structure, and a global path P containing a node sequence and driving parameters is generated; S5, obstacle data detected by the vehicle side through local sensors is obtained, and obstacle avoidance correction is performed through a local path optimization algorithm based on the obstacle data and the global path P, to generate a local path that meets the safety distance constraint and path deviation constraint; the execution status of the local path is obtained and fed back to the cloud control platform for dynamic path optimization.

[0012] Furthermore, S2 uses the Bayesian fusion algorithm to fuse the multi-source data and construct the traffic semantic graph data structure. , including: fusing multi-source data through Bayesian fusion algorithm to obtain the fusion estimation value of each traffic element , including: time synchronization and spatial synchronization of observation data from different V2X devices; obtaining observation values ​​from n data sources for each traffic element and the corresponding variance ; According to the variance of each data source , calculate the fusion weight , where the fusion weight Reflects the confidence level of each data source;

[0013] According to the fusion weight Calculate the fused estimate , ,in, Represents the fused estimate of a traffic element. The fused estimate for each traffic element is the optimal estimate obtained by weightedly fusing observation data from multiple V2X devices (such as OBUs and RSUs) for the same traffic element using a Bayesian fusion algorithm. In underground mining environments, single sensors often have blind spots or measurement errors. By integrating information from multiple sources, the fused estimate provides a more accurate and reliable description of the environmental state than any single data source.

[0014] The extended Kalman filter algorithm is used to track the multi-source data after fusion processing to obtain the state prediction information of the dynamic target, including: initializing the state vector of the dynamic target and the error covariance matrix ; Based on the state transfer matrix F and process noise covariance Q, through the state prediction equation Calculate the prior state estimate at the current moment and use the error covariance prediction equation , calculate the prior error covariance; where B represents the input control matrix, Represents the input control vector; obtains the observation data at the current moment , calculate the Kalman gain based on the observation matrix H , where R is the observation noise covariance; through the state update equation Calculate the posterior state estimate and update the equation through the error covariance Calculate the posterior error covariance;

[0015] Output the state estimate of the dynamic target at the current moment State prediction information. Dynamic target state prediction refers to the prediction of the future state of moving objects (such as other vehicles, mobile equipment, and pedestrians) in underground mining environments, using the extended Kalman filter (EKF) algorithm based on historical observation data and motion models. In narrow underground tunnels, predicting the movement trends of other vehicles or obstacles is crucial for avoiding collisions, optimizing meeting strategies, and planning safe overtaking paths.

[0016] Construct a traffic semantic graph data structure based on the fusion estimation value of each traffic element, the state prediction information of the dynamic target, and the road topology structure extracted from the underground mining map data , where V represents the traffic element node and E represents the spatial relationship between elements; the traffic semantic graph is a structured environmental expression model that abstracts the underground mining road network and traffic elements into a graph structure, where the node V represents various traffic entities and the edge E represents the spatial association relationship between entities.

[0017] Traffic element nodes include lanes, vehicles, traffic lights, and obstacles; spatial relationships between elements include lane connections, lane changes, and following relationships; traffic elements include lane position, vehicle position, vehicle speed, traffic light status, and obstacle position;

[0018] In particular, in the complex environment of underground mines, a single sensor or data source often has blind spots, measurement errors, and local viewing angle limitations. Through the Bayesian fusion algorithm, the system uses the variance of each data source to Calculate fusion weights , weighted fusion of observation data from multiple V2X devices such as OBU and RSU to obtain the optimal estimate of each traffic element This confidence-based fusion mechanism ensures the accuracy and integrity of environmental information. Multi-source fusion eliminates the uncertainty of a single data source, enabling the system to accurately identify the current environmental state and avoid constraint misjudgment caused by information errors. The fused estimate Compared with a single data source, the reliability of constraint evaluation is significantly improved.

[0019] The Extended Kalman Filter (EKF) algorithm, through its state prediction and update mechanism, not only captures the current target state but, more importantly, predicts the state's evolutionary trends in the future. This predictive capability enables the system to proactively identify conflicts between temporal and spatial constraints (e.g., predicting congestion on a particular road section at a specific time); anticipate resource competition (e.g., multiple vehicles simultaneously requesting the same road section); and assess evolving safety risks (e.g., the potential intersection of an obstacle's trajectory with the planned path).

