Track traffic inspection task scheduling and data management platform

Abstract

The invention relates to the technical field of rail transit inspection, and discloses a rail transit inspection task scheduling and data management platform, which comprises a central server, at least one relay node and one or more groups of inspection equipment. The central server is configured with a central task distribution strategy, and a main task is split into sub-tasks and distributed to the relay nodes; the relay node is arranged on an inspection site, receives the subtasks, splits the subtasks into task units based on a node task splitting strategy, and allocates the task units to inspection equipment for execution; and the inspection equipment executes the task unit and returns inspection data, the relay node pre-inspects the data, qualified data is transmitted to the central server for complete comparison, and unqualified data is recorded. Through mechanisms such as two-stage task splitting, data pre-check and dynamic priority adjustment, efficient scheduling of inspection tasks and intelligent management of data are realized, the inspection efficiency and the fault early warning capability are improved, and resource utilization is optimized.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of rail transit inspection technology, and particularly relates to a rail transit inspection task scheduling and data management platform. BACKGROUND

[0002] With the acceleration of urbanization, the scale of urban rail transit systems is continuously expanding, and its operation safety and reliability are increasingly important. The safe operation of rail transit systems relies on regular inspection and timely maintenance to detect potential faults and abnormalities in infrastructure such as tracks, signal equipment, power supply systems, etc. However, traditional manual inspection methods have low efficiency, high omission rate, poor real-time performance, and other problems, making it difficult to meet the large-scale and complex inspection needs of modern rail transit.

[0003] However, existing inspection systems still have deficiencies in task scheduling and data management. For example, existing systems are difficult to intelligently allocate according to actual conditions and equipment status at the inspection site, which can lead to uneven task allocation and low resource utilization efficiency. Meanwhile, centralized data processing results in a heavy data processing burden. SUMMARY

[0004] The present application aims to overcome the above-mentioned deficiencies of the prior art and provide a rail transit inspection task scheduling and data management platform. Through mechanisms such as two-level task splitting, data pre-inspection, and dynamic priority adjustment, the platform achieves efficient scheduling of inspection tasks and intelligent management of data, improving inspection efficiency and fault warning capability, and optimizing resource utilization.

[0005] To achieve the above-mentioned application purposes, the rail transit inspection task scheduling and data management platform provided by the present application comprises a central server configured with a central task distribution strategy to split a main task into several sub-tasks; at least one relay node set at the inspection site, the relay node being in communication with the central server, the relay node receiving the sub-tasks allocated by the central server and splitting the sub-tasks into several task units based on a preset node task splitting strategy; one or more groups of inspection equipment, the inspection equipment being in communication with the relay node, receiving and executing the task units, the inspection equipment executing the task units and returning inspection data obtained from the inspection to the corresponding relay node, the relay node receiving the inspection data and performing data pre-inspection, defining the inspection data as pre-inspection qualified data if the inspection data meets the basic inspection requirements and transmitting the data to the central server for complete comparison, or defining the inspection data as pre-inspection unqualified data if the inspection data does not meet the basic inspection requirements and transmitting the data to the central server for recording.

[0006] Further, the central task distribution strategy comprises dividing the inspection range into a plurality of inspection areas, dividing the main task into a plurality of pending sub-tasks according to the geographical position attribute of the inspection range, each pending sub-task corresponding to a different inspection area, and each pending sub-task being associated with a set of spatial position parameters; obtaining the processing capacity and resource configuration of each relay node, matching each relay node with the pending sub-tasks, constructing a deployment score relationship, selecting an adaptive relay node and delivering the corresponding pending sub-task as the sub-task.

[0007] Further, the center task distribution strategy is configured with a preset priority adjustment sub-strategy, and the priority adjustment sub-strategy includes obtaining the historical fault records of the inspection, evaluating the risk index, and task timeliness requirement, sorting the execution priority of the sub-task to obtain a priority label, allocating the execution order of the sub-task based on the order of the priority label.

[0008] Further, the node task splitting strategy includes the relay node receiving the sub-task; obtaining node device information of the current relay node, the node device information including the number of available inspection devices, the inspection device configuration, and the state information of the inspection device of the current relay node; obtaining the operation area and the expected operation duration of each inspection device; evaluating the fault history density distribution of the current inspection area; sorting the spatial position, task duration, and perception modality requirements of the task unit, and delivering the task unit to the inspection device that meets the conditions for execution.

[0009] Further, the relay node is configured with a sample pre-check comparison module, which is used for sample comparison of the returned inspection data for pre-check screening, and the sample pre-check comparison module is configured with a local database, which stores a local sample set delivered by the center database. The received inspection data is compared with the same type of sample in the local sample set to obtain a similarity score, and it is judged whether it meets the basic inspection requirement based on the similarity score.

