An OHT track node automatic identification and path planning method and system

Abstract

The application discloses an OHT track node automatic identification and path planning method and system, relates to the technical field of OHT tracks, and comprises the following steps: identifying track node positions by a multimode verification method based on vehicle motion trajectory feature analysis, point cloud geometry clustering processing and the combination of visual and magnetic features; obtaining node types by performing semantic identification according to the node positions, and constructing a track topology graph according to the node types; generating a candidate path set through the track topology graph, and locally optimizing the task execution order and vehicle matching relationship of the candidate path set; based on the local optimization result, synchronously adjusting path selection and task allocation, and outputting an integrated scheduling scheme; the application can realize high-precision positioning of nodes, semantic classification, construction of a structured track topology, and integrated optimization of multi-vehicle path and task scheduling, and significantly improves the operation efficiency, scheduling reliability and resource utilization rate of an OHT system.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of OHT track, more particularly, the present application relates to an OHT track node automatic identification and path planning method and system. BACKGROUND

[0002] In modern intelligent manufacturing and automated warehouse systems, the automated track transportation system (Overhead Hoist Transport, OHT for short) as a key logistics transportation equipment is widely used in material handling in factories, warehouses and distribution centers. However, the OHT system faces core technical challenges such as track node identification, path planning and multi-vehicle scheduling coordination in actual operation.

[0003] In the prior art, the node identification of the OHT track system mainly relies on manual measurement or single sensor collection, such as identifying track bifurcation, intersection and termination node positions by laser ranging, vision or magnetic field sensor alone. Although these methods can obtain part of the information of the track node, they have limited precision, are greatly affected by environmental noise, and have low automation degree. In addition, in the aspects of node semantic identification and track topology graph construction, the existing methods mostly use fixed rules or simple geometric matching means, which cannot accurately classify the node types in complex track networks, and it is also difficult to generate structured and computable track topology graphs.

[0004] In the aspect of path planning and scheduling, traditional methods are usually based on static topology or single-vehicle path search, lacking real-time consideration of dynamic resource occupation, node congestion and time window constraints in multi-vehicle and multi-task environment, which easily leads to path conflict, task delay or resource waste. For example, when multiple OHT vehicles run on the same track segment, the existing scheduling method cannot effectively coordinate the path selection and task allocation between vehicles, leading to decreased running efficiency and even safety risks.

[0005] In view of the above problems, the present application provides a solution. SUMMARY

[0006] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present application provide an OHT track node automatic identification and path planning method and system, which can realize high-precision positioning, semantic classification, structured track topology construction, and integrated optimization of multi-vehicle path and task scheduling, significantly improving the running efficiency, scheduling reliability and resource utilization of the OHT system.

[0007] To achieve the above purpose, the present application provides the following technical solutions:

[0008] In a first aspect, the application provides an OHT track node automatic identification and path planning method, which comprises: identifying track node positions based on a vehicle motion trajectory feature analysis, point cloud geometric clustering processing, and a multimodal verification method combining visual and magnetic features; performing semantic recognition according to the node positions to obtain node types, and constructing a track topology graph according to the node types; generating a candidate path set through the track topology graph, and locally optimizing the task execution order and vehicle matching relationship of the candidate path set; based on the local optimization result, synchronously adjusting path selection and task allocation, and outputting an integrated scheduling scheme.

[0009] In one embodiment, the track node positions are identified based on a vehicle motion trajectory feature analysis, point cloud geometric clustering processing, and a multimodal verification method combining visual and magnetic features, specifically: trajectory data is obtained, and the trajectory data is smoothed to obtain smoothed trajectory data; based on the smoothed trajectory data, the difference values between consecutive time instants are calculated in time sequence to obtain speed and steering rate of change; the trajectory abnormal section is identified according to the speed and steering rate of change; taking the center point of the abnormal section as a reference, a fixed trajectory window is extended forward and backward, and the minimum value of the steering rate of change of the front and rear endpoints is taken as a boundary correction point to correct the start point and end point coordinates of the abnormal section to obtain a candidate section; the confidence score is obtained according to the maximum steering rate of change, maximum speed and number of abnormal points in the candidate section; the core candidate node section is obtained by judging according to the confidence score; the candidate node section is clustered and geometric features are extracted to obtain a candidate node cluster set.

[0010] In one embodiment, the candidate node section is clustered and geometric features are extracted to obtain a candidate node cluster set, specifically: based on the candidate node section, the point cloud data within each section range is extracted and preprocessed; the neighborhood point set is determined according to the processed point cloud data, and the curvature is obtained; each point cloud is labeled with a normal vector and a curvature; the labeled point cloud is divided into several cluster blocks using a density clustering algorithm; the geometric morphology of each cluster block is analyzed, the track geometric features are identified, and the cluster confidence is calculated; the candidate node cluster set is obtained by judging according to the cluster confidence; the node position is obtained by a node multimodal verification method combining visual and magnetic features on the candidate node cluster set.