[0020] Constructed traffic semantic graph The complex underground mining environment is abstracted into a structured graph model: nodes V contain traffic elements such as lanes, vehicles, traffic lights, and obstacles, and each node carries semantic information such as position, speed, and status. Edges E express spatial relationships such as lane connections, lane changes, and following relationships, and implicitly contain traffic constraints and dependencies. This graph structure enables the modeling and solution of previously coupled multidimensional constraints within a unified framework: time constraints are mapped to the travel time attributes of paths in the graph; spatial constraints are reflected in the connectivity of the graph and the availability of edges; and safety constraints are manifested as minimum distance requirements between nodes.

[0021] The traffic semantic graph unifies heterogeneous constraints within a graph structure, transforming complex multi-objective optimization problems into graph search problems, significantly reducing solution complexity. Furthermore, the graph's dynamic update mechanism ensures that constraints reflect environmental changes in real time. The constructed traffic semantic graph not only incorporates static road topology but also incorporates dynamic traffic status and forecast information. This provides a comprehensive, accurate, and real-time environmental model for subsequent global path planning and local path optimization, ensuring optimal feasible solutions can be found even under multi-dimensional constraints.

[0022] Further, S3, based on the task issuance or job plan adjustment, generates path request data R, and verifies the time constraint, resource constraint and path optimality constraint of the path request data R, including: when receiving the task issuance instruction or job plan adjustment instruction, extracting the task parameters and generating the path request data R ; Perform time constraint verification to verify the estimated travel time in the path request data R Whether it meets: ; Perform resource constraint verification and verify the road segments in the path request data R Whether it meets: and , where path represents the starting point To the end position A set of feasible paths; perform path optimality constraint verification to verify whether: ,in, Represents the path cost of the i-th path; encapsulates the path request data R that passes all constraint verifications into a structured request set , and sort them according to task priority Q;

[0023] Time constraints refer to the time windows that route planning must meet to ensure that the vehicle can begin and complete the transport mission within the required timeframe. Time constraints are crucial in underground mining operations: The continuity of mining operations requires timely material delivery to avoid downtime; traffic is prohibited during the ventilation window after blasting operations; and the coordinated operation of multiple working faces requires precise timing. Violations of time constraints can lead to production interruptions, increased costs, and even safety incidents.

[0024] Resource constraints refer to the limitations on road resource availability that must be considered during route planning, ensuring that the planned route only uses currently accessible sections and avoids entering closed or restricted areas. Specific types of resource constraints include: physical constraints: restrictions on vehicle types due to lane width, height, and carrying capacity; dynamic constraints: temporary closures (for landslides, maintenance), and one-way traffic periods; capacity constraints: the upper limit on the number of vehicles a section can accommodate at a time; and access constraints: access rights to specific sections (such as dedicated lanes for hazardous materials). Lane resources are limited, and incorrect resource allocation can lead to widespread congestion; some lanes are dynamically opened / closed due to geological conditions, maintenance operations, and other reasons; different types of vehicles (personnel transport, material transport, and special vehicles) have different requirements for road resources; and violating resource constraints can lead to vehicle entrapment, collisions, or affect other operations.

[0025] Path optimality constraints mean that, while satisfying other hard constraints, path planning should minimize overall costs and achieve an optimal balance among multiple objectives, including efficiency, safety, and energy consumption. In underground mining environments, path optimality constraints balance the demands of different tasks, ensuring that critical production tasks are completed first. Efficiency is pursued while ensuring safety, reducing operating costs. Multi-vehicle collaboration is considered to avoid local optima that could lead to global efficiency degradation. Prioritization mechanisms are employed to flexibly address resource competition between emergency and routine tasks.

[0026] Furthermore, the starting point Indicates the current vehicle position coordinates, which are derived from the vehicle-mounted GNSS+IMU integrated navigation or the underground UWB positioning system; the end position Indicates the coordinates of the mission destination, including loading point, unloading point or maintenance station; start time Indicates the scheduled start time of the task, used to determine whether the vehicle can start in time; the latest completion time Indicates that the destination must be reached before this time, which is used for task timeout judgment and path optimization; task priority Q is a numerical priority identifier, which is used for multi-task path competition or scheduling sorting; access restriction L represents the set of road sections currently allowed to pass, which is restricted by landslides, maintenance or traffic flow; safety strategy parameter S represents the safety constraints of obstacle avoidance level, minimum vehicle distance, maximum downhill slope or braking response time; path preference E represents the preferred route type, including main lane priority, avoidance of slippery areas or avoidance of manual work areas.