[0010] Further, the pre-check comparison module is configured with a key feature comparison strategy, and the key feature comparison strategy includes extracting unqualified high-frequency features based on historical inspection data, The inspection data is compared with the unqualified high-frequency features for similarity, and an unqualified feature similarity matching value is calculated. If the unqualified feature similarity matching value is lower than a first preset threshold, the inspection data is marked as pre-inspection normal; If the unqualified feature similarity matching value is higher than a second preset threshold, the inspection data is marked as pre-inspection unqualified; If the unqualified feature similarity matching value is between the first preset threshold and the second preset threshold, the inspection data is marked as to-be-rechecked data.

[0011] Further, the center server is configured with a data comparison strategy, which is used for hierarchical processing of pre-inspection qualified data uploaded by the relay node, including a basic verification step, comparing the pre-inspection qualified data with a preset device standard parameter library to verify whether the pre-inspection qualified data is in a safe operation interval; a trend verification step, extracting historical inspection data in a preset time interval to construct a trend model, and comparing the deviation rate of the current pre-inspection qualified data with the trend model; an abnormal feature extraction step, based on the comparison result of the basic verification step and the deviation analysis of the trend verification step, extracting potential abnormal features from the pre-inspection qualified data; a fault association step, calling a historical fault database to obtain a fault precursor feature, the fault precursor feature being specifically an associated feature with an occurrence rate higher than a preset occurrence threshold in a preset period before the fault occurs, and matching the potential abnormal features with the fault precursor feature; When the basic verification step finds that the parameter is out of limit, or the trend verification step finds that the trend is abnormal, or the fault association step finds a high association risk, a fault warning information is generated, and a key review task is pushed to the corresponding relay node, the key review task including a review area and a detection accuracy requirement, and at the same time, the priority of the next inspection task in the review area is improved.

[0012] Further, when the sample comparison module performs sample comparison, the number of sample comparisons is dynamically determined according to the historical pre-inspection error rate corresponding to the inspection area or task type. The higher the error rate, the more the initial number of comparison samples.

[0013] Further, the center server dynamically adjusts the receiving priority and processing priority of the pre-inspection qualified data from the relay node according to the current system load state. When the center server load is high, high-priority task data is preferentially received and processed, and low-priority task data is delayed.

[0014] Further, the relay node is configured with a task reassignment strategy, when the relay node detects that the inspection device is offline or does not respond for more than a preset waiting time, the inspection device is defined as an offline device and the task reassignment strategy is triggered, the task reassignment strategy comprises defining the task unit currently executed by the offline device as an offline task, evaluating the priority of the offline task, and reassigning the offline task and the unexecuted task unit based on the priority of the task unit, if the corresponding relay node cannot be reassigned, generating a task fallback signal to the central server and redefining a subtask.

[0015] Advantages: 1. Through the two-level task splitting strategy of the central server and the relay node, the main task is decomposed into appropriate task units, and intelligent allocation is performed according to the state and ability of the inspection device, realizing the optimal configuration of the inspection resource and significantly improving the inspection efficiency.

[0016] 2. The data pre-checking mechanism of the relay node can preliminarily filter the inspection data locally, timely find data that does not meet the basic requirements, reduce the transmission and processing of invalid data, and improve the data quality and accuracy of subsequent analysis. The central task distribution strategy and the node task splitting strategy comprehensively consider factors such as the geographical position of the inspection area, the processing capacity of the relay node, and the state of the inspection device, realizing the reasonable allocation and efficient utilization of resources.

[0017] 3. The data comparison strategy of the central server adopts a multi-level verification mechanism, combines historical fault data and trend analysis, can discover potential abnormal features in advance, and improves the timeliness and accuracy of fault warning. The task reassignment strategy and the dynamic priority adjustment mechanism enable the system to adapt to sudden situations such as the offline of the inspection device and the change of the task priority, guarantee the smooth execution of the inspection task, and enhance the flexibility and reliability of the system. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 is the system architecture schematic diagram of the rail transit inspection task scheduling and data management platform provided by the application.

[0019] Figure 2 is the process schematic diagram of the central task distribution strategy in the application.

[0020] Figure 3 is the process schematic diagram of the node task splitting strategy in the application.

[0021] Figure 4 is the process schematic diagram of the data comparison strategy in the application.

[0022] Figure 5This is a flowchart illustrating the task redistribution strategy in this invention. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] It should be noted that when a component is described as "fixed to" another component, it can be directly on the other component or may have a component in between. When a component is considered "connected to" another component, it can be directly connected to the other component or may have a component in between. When a component is considered "set on" another component, it can be directly set on the other component or may have a component in between. The terms "vertical," "horizontal," "left," "right," and similar expressions used in this document are for illustrative purposes only.