[0011] In one of the embodiments, the node position is obtained through the node multimodal verification method of visual and magnetic features for the candidate node cluster set, specifically: based on the candidate node cluster set, the spatial coordinate range is selected, and according to the spatial coordinate range, the image sequence and the magnetic field signal data of each candidate node are obtained; the track key identifier is detected for the image sequence, and is aligned with the center point of the spatial coordinate range to obtain the visual verification feature; the magnetic flux change rate is obtained through the magnetic field signal data, and the magnetic anomaly point is marked; the continuous magnetic anomaly points are aggregated, and the track coordinates corresponding to the center time are taken as the magnetic feature center point; the deviation of the geometric center point from the visual verification feature and the magnetic feature center point is obtained, and the node position is determined.

[0012] In one of the embodiments, the node type is obtained through semantic recognition according to the node position, specifically: according to the recognized node position, the geometric feature and the behavior feature of the node are extracted; the geometric feature and the behavior feature are normalized and then fused to obtain a node feature comprehensive vector, wherein the geometric feature is used to describe the spatial form of the node, and the behavior feature is used to reflect the dynamic performance of the vehicle at the node; based on the node feature comprehensive vector, the node type is determined by inputting into a pre-trained node semantic classification model, and the node type label is output.

[0013] In one of the embodiments, the track topology graph is constructed according to the node type, specifically: attribute information is assigned to each node; a candidate set of inter-node links is generated according to the node type and the spatial position; the candidate set of inter-node links is scored for reliability, and the reliability score includes the deviation of the link from the historical track, the point cloud density, the track direction consistency index, and the node endpoint position deviation; high-reliability links are selected through the reliability score, and the link priority is adjusted according to the node type to generate an optimized link set; the node set with assigned attribute information and the optimized link set are combined to generate the track topology graph.

[0014] In one of the embodiments, the candidate path set is generated through the track topology graph, specifically: the current node and the task target node of each vehicle are obtained and marked as the starting point and the ending point; based on the track topology graph, an incremental graph search algorithm is used to search the feasible path from the starting point to the ending point; a plurality of feasible paths searched are formed into a preliminary candidate path set; the path dynamic priority of each path in the preliminary candidate path set is calculated, and the path dynamic priority includes the reliability score of the path, the node load index, and the track resource occupancy rate; the preliminary candidate paths are sorted from high to low according to the path dynamic priority to obtain a sorted path priority queue, and the candidate path set corresponding to each vehicle is output.

[0015] In one of the embodiments, the task execution sequence and the vehicle matching relationship of the candidate path set are locally optimized, specifically: obtaining an initial scheduling data set, and establishing a time window for each task; obtaining the expected execution time sequence of the path for each task's candidate path; matching the task expected execution time sequence with the task time window to determine whether the task's candidate path is feasible in time; performing overlap detection on the path that is feasible in time to determine whether there is a conflict; judging whether the path that has a conflict is feasible according to the path feasibility score; filtering an effective candidate path set based on the feasibility detection of time and conflict; obtaining a task dynamic priority based on the effective candidate path set, and generating a task execution sequence; preliminarily matching the task execution sequence with the candidate vehicle to obtain a task-vehicle mapping table; establishing a local optimization objective function based on the task-vehicle mapping table, the local optimization objective function including task completion time, corresponding path cost, and node congestion degree; solving the local optimization objective function to minimize to obtain a locally optimal task execution sequence and vehicle allocation scheme.

[0016] In one of the embodiments, the path selection and task allocation are synchronously adjusted based on the local optimization result to output an integrated scheduling scheme, specifically: updating the edge weight of the track topology graph based on the local optimization result, and real-time correcting the task priority and the vehicle matching strategy according to the updated edge weight, the correction including: adjusting the task execution sequence according to the node real-time congestion degree, locally re-matching the task execution sequence and the vehicle allocation scheme, and selecting a vehicle with the minimum path cost and meeting the task time window constraint; generating a new candidate path set according to the updated track topology graph; re-optimizing the task execution sequence and the vehicle matching relationship on the new candidate path set; and iteratively looping until convergence to output the integrated scheduling scheme.

[0017] In a second aspect, the present application provides an OHT track node automatic identification and path planning system, which comprises: a node position identification module for identifying track node positions based on vehicle motion trajectory feature analysis, point cloud geometry clustering processing, and multi-mode verification method combining visual and magnetic features; a topology graph construction module for obtaining node types through semantic recognition based on node positions, and constructing a track topology graph according to the node types; a candidate path set generation and optimization module for generating a candidate path set through the track topology graph, and locally optimizing the task execution sequence and the vehicle matching relationship of the candidate path set; and a feedback adjustment module for synchronously adjusting the path selection and the task allocation based on the local optimization result to output an integrated scheduling scheme.