[0027] In particular, in traditional path planning, time, space, safety, efficiency and other constraints are often intertwined and difficult to distinguish and handle. This solution decomposes the complex constraints into 8 independent parameters: Time dimension: start time Ts, latest completion time ; Spatial dimension: starting position , end position , access restriction L; priority dimension: task priority Q; safety dimension: safety policy parameter S; efficiency dimension: path preference E; This parameterized expression transforms the originally coupled constraints into independent data fields , giving each constraint dimension a clear numerical representation, creating the conditions for hierarchical processing. By decomposing coupled multidimensional constraints into independent parameters, the originally complex multi-objective optimization problem is transformed into a parameterized constraint satisfaction problem. Each constraint condition has clear judgment criteria and thresholds, making automated verification possible.

[0028] In addition, this application adopts a three-tier progressive verification architecture:

[0029] First layer: time constraint verification ,The feasibility of the time window is verified before path search,,avoiding tasks that are impossible to complete in time from entering,subsequent calculations and reducing invalid calculations.

[0030] Second layer: resource constraint verification and , checks spatial access restrictions to ensure that the planned path does not pass through closed or restricted areas. This layer separates time constraints from spatial constraints, avoiding the coupling problem of the shortest path being blocked and causing time limit violations.

[0031] The third layer: path optimality verification ,On the basis of satisfying hard constraints, the priority-weighted comprehensive cost is minimized to achieve efficiency optimization.

[0032] This hierarchical verification breaks down complex multi-objective optimization problems into ordered, single-objective decisions, significantly reducing solution complexity. This hierarchical verification mechanism filters out requests that fail to meet hard constraints at an early stage, preventing the subsequent path planning algorithm from searching in an infeasible solution space. Completing constraint verification before path planning execution promptly identifies potential constraint conflicts (such as narrow time windows or closed road sections), allowing time for task rescheduling or parameter adjustments, and avoiding reactive responses during execution.

[0033] Further, S4, generates a global path P including node sequence and driving parameters, including: Extract road topology data and use lane nodes in traffic element node V as vertex set for path planning , the lane connection relationship in the spatial relationship E between elements is used as the edge set , build a path search graph ,in, Represents edge weight; based on the starting point location in the path request data R and end position , in the path search graph Determine the corresponding starting point node and target node ; According to the access restriction L, from the edge set Eliminate the inaccessible edges and get the set of available edges ; Establish heuristic cost function ,in, Represents the cumulative path cost from the starting node vs to the current node n, Indicates the transition from the current node n to the target node The heuristic estimation cost is , Q represents the task priority; the A* algorithm is used to search the global path P.

[0034] in: , where e represents the edge connecting two nodes; represents the road segment distance corresponding to edge e, Represents the fusion estimation value in the traffic semantic graph data structure The estimated travel time is calculated based on the vehicle speed in Indicates the risk value of the road section; is the weight coefficient;

[0035] In particular, traditional path planning is difficult to handle heterogeneous constraints such as distance, time, and safety at the same time. , converting multidimensional constraints into comparable numerical values: spatial dimension: Indicates the physical distance of the road section and reflects the occupation of spatial resources; Time dimension: Calculate travel time based on real-time traffic status to reflect time constraints; Safety dimension: Quantify the risk value of the road section, including safety factors such as slope, curve, accident probability, etc.; through the weight coefficient The system can flexibly balance the importance of different constraint dimensions. For example, urgent tasks can increase reduce , giving priority to time efficiency; the transportation of dangerous goods increases , giving priority to ensuring safety. Specifically, daily transportation: balanced consideration of various indicators ; Emergency rescue: time priority ; Dangerous goods transportation: safety first .

[0036] In addition, this application reduces the search space through a two-layer mechanism to improve the efficiency of solving under constraints: static pruning: according to the access restriction L, pre-select the edge set Eliminate the inaccessible edges and get This avoids repeatedly checking the availability of edges during the search process and pre-processes the spatial constraints. Dynamic pruning: During the A* search process, the heuristic function h(n) guides the search direction, giving priority to expanding more promising nodes and avoiding ineffective exploration in low-quality solution spaces.

[0037] Finally, this application incorporates the task priority Q into the heuristic cost function :High-priority tasks (large Q values) are given additional priority rights of way and tend to choose better road sections during path search; the λ parameter controls the influence of priority, achieving a quantitative trade-off between priority constraints and other constraints; this design enables the system to handle ordinary transportation and emergency tasks within the same framework without switching algorithms.