[0025] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0026] This application provides a rail transit inspection task scheduling and data management platform. Please refer to [link / reference]. Figures 1 to 5 This platform achieves precise scheduling of inspection tasks and end-to-end data management through hierarchical collaboration between a central server, relay nodes, and inspection equipment. The central server, as the core decision-making layer, is responsible for task breakdown and global coordination; relay nodes, as the on-site scheduling layer, undertake sub-tasks and allocate them to specific equipment; and inspection equipment, as the execution layer, completes the detection and transmits the data back.

[0027] The core function of the center server is to split the main task based on the center task distribution strategy, the main task refers to the complete inspection requirement covering a certain rail transit line such as subway line 1, and the subtask is an independent task unit split by area or equipment type such as track inspection in the interval between station 1 and station 3. The relay node is deployed at the inspection site such as the station control room or the interval signal room, and is connected with the center server through wireless communication, receives the subtask, generates smaller task units such as single track fastener detection tasks according to the node task splitting strategy, and issues them to the inspection equipment. The inspection equipment includes track robot vehicle detection devices and the like, executes the task unit, collects data through sensors such as high-definition cameras, infrared detectors, etc., and then transmits it back through the relay node.

[0028] The data flow follows the collection pre-inspection comparison process: the data returned by the inspection equipment is first subjected to data pre-inspection by the relay node, the basic inspection requirement is a pre-set judgment standard to reduce the data processing pressure of the center server, the detected inspection data is first subjected to preliminary judgment at the relay node, the pre-inspection qualified data meeting the requirement is uploaded to the center server for complete comparison, and the unqualified data such as missing data of fuzzy image parameters is recorded and marked by the center server for re-inspection. The basic inspection requirement includes: if the similarity of the inspection data to the key features of the same type of sample in the local sample set of the relay node is greater than or equal to the pre-set basic similarity threshold, it is defined as pre-inspection qualified data; otherwise, it is defined as pre-inspection unqualified data, and the pre-set basic similarity threshold is determined based on the historical normal sample statistics of the inspection area.

[0029] The center task distribution strategy includes The inspection range is divided into a plurality of inspection areas, the main task is divided into a plurality of pending subtasks according to the geographical location attribute of the inspection range, each pending subtask corresponds to a different inspection area, and each pending subtask is associated with a group of spatial location parameters. The processing capacity and resource configuration of each relay node are obtained, each relay node is matched with the pending subtask, a deployment score relationship is constructed, and the adaptive relay node is selected and transported to the corresponding pending subtask as the subtask.

[0030] The core of the center task distribution is the accurate allocation based on the spatial attribute and resource adaptation. First, the inspection area is divided according to the geographical location attribute of the inspection range, for example, the subway line can be divided into three levels according to the station interval equipment type, such as the inspection area of the interval between station A and station B on subway line 2, and the interval between station B and station C, each area is associated with spatial location parameters such as the location coordinates in the station, which can be set in advance according to the distribution of station facilities.

[0031] The relay node matching needs to comprehensively process the capability and resource configuration. The processing capability refers to the data forwarding rate of the relay node, such as 100 pieces of data per second, and the number of concurrent connections, such as 20 pieces of inspection equipment at the same time. The resource configuration includes storage capacity, such as 100 GB of local cache data, and computing power, such as supporting 50 times of image comparison per second. The deployment score relationship is calculated by weighting three indexes. The position matching degree is 0 to 10 points. The closer the distance, the higher the score. The processing capability adaptation degree is 0 to 10 points. The higher the matching degree of the capability and the task demand, the higher the score. The load redundancy is 0 to 10 points. The lower the current load, the higher the score. The total score is equal to the position matching degree multiplied by the position matching degree weight, plus the processing capability adaptation degree multiplied by the processing capability weight, plus the load redundancy multiplied by the load weight. The subtask is assigned to the relay node with the highest score.

[0032] The priority adjustment sub-strategy is used for dynamically sorting subtasks. The priority adjustment sub-strategy includes obtaining the historical fault record of the inspection, evaluating the risk index, and task timeliness requirement; sorting the execution priority of the subtask to obtain a priority label; assigning the execution order of the subtask based on the order of the priority label.