[0018] From the above technical solutions, it can be seen that the embodiments of the present application have the following advantages:

[0019] By high-precision identification of the track node of the OHT system, semantic classification of the node type, construction of the track topology graph, and multi-vehicle path planning and task scheduling optimization based on the topology graph, the intelligent management of the whole process from data acquisition to integrated scheduling is realized. The advantages are that: by fusing the vehicle motion trajectory, point cloud geometric features, and multi-mode verification of vision and magnetic field, high precision and robustness of node positioning are realized; based on the node type and topology structure, reliable paths are generated, combined with real-time track resource occupation, node load and vehicle state for dynamic priority sorting and local optimization, which can effectively avoid task conflicts, node congestion and path delay; finally, a closed-loop iterative integrated scheduling scheme is formed, realizing safe, efficient and intelligent transportation in a multi-vehicle and multi-task environment, improving the overall operation efficiency, scheduling reliability and resource utilization rate of the OHT system. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 An OHT track node automatic identification and path planning method flowchart is provided for the embodiments of the present application.

[0021] Figure 2 An OHT track node automatic identification and path planning system structure diagram is provided for the embodiments of the present application. DETAILED DESCRIPTION

[0022] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor are within the scope of protection of the present application.

[0023] Referring to Figure 1 The OHT track node automatic identification and path planning method flowchart provided by the present application includes the following steps:

[0024] S1, based on the combination of vehicle motion trajectory feature analysis, point cloud geometric clustering processing and multi-mode verification of visual and magnetic features, the track node position is identified.

[0025] In this embodiment, the track node position is identified based on the combination of vehicle motion trajectory feature analysis, point cloud geometric clustering processing and multi-mode verification of visual and magnetic features, specifically:

[0026] Trajectory data is obtained, and the trajectory data is smoothed to obtain smoothed trajectory data, the trajectory data including three-dimensional displacement velocity variation curves and attitude angle data of the vehicle continuously moving along the track;

[0027] The trajectory data can be collected by an IMU (inertial measurement unit) and an odometer mounted on the vehicle; the smoothing processing includes applying Kalman filtering or a weighted moving average filter to the continuous displacement and attitude angle data to remove high-frequency noise, filtering the speed change curve to reduce the influence of sensor measurement errors on node identification, and obtaining smooth, continuous and reliable smoothed trajectory data to provide a stable basis for anomaly detection;

[0028] Based on the smoothed trajectory data, the difference value between consecutive time points is calculated in time sequence to obtain the speed and steering change rate;

[0029] The displacement difference between consecutive time points is calculated to obtain the speed, and the specific calculation formula is as follows:

[0030]

[0031] In the formula, is the speed of the i-th time point, is the trajectory coordinate of the i-th time point, is the time interval, is the trajectory coordinate of the i+1-th time point.

[0032] The attitude angle difference between consecutive time points is calculated to obtain the vehicle steering change rate, and the specific calculation formula is as follows:

[0033]

[0034] In the formula, is the steering change rate, is the yaw angle of the trajectory point of the vehicle at the i-th time point, is the yaw angle of the trajectory point at the i+1-th time point.

[0035] The trajectory anomaly section is identified according to the speed and steering change rate;

[0036] If the steering change rate is greater than a set steering change threshold, it is marked as a direction angle mutation section;

[0037] If the speed is greater than a set speed change threshold, it is marked as a speed anomaly section;

[0038] Taking the center point of the anomaly section as a reference, a fixed trajectory window is expanded forward and backward, and the minimum value of the steering change rate of the front and rear endpoints is taken as a boundary correction point to correct the start and end coordinates of the anomaly section, so that it completely covers the trajectory range where the potential node is located, and a candidate section is obtained;

[0039] The confidence score is obtained by considering the maximum steering change rate, maximum speed and number of abnormal points in the candidate section;

[0040] The confidence score, specifically the calculation formula is as follows:

[0041]

[0042] In the formula, is the confidence score, is the maximum steering rate of change, is the steering change threshold, is the maximum speed, is the speed change threshold, is the number of consecutive trajectory points marked as abnormal in the candidate section, is the minimum reference value of the number of abnormal points, , , are weight coefficients, respectively.

[0043] According to the confidence score, a core candidate node section is obtained; if the confidence score is greater than a preset confidence threshold, it is a core candidate node section; otherwise, it is a suspicious section;

[0044] The candidate node sections are clustered and geometric features are extracted to obtain a candidate node cluster set.

[0045] Further, the candidate node sections are clustered and geometric features are extracted to obtain a candidate node cluster set, specifically:

[0046] Based on the candidate node sections, laser radar point cloud data within each section range is extracted and preprocessed, the preprocessing including removing isolated noise points using a statistical outlier removal method, downsampling the point cloud using a voxel grid wave to reduce the amount of calculation and maintain the spatial structure, and performing local normalization or alignment processing on the point cloud coordinates to make the trajectory direction consistent;

[0047] A neighborhood point set is determined according to the processed point cloud data, and the neighborhood can adopt a fixed number of K-neighbors;

[0048] According to the neighborhood point set, curvature is obtained by a local plane fitting method, and each point cloud is labeled with a normal vector and curvature; the greater the curvature, the more obvious the local bending;

[0049] The labeled point cloud is divided into a plurality of cluster blocks using a density clustering algorithm;

[0050] Geometric morphological analysis is performed on each cluster block to identify trajectory geometric features, and a cluster confidence is calculated, the trajectory geometric features including bifurcation, intersection, termination or bending features;

[0051] The cluster confidence, specifically the calculation formula is as follows:

[0052]

[0053] wherein, is a clustering confidence, is a clustering average curvature, is a curvature threshold, is a number of in-cluster points, is a minimum number of points threshold, is a track direction consistency index (cosine of the angle between the principal direction and the historical track direction), , , are weight coefficients, respectively.