[0038] Furthermore, the A* algorithm is used to search for the global path P, including: initializing the open list and the closed list, and placing the starting node Add to the open list; loop to select the cost function in the open list The smallest node is expanded, and each edge of the expansion is constrained according to the obstacle avoidance level, minimum vehicle distance, maximum downhill slope and braking response time in the safety policy parameter S; the adjacent nodes that meet the constraints are added to the open list, and the current node is moved to the closed list until the target node Added to the closed list or the open list is empty; backtrack from the closed list to get the starting node To the target node The optimal node sequence For each road segment , calculate the global expected speed based on the path preference E , based on the state prediction information and road section length of the dynamic target Calculate estimated travel time , according to the edge weight Calculate road segment cost ; The optimal node sequence And the corresponding driving parameter set Encapsulated as a global path P.

[0039] Furthermore, S5 generates a local path that satisfies the safety distance constraint and the path deviation constraint; obtains the execution status of the local path and feeds it back to the cloud control platform for dynamic path optimization, including: representing the global path P as a node sequence ,in, ,in, Represents the three-dimensional coordinate data of the i-th path point, represents the global expected velocity, Indicates the estimated time of travel; obtains the set of obstacles detected by the vehicle ,in, ,in, Indicates the location of the obstacle. Represents the obstacle radius or influence range; based on the global path P and the obstacle set O, a local path that meets the obstacle avoidance constraints is generated through the local path optimization algorithm ;

[0040] Furthermore, based on the global path P and the obstacle set O, a local path that satisfies the obstacle avoidance constraints is generated through the local path optimization algorithm. , including: constructing the local path optimization objective function J: ,in, Represents the coordinates of the kth node on the local path; for The corresponding reference point on the global path P; represents the planned speed of the kth node on the local path, Indicates location The obstacle cost function at ; K represents the total number of nodes in the local path; represents the weight coefficient;

[0041] Set path fusion constraints: ;in, Indicates the maximum allowed path deviation threshold; sets obstacle avoidance constraints: ,in, is the minimum safe distance threshold between the vehicle and the obstacle; solve the objective function J with constraints and generate a local path that satisfies both the obstacle avoidance constraint and the path fusion constraint. .

[0042] In particular, the local path optimization objective function J integrates three key objectives: the first (path tracking): maintaining consistency with the global path and inheriting the efficiency optimization results of the cloud; the second (speed smoothing): avoiding frequent acceleration and deceleration to improve ride comfort and energy efficiency; the third (obstacle avoidance safety): dynamically adjusting the path according to the distribution of obstacles to ensure driving safety; by adjusting the weight coefficients α, β, and γ, the system can dynamically balance the weights of safety and efficiency according to the real-time environment.

[0043] In addition, this application sets up two layers of security: soft constraints (objective function): through the obstacle cost function Guide the path away from obstacles. Hard constraints (mandatory conditions): Path deviation constraints: , to prevent local adjustments from deviating excessively from the global path; obstacle avoidance safety constraints: , ensuring a minimum safe distance from obstacles; this combination of soft and hard constraint design not only ensures that the safety bottom line is not broken, but also reserves space for efficiency optimization.

[0044] Compared with the existing technology, the advantages of this application are:

[0045] This paper aims to address the decision conflicts and optimization difficulties caused by multi-dimensional constraint coupling in unmanned driving path planning in underground mining environments in existing technologies.

[0046] This application first integrates multi-source V2X data such as OBU and RSU through the Bayesian fusion algorithm to build a real-time traffic semantic graph G=(V, E), and considers time constraints in the path request verification stage. and space access restrictions , using graph theory-based heuristic search in the set of available road segments Find a feasible path that meets the time limit.

[0047] Secondly, the task priority factor is introduced into the heuristic cost function , so that high-priority tasks can obtain lower cost evaluation during path search, while verifying resource constraints and path optimality constraints Ensure proper allocation of resources among multiple tasks.

[0048] Finally, a cloud-vehicle layered optimization architecture is adopted. The cloud generates the most efficient global path P based on global information, and the vehicle optimizes the local path through the objective function J, taking into account the path tracking accuracy, speed smoothness, and obstacle avoidance safety. and path deviation constraints Under certain conditions, a dynamic balance between safety and traffic efficiency can be achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The present application 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:

[0050] Figure 1 This is an exemplary flow chart of a global path planning method for a V2X-based underground mining unmanned driving cloud control platform according to some embodiments of the present application;

[0051] Figure 2 This is an exemplary flowchart of generating a global path according to some embodiments of the present application. DETAILED DESCRIPTION