[0054] According to the clustering confidence, a core candidate node is determined when the clustering confidence exceeds a preset clustering confidence threshold, and a candidate node clustering set is obtained, wherein the candidate node clustering set includes a point cloud geometry feature extraction result.

[0055] A node multi-modal verification method based on visual and magnetic features is used on the candidate node clustering set to obtain a node position.

[0056] It should be noted that through a multi-level, parameterized, and computable point cloud geometry analysis process, high-precision and verifiable automatic identification of OHT track nodes is achieved. Specifically, first, the point cloud acquisition and preprocessing steps ensure the integrity and noise suppression of the input data, providing a reliable foundation for subsequent identification. Second, local curvature and normal vector calculation converts track spatial morphology into quantifiable geometric features, making node bending, branching, and other features mathematically distinguishable. Third, density-based clustering effectively separates different track structure sections and accurately identifies potential node regions. Finally, a confidence score mechanism is introduced to integrate curvature features, point cloud density, and direction consistency for evaluation, enabling quantitative decision-making and reliable filtering for node identification. This significantly improves the robustness, accuracy, and practicality of node detection.

[0057] Further, a node multi-modal verification method based on visual and magnetic features is used on the candidate node clustering set to obtain a node position, specifically:

[0058] Based on the trajectory and point cloud geometry feature extraction results in the candidate node clustering set, the spatial coordinate range of each candidate node clustering region is selected, and the corresponding camera image frames and magnetic field sensor measurement signals are associated with the spatial coordinate range to obtain image sequences and magnetic field signal data for each candidate node.

[0059] For the image sequence, computer vision methods are used to detect track key identifiers, and the center point of the spatial coordinate range of the candidate node is aligned to obtain visual verification features of the candidate node. The track key identifiers include track intersection points, track branching markers, or connecting bracket positions.

[0060] Obtain the magnetic flux rate of change through the magnetic field signal data, and mark the magnetic anomaly point;

[0061] The magnetic flux rate of change is specifically calculated according to the following formula: , wherein, is the magnetic flux rate of change, is the magnetic induction intensity value collected at the time point ; if , wherein, is an empirical threshold coefficient, is the standard deviation of the magnetic signal, and the magnetic anomaly point is marked;

[0062] Aggregate the continuous magnetic anomaly points, and take the trajectory coordinates corresponding to the center time as the magnetic feature center point;

[0063] Obtain the deviation of the geometric center point from the visual verification feature and the magnetic feature center point respectively;

[0064] It should be noted that obtaining the deviation of the geometric center point from the visual verification feature and the magnetic feature center point respectively can be understood as the difference between the coordinates of the geometric center point and the coordinates of the visual verification feature, and the difference between the coordinates of the geometric center point and the coordinates of the magnetic feature center point.

[0065] If the deviations do not exceed the preset empirical threshold, it is determined that the candidate node is a real node position.

[0066] It should be noted that the node position obtained by the above technical solution has the significant advantages of high precision, strong robustness and multi-source credible verification. The scheme combines trajectory motion features (speed, turning change), point cloud geometric shape (curvature, direction consistency) and visual and magnetic field multi-mode verification organically, forming a layer-by-layer progressive identification process from data acquisition, anomaly detection, candidate section clustering to multi-source verification. Through smoothing the trajectory and confidence, the sensor noise and local false detection are effectively suppressed; through point cloud curvature analysis and density clustering, the geometric interpretability of the node feature is ensured; and through the double matching of visual and magnetic features, the multi-mode verification of spatial consistency is realized, thereby avoiding single sensor misidentification. The node position finally output by the method not only has high positioning accuracy and strong repeatability, but also can maintain stable identification ability in complex environments, providing a reliable structured space reference for path planning and scheduling of the OHT system, and significantly improving the intelligent level and operation efficiency of the overall automated transportation system.

[0067] S2, according to the node position, semantic recognition is performed to obtain a node type, and an orbit topology graph is constructed according to the node type.

[0068] In this embodiment, the node type is obtained according to the semantic recognition of the node position, specifically:

[0069] extracting geometric features of the node according to the identified node position;

[0070] extracting behavior features related to the node interaction based on historical motion data of the vehicle within the node position, the behavior features including speed, steering rate of change, deceleration / stop, and steering or diverging behavior of the vehicle;

[0071] normalizing the geometric features and the behavior features to obtain a node feature comprehensive vector, wherein the geometric features are used to describe the spatial form of the node, and the behavior features are used to reflect the dynamic performance of the vehicle at the node;

[0072] inputting the node feature comprehensive vector into a pre-trained node semantic classification model to determine the type of the node;

[0073] The node semantic classification model includes a classification algorithm based on machine learning or rule determination, which is used to divide the node into a cross node, a diverging node, a converging node, a terminating node, or a normal straight line node, etc.

[0074] The node type label includes a cross node, a diverging node, a converging node, a terminating node, and a straight line node.

[0075] It should be noted that through the fusion recognition mechanism of geometric structure features and behavior dynamic features, the automatic semantic classification of node types is realized, which can accurately distinguish cross, diverging, converging, and terminating node types in complex track environments, thereby providing a structured and computable semantic basis for subsequent path planning and scheduling strategies, and significantly improving the intelligent decision-making ability and operation safety of the system.

[0076] Further, a track topology graph is constructed according to the node type, specifically:

[0077] attribute information is assigned to each node, including node type label, node number and spatial coordinates, node direction vector, and number of connectable tracks;

[0078] According to the node type and the spatial position, a candidate set of connections between nodes is generated, specifically:

[0079] For straight or terminating nodes, the shortest connection candidate is generated;

[0080] For diverging or converging nodes, multi-branch candidate connections are generated, considering that the branch angle is within a preset range and the track direction is consistent;

[0081] For cross nodes, cross connection candidates are generated to ensure that the cross section is perpendicular to the node direction vector;

[0082] The reliability score of the candidate set of inter-node connections includes a deviation of the connection from historical trajectories, a point cloud density and track direction consistency index, and a node end position deviation;

[0083] The reliability score is calculated according to the following formula:

[0084]

[0085] In the formula, is the reliability score, is a normalized deviation of the connection from historical trajectories, is a point cloud density and track direction consistency index, is an end position deviation, is a maximum end position deviation, , , are weight coefficients, respectively.

[0086] High-reliability connections are screened through the reliability score, connections with a reliability score lower than a preset threshold are removed, high-reliability connections with a reliability score higher than the preset threshold are retained, the connection priority is adjusted according to the node type, and an optimized connection set is generated;

[0087] The connection priority is adjusted according to the node type to generate the optimized connection set, specifically:

[0088] For straight-line nodes, the priority is inversely proportional to the connection length, and the shorter the length, the higher the priority;

[0089] For terminal nodes, the connection with the highest priority is the only exit, and other connections are set to low priority or removed;

[0090] For diverging / converging nodes, the priority is adjusted according to the angle between the node direction vector and the connection direction, and the connection with an angle close to the main direction of the node has a high priority;

[0091] For intersection nodes, the direction vectors of the intersection segments are ensured to be orthogonal, and the priority is adjusted according to the spatial overlap degree and historical trajectory consistency. The higher the spatial overlap degree, the more likely the connection is a real existing track segment, so the priority is improved. Connections with a high number of passes and good trajectory continuity indicate that they are actual passable track segments, and the priority should be improved accordingly;

[0092] It should be noted that an initial priority is set for each connection, which is consistent with the preliminary reliability score of the connection;

[0093] The node set with assigned attribute information is combined with the optimized connection set to generate a track topology graph.

[0094] Wherein, the node is the vertex of the graph, the connecting line is the edge of the graph, and the connecting line weight is the comprehensive calculation of the spatial length, the node load index and the reliability score weighted sum; the node load index represents the number of task requests of the node within a preset time window.

[0095] It should be noted that the track topology graph constructed by the above scheme can realize high-precision and structured representation of the OHT track system nodes and connecting lines, and the connecting lines are screened and priority optimized by using the node type information, spatial position, direction vector and reliability score of multi-source data, so as to eliminate low reliability or redundant connecting lines, and ensure that the generated topology graph conforms to the actual track connection structure; this method not only improves the accuracy and robustness of the track topology graph, but also supports path planning, scheduling optimization and track management, realizes automation, verifiability and update of track information, and greatly improves the overall operation efficiency and safety of the OHT system.

[0096] S3, generating a candidate path set through the track topology graph, and locally optimizing the task execution order and vehicle matching relationship of the candidate path set.

[0097] In this embodiment, the candidate path set is generated through the track topology graph, specifically:

[0098] The current node and task target node of each vehicle are obtained and marked as the starting point and the end point;

[0099] Based on the track topology graph, an incremental graph search algorithm is used to search the feasible path from the starting point to the end point, and each path in the feasible path includes node order, connecting line sequence, cumulative weight and track resource occupation rate;

[0100] Wherein, the track resource occupation rate represents the time proportion of the path currently occupied by the vehicle, and the incremental search algorithm updates the path cost function in real time according to the dynamic weight of the edge at each expansion, so that the search result considers both real-time and scalability, wherein the incremental graph search algorithm is a dynamic version based on hierarchical A* algorithm, which dynamically adjusts the heuristic function weight according to the change of track resource state in the search process, so as to ensure the real-time of the search;

[0101] It should be noted that in the path planning algorithm, the path cost function is a calculation model for comprehensively evaluating the advantages and disadvantages of a path, and is used to guide the expansion direction of incremental graph search, including: track length cost, representing the cumulative length of the path, calculated by summing the lengths of the connected lines, topological direction cost, representing the consistency of the path direction, calculated by taking the average of the cosine of the angle between the direction of each connected line in the path and the main direction of vehicle operation, reliability cost, reflecting the reliability of the path according to the reliability score of the connected line, time cost, estimating the path execution time in combination with the current speed of the vehicle and the track passage state, the path cost function is represented in the form of weighted sum of track length cost, topological direction cost, reliability cost and time cost;