[0052] The method and system provided in the embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0053] like Figure 1 As shown in the figure, S1 collects multi-source data, which includes vehicle status data collected by OBU equipment, road traffic data collected by RSU equipment, traffic signal status data and underground mining map data; S2 fuses the multi-source data through the Bayesian fusion algorithm to construct the traffic semantic graph data structure. ; S3, according to the task issued or the work plan adjustment, generate the path request data R, and perform time constraint, resource constraint and path optimality constraint verification on the path request data R; wherein the path request data R includes: starting position , end position , start time , latest completion time , task priority Q, access restriction L, safety policy parameter S and path preference E; S4, based on the verified path request data R, and traffic semantic graph data structure , a heuristic search algorithm based on graph theory is used to search for the optimal path of the road topology structure, and a global path P containing a node sequence and driving parameters is generated; S5, obstacle data detected by the vehicle side through local sensors is obtained, and obstacle avoidance correction is performed through a local path optimization algorithm based on the obstacle data and the global path P, to generate a local path that meets the safety distance constraint and path deviation constraint; the execution status of the local path is obtained and fed back to the cloud control platform for dynamic path optimization.

[0054] Specifically, the cloud control center, as the core processing platform of the system, performs the following initialization tasks during the system startup phase: (1) Road topology map construction: Import high-precision map data (HD map) of the target area, including road topology, lane information, intersection connection relationships, speed limit information, etc., and construct a graph structure data model for path calculation. (2) Basic traffic model loading: Initialize the traffic rule configuration in the area, establish traffic priority, restricted areas, lane type classification (main road / branch road), static obstacle location, etc. (3) Communication channel verification and synchronization: Establish communication with each RSU and OBU, test data transmission and return channels, and ensure that network reliability and latency meet the real-time requirements of path planning.

[0055] Through the V2X communication network, the system begins to continuously collect and summarize the following real-time traffic information: (1) vehicle status information; (2) road traffic information (from RSU equipment); (3) traffic signal status information;

[0056] Since the information collected comes from a variety of sources, there are problems such as time delay and format differences. Therefore, the system performs the following data fusion preprocessing: (1) Time synchronization: uniformly adjust the time accuracy of all data sources based on timestamps to eliminate timing conflicts. (2) Spatial alignment: align data in a unified coordinate system to ensure consistent map matching. (3) Data cleaning and anomaly removal: remove outliers and noise interference data to improve data quality. After completing this step, the system forms a fused "dynamic traffic perception map" as the input basis for path planning and decision control.

[0057] The goal of this step is to fuse information from multiple V2X devices (including roadside units (RSUs), cameras, radars, unmanned vehicle sensors, cloud-based traffic platforms, etc.) to form a structured traffic semantic graph (Traffic Semantic Graph), providing a global understanding of the traffic environment for subsequent path planning.

[0058] The main input data includes V2X device perception information (including vehicle position, speed, acceleration, direction, etc.), traffic infrastructure information (including traffic light status, speed limit signs, road geometry), road topology (including lanes, intersections, ramps, etc. connection relationships), and real-time traffic event information.

[0059] The data processing process is as follows: time synchronization and coordinate unification; using timestamps to align observation data from different devices; and coordinate conversion based on high-precision maps (HD maps).

[0060] Integrate multiple sensory information using weighted Bayesian fusion: ,in, : Estimated value after fusion (position, speed, etc.); : No. Data source observations; : No. The variance of each data source; : Fusion weight, reflecting confidence.

[0061] Object tracking and recognition, using extended Kalman filter (EKF) to achieve dynamic target tracking:

[0062] ;

[0063] ;

[0064] ;

[0065] ;

[0066] in: : current state estimate; F: state transfer matrix; H: observation matrix; Q, R: process noise covariance.

[0067] Traffic semantic graph modeling and graph model construction , where: V: represents traffic elements (lanes, vehicles, traffic lights, obstacles, etc.); E: represents the spatial relationship between elements (lane connection, lane change, following relationship); each node Contains semantic tags: .

[0068] After modeling, the output is a structured traffic semantic graph (Graph G), which can be used by path planning algorithms to make decisions. It also provides predictions of vehicle behavior for each lane, intersection, and vehicle, including predicted trajectory, congestion probability, and traffic priority.

[0069] This step aims to trigger path planning requests during unmanned transport operations in underground mines based on the task, vehicle status, and real-time mining environment. It also specifies the target parameters required for path planning, including the starting point, end point, time requirements, job priority, and environmental restrictions, thereby providing precise constraints for subsequent global path planning calculations.

[0070] In underground mining environments, route requests for unmanned vehicles are typically triggered in the following situations:

[0071] (1) Task dispatch trigger: When the scheduling system or cloud control platform assigns a new transportation task, a route request is automatically generated.

[0072] (2) When the job plan adjustment is triggered, the job scheduling changes (task time changes, priority adjustment), and the path needs to be replanned.

[0073] The successful implementation of path planning depends on accurate target parameter input, which mainly includes the following:

[0074]

[0075] The above parameters will be uniformly structured into a path request structure: .

[0076] All path requests are encapsulated as a request set , the system allocates according to the vehicle identification: .

[0077] Every request It is automatically generated by the system dispatch center or edge computing unit according to task requirements and sent to the path planning module as input.

[0078] Path requests must meet certain time and space constraints, including:

[0079] Time constraints: .

[0080] Resource constraints: and .

[0081] Path optimality goal: ,in Represents the path cost.

[0082] This step outputs structured path request parameters, which are passed as input to the subsequent path planning module. The platform performs path calculation, scheduling, and conflict detection based on priority and resource usage.

[0083] like Figure 2 As shown in the figure, global path planning aims to calculate the optimal driving path that meets the requirements of safety, efficiency and task priority based on the path request parameters defined in the previous step, combined with the global traffic environment of the mining area and multi-vehicle cooperation constraints.

[0084] From the traffic semantic graph data structure Extract road topology data and use lane nodes in traffic element node V as vertex set for path planning , the lane connection relationship in the spatial relationship E between elements is used as the edge set , build a path search graph ,in, represents the edge weight; , where e represents the edge connecting two nodes; represents the road segment distance corresponding to edge e, Represents the fusion estimation value in the traffic semantic graph data structure The estimated travel time is calculated based on the vehicle speed in Indicates the risk value of the road section; is the weight coefficient;

[0085] Request the starting point location in the data R according to the path and end position , in the path search graph Determine the corresponding starting point node and target node ; According to the access restriction L, from the edge set Eliminate the inaccessible edges and get the set of available edges ;

[0086] Establishing heuristic cost function ,in, Represents the cumulative path cost from the starting node vs to the current node n, Indicates the transition from the current node n to the target node The heuristic estimated cost is , and Q represents the task priority;

[0087] Use the A* algorithm to search for the global path P:

[0088] Initialize the open list and closed list, and add the starting node vs to the open list;

[0089] Loop through the open list to select the cost function The smallest node is expanded, and each expanded edge is constrained according to the obstacle avoidance level, minimum vehicle distance, maximum downhill slope, and braking response time in the safety policy parameter S;

[0090] Add the adjacent nodes that meet the constraints to the open list and move the current node to the closed list until the target node Added to the closed list or the open list is empty;

[0091] Backtrack from the closed list to get the starting node To the target node The optimal node sequence ;

[0092] For each road segment , calculate the global expected speed based on the path preference E , based on the state prediction information and road section length of the dynamic target Calculate estimated travel time , according to the edge weight Calculate road segment cost ;

[0093] The optimal node sequence And the corresponding driving parameter set Encapsulated as a global path P.

[0094] This step is mainly responsible for sending the global path solution calculated in the cloud to the unmanned vehicle, and combining it with the vehicle's local perception and control system to achieve effective integration of the global path and the local dynamic environment, ensuring that the vehicle executes the path safely, smoothly and efficiently.

[0095] The path is distributed, and the global path planned on the cloud is represented as a node sequence: ;in Represents the three-dimensional coordinates of the path point and the expected speed and planned arrival time .

[0096] The path delivery process is to Transmitted to the vehicle via V2X communication.

[0097] Local path perception and adjustment: a set of obstacles is detected within the vehicle's local perception area: , ,in, is the radius or influence range of the obstacle.

[0098] Based on the global path and obstacles, the vehicle generates a local path , satisfying the obstacle avoidance constraints: ,in, is the safety distance threshold.

[0099] The vehicle local path planning is performed by optimizing the following objective function: ,in, Represents the coordinates of the kth node on the local path; for The corresponding reference point on the global path P; represents the planned speed of the kth node on the local path, Indicates location The obstacle cost function at ; K represents the total number of nodes in the local path; represents the weight coefficient;

[0100] The local path must be consistent with the global path: ,in, It is the local path deviation threshold, which ensures that the local adjustment does not deviate too far from the overall plan.

[0101] The vehicle feeds local perception and path adjustment information back to the cloud: , where s represents the current state of the vehicle and the path execution status, which is used for subsequent dynamic path optimization in the cloud.

[0102] This step achieves the seamless integration of the cloud-based global path and the vehicle's local dynamic perception, ensuring the global optimization of path planning while fully adapting to the complex and changing local environment of underground mines, thereby improving the safety and execution efficiency of unmanned vehicles.