[0102] The plurality of feasible paths searched are formed into a preliminary candidate path set;

[0103] The path dynamic priority of each path in the preliminary candidate path set is calculated, the path dynamic priority including the reliability score of the path, the node load index and the track resource occupancy rate;

[0104] The path dynamic priority is determined by linear weighting;

[0105] The preliminary candidate paths are sorted according to the path dynamic priority from high to low, to obtain a sorted path priority queue;

[0106] The candidate path set corresponding to each vehicle is output, the candidate path set including the path node sequence, the cumulative weight, the track resource state and the path dynamic priority.

[0107] Further, the task execution order and the vehicle matching relationship of the candidate path set are locally optimized, specifically:

[0108] The scheduling parameters are initialized according to the current system running state, to obtain an initial scheduling data set containing vehicle state, task constraint and path feasibility, the scheduling parameters including the task queue, position coordinate and load state of each vehicle, the urgency, target node and deadline of each task, and the occupancy state and passable time slot of each track node;

[0109] According to the initial scheduling data set, a time window is established for each task;

[0110] The specific calculation formula of the time window is as follows:

[0111]

[0112] In the formula, is the time window, is the earliest start time allowed for the task, is the latest end time that the task must be completed.

[0113] For each candidate path of a task, the expected execution time sequence of the path is calculated based on the track topology, vehicle speed and node dwell time;

[0114] The specific calculation formula of the expected execution time sequence is as follows:

[0115]

[0116] In the formula, is the time when the vehicle is expected to leave the end node of the path, is the current system time of the vehicle, is the length of the nth track on the path, is the expected driving speed of the vehicle on the segment, is the waiting time of the fth stop node on the path, is the number of path segments, is the number of stop nodes.

[0117] The expected execution time sequence of the task is matched with the task time window to determine whether a preset condition is met, and if the preset condition is met, the candidate path of the task is feasible in time;

[0118] The preset condition is that the time when the vehicle is expected to leave the end node of the path should not exceed the latest end time when the task must be completed, and the time when the vehicle is expected to enter the start node of the path should be greater than or equal to the earliest start time allowed by the task.

[0119] For the path feasible in time, the resource occupation time interval of each node and track segment, i.e. the time range from when the vehicle enters to when the vehicle leaves each node or track segment, is obtained, and overlap detection is performed with the system global occupation table, i.e. if the occupation time interval of any node or track segment in the path overlaps in time with a registered occupation interval in the global track occupation table, it indicates that there is a resource conflict or potential collision risk in the path in the time period, and it is determined that there is a conflict in the current node for the current path;

[0120] For the path with a conflict, the path feasibility score is calculated, and the path feasibility score includes conflict duration and detour cost;

[0121] The specific calculation formula of the path feasibility score is as follows:

[0122]

[0123] In the formula, is the path feasibility score, is the conflict duration, is the time delay amount, is the detour cost obtained after the path is re-planned, to represent the total path cost of the candidate path, , are weight parameters respectively.

[0124] If the path feasibility score is higher than a set threshold, the path is reserved as a conditionally feasible path.

[0125] According to the feasibility detection of time and resource occupation conflicts, an effective candidate path set of each task in the current time period is screened;

[0126] Based on the effective candidate path set, task dynamic priority calculation is performed on the task;

[0127] The specific calculation formula of the task dynamic priority is as follows:

[0128]

[0129] In the formula, is the task dynamic priority, is the task urgency, is the path cost, is the waiting time, is the congestion degree, , , , are weight parameters respectively.

[0130] According to the task dynamic priority from high to low, a task execution sequence is generated;

[0131] According to the task execution sequence, a candidate vehicle with the closest position and the minimum path cost is selected from the current idle vehicle set for preliminary matching, and a task-vehicle mapping table is obtained;

[0132] Based on the task-vehicle mapping table, a local optimization objective function is established, the local optimization objective function includes task completion time, corresponding path cost and node congestion degree, and the local optimization objective function is obtained by weighted summation;

[0133] An optimal combination is iteratively searched in the local solution space by a heuristic algorithm to minimize the local optimization objective function, and a locally optimal task execution order and vehicle allocation scheme are obtained.

[0134] It should be noted that through the above technical solution, the system can generate multiple candidate paths based on the track topology graph, and dynamically prioritize the paths in combination with the current position of the vehicle, path cost, track resource occupation and node load, and perform feasibility screening based on the time window and resource occupation, thereby realizing local optimization of the task execution order and vehicle matching relationship. The method comprehensively considers multiple levels of constraints, so that each vehicle can select a path with high reliability and high feasibility, and coordinate the scheduling relationship of multiple tasks and multiple vehicles, thereby significantly improving the integrated efficiency, real-time performance and robustness of path planning and scheduling, effectively avoiding vehicle congestion, task delay and resource conflict, and improving the operation reliability and execution efficiency of the entire OHT system.

[0135] S4, based on the local optimization result, synchronously adjusting the path selection and task allocation, and outputting an integrated scheduling scheme.

[0136] In this embodiment, based on the local optimization result, the path selection and task allocation are synchronously adjusted, and an integrated scheduling scheme is output, specifically:

[0137] Based on the local optimization result, first data is obtained, the first data including an optimized path set of each vehicle, a task execution sequence and a vehicle-task matching relationship, and a real-time node load index;

[0138] The optimized path set includes the node order, the cumulative weight, the path dynamic priority and the track resource occupation rate of each path, and the task execution sequence includes the expected completion time, the remaining time window and the task dynamic priority of each task.

[0139] According to the real-time node load index, the edge weight of the track topology graph is updated, and according to the updated edge weight, the task priority and the vehicle matching strategy are real-time corrected, specifically:

[0140] According to the real-time congestion degree of the node, the task execution order is adjusted, and according to the high and low of the resource occupation rate, the low path is preferentially allocated;

[0141] The assigned task execution order and vehicle allocation scheme are locally re-matched, and the vehicle with the minimum path cost and meeting the task time window constraint is selected;

[0142] According to the updated track topology graph, a new candidate path set is generated;

[0143] The task execution order and the vehicle matching relationship are re-optimized on the new candidate path set;

[0144] The loop iteration is performed until convergence, i.e. the preset maximum iteration number is reached.

[0145] After the closed-loop iteration ends, an integrated scheduling scheme is output, including the final path node sequence and cumulative cost of each vehicle, the final execution order, vehicle allocation and expected completion time of each task, the final congestion degree and track resource occupation rate of each track node and connection.

[0146] It should be noted that by realizing the synchronous optimization of path planning and task scheduling in the OHT system, through dynamic feedback based on the local optimization result, real-time node load, path priority and track resource occupation information are fused into track topology graph updating and task-vehicle matching adjustment, forming a closed-loop iteration mechanism, so as to continuously optimize the candidate path and task execution order. The advantage is that it can reflect the track congestion and vehicle distribution in real time, reasonably allocate vehicles and tasks, ensure efficient completion of tasks within the time window, while avoiding track resource conflicts and node congestion, improving the operation efficiency, scheduling robustness and resource utilization of the entire OHT system, realizing safe, reliable and efficient transportation in a multi-vehicle and multi-task environment.

[0147] Referring to Figure 2 The OHT track node automatic identification and path planning system structure schematic diagram provided by the application comprises a node position identification module, a topology graph construction module, a candidate path set generation and optimization module and a feedback adjustment module, and there is a connection between the modules.

[0148] The node position identification module is used for identifying the track node position based on a multi-mode verification method combining vehicle motion trajectory feature analysis, point cloud geometric clustering processing and visual and magnetic features.

[0149] The topology graph construction module is used for obtaining the node type through semantic recognition according to the node position, and constructing the track topology graph according to the node type.

[0150] The candidate path set generation and optimization module is used for generating the candidate path set through the track topology graph, and locally optimizing the task execution order and vehicle matching relationship of the candidate path set.

[0151] The feedback adjustment module is used for synchronously adjusting the path selection and task allocation based on the local optimization result, and outputting an integrated scheduling scheme.

[0152] The above formulas are all dimensionless numerical calculations, and the formulas are obtained by software simulation of a large amount of data to obtain a formula closest to the real situation. The preset parameters in the formula are set by a person skilled in the art according to the actual situation.

[0153] The above embodiments can be realized wholly or partially by software, hardware, firmware or any other combination. When realized by software, the above embodiments can be realized in the form of a computer program product wholly or partially.

[0154] Those skilled in the art can understand that the modules and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software manner depends on the specific application and design constraints of the technical solution. The skilled person can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0155] In addition, each functional module in each embodiment of the present application can be integrated in one processing module, or each module can exist physically alone, or two or more modules can be integrated in one module.