[0103] The invention of the present application and its implementation methods are described schematically above. This description is not restrictive. Without departing from the spirit or basic features of the present application, the present application can be implemented in other specific forms. What is shown in the accompanying drawings is only one of the implementation methods of the invention of the present application, and the actual structure is not limited to this. Therefore, if a person of ordinary skill in the art is inspired by it, without departing from the purpose of the invention, a structural method and embodiment similar to the technical solution are designed without creativity, which should all fall within the scope of protection of the present application. In addition, the word "including" does not exclude other elements or steps, and the word "one" before an element does not exclude the inclusion of "multiple" elements. Words such as first and second are used to indicate names and do not indicate any specific order.

Claims

1. A global path planning method for an unmanned driving cloud control platform in underground mines based on V2X, characterized in that: include: S1, collect multi-source data, including: vehicle status data collected by OBU equipment, road traffic data collected by RSU equipment, traffic signal status data and underground mine map data; S2, using the Bayesian fusion algorithm to fuse multi-source data and construct a traffic semantic graph data structure ; S3, based on the task delivery or job plan adjustment, generate path request data R, and verify the time constraint, resource constraint and path optimality constraint of the path request data R; wherein the path request data R includes: starting point location , end position , start time , latest completion time , task priority Q, access restriction L, security policy parameter S and path preference E; S4, based on the verified path request data R and traffic semantic graph data structure , a heuristic search algorithm based on graph theory is used to search for the optimal path of the road topology structure, and a global path P containing node sequence and driving parameters is generated; S5 obtains obstacle data detected by local sensors on the vehicle side. Based on the obstacle data and the global path P, the vehicle performs obstacle avoidance corrections using a local path optimization algorithm to generate a local path that meets the safety distance constraint and path deviation constraint. The local path execution status is obtained and fed back to the cloud control platform for dynamic path optimization.

2. The global path planning method for the V2X-based underground mining unmanned driving cloud control platform according to claim 1 is characterized by: S2, building traffic semantic graph data structure ,include: The multi-source data is fused through the Bayesian fusion algorithm to obtain the fusion estimation value of each traffic element ; The extended Kalman filter algorithm is used to track the fused multi-source data and obtain the state prediction information of the dynamic target; Construct a traffic semantic graph data structure based on the fusion estimation value of each traffic element, the state prediction information of the dynamic target, and the road topology structure extracted from the underground mining map data , where V represents the traffic element node and E represents the spatial relationship between elements; Among them, traffic element nodes include: lanes, vehicles, traffic lights and obstacles; The spatial relationships between elements include: lane connections, lane changes, and following relationships; Traffic elements include: lane position, vehicle position, vehicle speed, traffic light status, and obstacle position.

3. The global path planning method for the V2X-based underground mining unmanned driving cloud control platform according to claim 2 is characterized by: Obtain the fusion estimate of each traffic element, including: Perform time and space synchronization on observation data from different V2X devices; For each traffic element, obtain observations from n data sources and the corresponding variance ; According to the variance of each data source , calculate the fusion weight , where the fusion weight Reflects the confidence level of each data source; According to the fusion weight Calculate the fused estimate , ,in, Represents the fused estimated value of traffic features.

4. The global path planning method for the V2X-based underground mining unmanned driving cloud control platform according to claim 3 is characterized by: Obtain state prediction information of dynamic targets, including: Initialize the state vector of the dynamic target and the error covariance matrix ; Based on the state transfer matrix F and process noise covariance Q, the state prediction equation Calculate the prior state estimate at the current moment and use the error covariance prediction equation Calculate the prior error covariance; where B represents the input control matrix, represents the input control vector; Get the observation data at the current moment , calculate the Kalman gain based on the observation matrix H , where R is the observation noise covariance; By state update equation Calculate the posterior state estimate and update the equation through the error covariance Calculate the posterior error covariance; Output the state estimate of the dynamic target at the current moment As state prediction information.

5. The global path planning method for an unmanned driving cloud control platform in an underground mine based on V2X according to any one of claims 2 to 4, characterized in that: S3: Generate path request data R based on task delivery or job plan adjustment, and verify the time constraints, resource constraints, and path optimality constraints of the path request data R, including: When receiving a task dispatch instruction or a job plan adjustment instruction, extract the task parameters and generate path request data ; Perform time constraint verification to verify the estimated travel time in the path request data R Whether it meets: ; Perform resource constraint verification and verify the road segments in the path request data R Whether it meets: and , where path represents the starting point To the end position The set of feasible paths; Perform path optimality constraint verification to verify whether: ,in, represents the path cost of the i-th path; Encapsulate the path request data R that passes all constraint validations into a structured request set , and sort them according to task priority Q.