[0156] The above is merely specific embodiments of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art can easily think of changes or replacements within the technical scope disclosed by the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0157] Finally: the above is only the preferred embodiment of the present application, and is not used to limit the present application, and any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. An OHT track node automatic identification and path planning method, characterized in that, The method comprises the following steps: The node position is identified by a multi-mode verification method based on vehicle motion trajectory feature analysis, point cloud geometric clustering processing, and the combination of visual and magnetic features, specifically: Trajectory data is acquired and smoothed to obtain smoothed trajectory data; Based on the smoothed trajectory data, the speed and steering change rate are calculated, and the trajectory abnormal section is identified according to the speed and steering change rate; Taking the center point of the abnormal section as a reference, a fixed trajectory window is extended forward and backward, and the minimum steering change rate of the front and rear endpoints is taken as the boundary correction point to correct the start and end coordinates of the abnormal section to obtain a candidate section; The confidence score is obtained according to the maximum steering change rate, maximum speed, and number of abnormal points in the candidate section, and a judgment is made to obtain a core candidate node section; Based on the candidate node section, point cloud data within the range of each section is extracted and preprocessed; The neighborhood point set is determined according to the processed point cloud data, and the curvature is obtained, and the normal vector and curvature of each point cloud are labeled; The labeled point cloud is divided into several clustering blocks using a density clustering algorithm; The geometric features of each clustering block are analyzed to identify the track geometric features, and the clustering confidence is calculated; The candidate node clustering set is obtained according to the clustering confidence, and the spatial coordinate range is selected; According to the spatial coordinate range, the image sequence and magnetic field signal data of each candidate node are obtained; The image sequence is detected to identify the track key marks, and is aligned with the center point of the spatial coordinate range to obtain the visual verification feature; The magnetic flux change rate is obtained through the magnetic field signal data, and the magnetic anomaly points are marked; The continuous magnetic anomaly points are aggregated, and the center point corresponding to the trajectory coordinate at the center time is taken as the magnetic feature center point; The deviations of the geometric center points from the visual verification feature and the magnetic feature center point are obtained, and the node position is determined; The node type is identified according to the node position, and the track topology graph is constructed according to the node type; A candidate path set is generated through the track topology graph, and the task execution order and vehicle matching relationship of the candidate path set are locally optimized, specifically: An initial scheduling data set is obtained, and a time window is established for each task; For each task candidate path, the expected execution time sequence of the path is obtained, and is matched with the time window to determine whether the task candidate path is feasible in time; For the feasible path, overlap detection is performed to determine whether there is a conflict; For the path with a conflict, it is determined whether it is feasible according to the path feasibility score; Based on the time and conflict feasibility detection, the effective candidate path set is screened; Based on the effective candidate path set, the task dynamic priority is obtained, and the task execution sequence is generated; The task execution sequence is preliminarily matched with the candidate vehicle to obtain a task-vehicle mapping table, and a local optimization objective function is established; The local optimal task execution order and vehicle allocation scheme are obtained by solving the minimum local optimization objective function; Based on the local optimization result, the path selection and task allocation are adjusted synchronously, and an integrated scheduling scheme is output.

2. The OHT track node automatic identification and path planning method of claim 1, wherein, The node type is identified according to the node position, specifically: The geometric features and behavior features of the node are extracted according to the identified node position; The geometric features and the behavior features are normalized and fused to obtain a node feature comprehensive vector, wherein the geometric features are used to describe the spatial form of the node, and the behavior features are used to reflect the dynamic performance of the vehicle at the node; Based on the node feature comprehensive vector, input into a pre-trained node semantic classification model to determine the node type, and output the node type label.

3. The OHT track node automatic identification and path planning method of claim 1, wherein, The track topology graph is constructed according to the node type, specifically: Attribute information is assigned to each node; According to the node type and the spatial position, a candidate set of inter-node connections is generated; The reliability score of the candidate set of inter-node connections is calculated, which includes the deviation of the connection from the historical trajectory, the point cloud density, the consistency index of the track direction, and the node endpoint position deviation; High-reliability connections are selected through the reliability score, and the connection priority is adjusted according to the node type to generate an optimized connection set; The node set with assigned attribute information and the optimized connection set are combined to generate the track topology graph.

4. The OHT track node automatic identification and path planning method of claim 1, wherein, The candidate path set is generated through the track topology graph, specifically: The current node and the task target node of each vehicle are obtained and marked as the starting point and the ending point; Based on the track topology graph, an incremental graph search algorithm is used to search for feasible paths from the starting point to the ending point; The multiple feasible paths obtained by searching form a preliminary candidate path set; The path dynamic priority of each path in the preliminary candidate path set is calculated, including the reliability score of the path, the node load index, and the track resource occupancy rate; The preliminary candidate paths are sorted according to the path dynamic priority from high to low to obtain a sorted path priority queue, and the corresponding candidate path set of each vehicle is output.

5. The OHT track node automatic identification and path planning method of claim 1, wherein, Based on the local optimization result, the path selection and task allocation are adjusted synchronously to output an integrated scheduling scheme, specifically: The edge weight of the track topology graph is updated based on the local optimization result, and the task priority and vehicle matching strategy are real-time corrected according to the updated edge weight, including adjusting the task execution order according to the real-time congestion degree of the node, and locally re-matching the task execution order and vehicle allocation scheme to select the vehicle with the minimum path cost and meeting the task time window constraint; A new candidate path set is generated based on the updated track topology graph; The task execution order and vehicle matching relationship are re-optimized on the new candidate path set; Iterate until convergence, and output the integrated scheduling scheme.

6. A system for automatic identification and path planning using the OHT track node as claimed in any one of claims 1-5, characterized in that, It includes: A node position recognition module for recognizing track node positions based on vehicle motion trajectory feature analysis, point cloud geometric clustering processing, and multi-mode verification method combining visual and magnetic features; A topology graph construction module for obtaining node types through semantic recognition based on node positions, and constructing a track topology graph according to the node types; A candidate path set generation and optimization module for generating a candidate path set through the track topology graph, and locally optimizing the task execution order and vehicle matching relationship of the candidate path set; A feedback adjustment module for synchronously adjusting the path selection and task allocation based on the local optimization result to output an integrated scheduling scheme.

Images