6. The global path planning method for the V2X-based underground mining unmanned driving cloud control platform according to claim 5 is characterized by: Starting point Indicates the current vehicle position coordinates, which are derived from the vehicle-mounted GNSS+IMU integrated navigation or the underground UWB positioning system; End position Indicates the coordinates of the mission destination, including loading point, unloading point or maintenance station; Start time Indicates the scheduled start time of the mission, used to determine whether the vehicle can start in time; Latest completion time Indicates that the destination must be reached before this time, which is used for task timeout judgment and path optimization; Task priority Q represents a numerical priority identifier, which is used for multi-task path competition or scheduling sorting; The access restriction L represents the set of road sections that are currently allowed to pass, which is restricted by landslides, maintenance or traffic flow; The safety policy parameter S represents the safety constraints of obstacle avoidance level, minimum vehicle distance, maximum downhill slope, or braking response time; Path preference E indicates the preferred route type, including main lane priority, avoiding slippery areas, or avoiding manual work areas.

7. The global path planning method for the V2X-based underground mining unmanned driving cloud control platform according to claim 6 is characterized by: S4, generating a global path P containing node sequences and driving parameters, including: From the traffic semantic graph data structure Extract road topology data and use lane nodes in traffic element node V as vertex set for path planning , the lane connection relationship in the spatial relationship E between elements is used as the edge set , build a path search graph ,in, represents the edge weight; Request the starting point location in the data R according to the path and end position , in the path search graph Determine the corresponding starting point node and target node ; According to the access restriction L, from the edge set Eliminate the inaccessible edges and get the set of available edges ; Establishing heuristic cost function ,in, Represents the cumulative path cost from the starting node vs to the current node n, Indicates the transition from the current node n to the target node The heuristic estimated cost is , and Q represents the task priority; Use A* algorithm to search for the global path P; in: , where e represents the edge connecting two nodes; represents the road segment distance corresponding to edge e, Represents the fusion estimation value in the traffic semantic graph data structure The estimated travel time is calculated based on the vehicle speed in Indicates the risk value of the road section; is the weight coefficient.

8. The global path planning method for the V2X-based underground mining unmanned driving cloud control platform according to claim 7 is characterized by: Use the A* algorithm to search for the global path P, including: Initialize the open list and closed list, and set the starting node Join the open list; Loop through the open list to select the cost function The smallest node is expanded, and each expanded edge is constrained according to the obstacle avoidance level, minimum vehicle distance, maximum downhill slope, and braking response time in the safety policy parameter S; Add the adjacent nodes that meet the constraints to the open list and move the current node to the closed list until the target node Added to the closed list or the open list is empty; Backtrack from the closed list to get the starting node To the target node The optimal node sequence ; For each road segment , calculate the global expected speed based on the path preference E , based on the state prediction information and road section length of the dynamic target Calculate estimated travel time , according to the edge weight Calculate road segment cost ; The optimal node sequence And the corresponding driving parameter set Encapsulated as a global path P.

9. The global path planning method for the V2X-based underground mining unmanned driving cloud control platform according to claim 8 is characterized by: S5, generating a local path that satisfies the safety distance constraint and path deviation constraint, including: Represent the global path P as a sequence of nodes ,in, ,in, Represents the three-dimensional coordinate data of the i-th path point, represents the global expected velocity, Indicates the estimated time of travel; Get the set of obstacles detected by the vehicle ,in, ,in, Indicates the location of the obstacle. Indicates the radius or impact range of the obstacle; According to the global path P and the obstacle set O, a local path that satisfies the obstacle avoidance constraints is generated through the local path optimization algorithm. .

10. The global path planning method for the V2X-based underground mining unmanned driving cloud control platform according to claim 9 is characterized in that: Generate a local path that meets the obstacle avoidance constraints through a local path optimization algorithm ,include: Construct the local path optimization objective function J: ,in, Represents the coordinates of the kth node on the local path; for The corresponding reference point on the global path P; represents the planned speed of the kth node on the local path, Indicates location The obstacle cost function at ; K represents the total number of nodes in the local path; represents the weight coefficient; Set path fusion constraints: ;in, Indicates the maximum allowed path deviation threshold; Set obstacle avoidance constraints: ,in, is the minimum safe distance threshold between the vehicle and the obstacle; Solve the objective function J with constraints and generate a local path that satisfies both obstacle avoidance constraints and path fusion constraints .

Images