[0033] Through the historical fault record, such as 5 times of rail fastener loosening in a certain interval in the past 3 months; the risk index is evaluated based on the importance of the equipment, such as the risk index of the power supply equipment is 8.5 / 10; the task timeliness requirement, such as the task timeliness weight before the morning and evening peak is calculated as 1.2 to calculate the priority. Here, a weighted sum model is used, for example, the priority of a certain subtask is calculated as: The historical fault record corresponds to a weight of 0.3, and 5 times of failure corresponds to a score of 8; The evaluation risk index corresponds to a weight of 0.4, and the evaluation risk index 8.5 corresponds to a score of 8.5; The task timeliness requirement corresponds to a weight of 0.3, and the task before the peak corresponds to a score of 9; The final priority = 8*0.3+8.5*0.4+9*0.3=8.5, and the execution order is assigned according to the score.

[0034] The node task splitting strategy includes The relay node receives the subtask; obtaining the node equipment information of the current relay node, the node equipment information including the number of available inspection equipment of the current relay node, the inspection equipment configuration and the state information of the inspection equipment; obtaining the operation area and the expected work duration of each inspection equipment; evaluating the fault history density distribution of the current inspection area; The spatial position, task duration and perception modality requirement of the task unit are sorted, and the task unit is dispatched to the inspection device that meets the condition for execution.

[0035] After receiving the sub-tasks, the relay node needs to implement fine allocation in combination with the state of the inspection device. The number of available inspection devices in the node device information indicates the number of devices currently in online and idle state. The inspection device configuration includes the type of sensor carried, such as whether to carry a laser radar, an industrial camera, and a moving speed such as a track robot speed of 3 km / h. The state information includes the remaining power and the current position.

[0036] The operating area refers to the range that can be covered by the device, such as a robot that can only work in the interval from station 2 to station 3. The estimated operation time is based on historical data, such as 100 fasteners taking about 15 minutes to detect. The fault history density distribution is calculated by counting the frequency of faults in a certain area in the past 6 months, such as an average of 3 fastener failures per kilometer of track. The higher the density, the more detection resources are allocated to the area.

[0037] The perception modality requirement of the task unit refers to the detection method required to complete the task, such as visual detection for appearance defects and infrared detection for temperature abnormalities. The relay node matches the device configuration according to the configuration matching model. For example, visual detection modality is required to detect track fastener looseness, and only robots equipped with high-definition cameras are allocated this task. Infrared modality is required to detect cable temperature, and only devices equipped with infrared sensors are allocated.

[0038] The relay node is configured with a sample pre-check comparison module for sample comparison of the returned inspection data for pre-check screening. The sample pre-check comparison module is configured with a local database that stores a local sample set dispatched by the central database. The received inspection data is compared with the same type of sample in the local sample set to obtain a similarity score, and whether it meets the basic inspection requirement is judged based on the similarity score.

[0039] The sample pre-check comparison module of the relay node realizes rapid screening through the local database. The local sample set is synchronized by the central database every day at 3 am, containing common qualified samples such as normal fastener image standard temperature range and unqualified samples such as loose fastener image over-temperature data in this area. Related local sample sets can also be sent synchronously when allocating sub-tasks.

[0040] The pre-check comparison module is configured with a key feature comparison strategy, which includes extracting unqualified high-frequency features based on historical inspection data, The inspection data is compared with the high-frequency non-conforming features to calculate the non-conforming feature similarity matching value. If the non-conforming feature similarity matching value is lower than a preset first preset threshold, the inspection data is marked as normal in the pre-inspection. If the similarity matching value of the non-conforming feature is higher than the second preset threshold, then the inspection data is marked as a pre-inspection non-conforming data. If the similarity matching value of the non-conforming feature is between the first preset threshold and the second preset threshold, then the inspection data is marked as data to be re-inspected.

[0041] The key feature comparison strategy focuses on high-frequency non-conforming features. These features are extracted from historical data and refer to defects that are prone to occur frequently during long-term track use. The system extracts key features from these defects, including shape, size, and location, as standard templates for comparison. Examples of high-frequency non-conforming features include: Rail cracks: a high-frequency defect characterized by linear or irregular cracks on the surface of the rail, usually longer than 5 mm, with obvious fracture marks at the edges; Loose / missing fasteners: Track fixing fasteners such as bolts and elastic clips may become loose due to vibration, manifested as a gap of >2mm between them and the sleepers, or complete absence, which is a high-frequency safety hazard; Track bed compaction: Track bed gravel becomes compacted due to rainwater erosion and soil infiltration. It is characterized by a seamless surface, loss of elasticity, and continuous blocky appearance in the image. Track gauge exceeds standard: The distance between two rails exceeds the standard range. For example, if the standard track gauge is 1435mm, the actual deviation is >3mm. The deviation value characteristics can be extracted through sensor data.

[0042] The relay node will compare new inspection data, such as images of a certain section of track and track gauge sensor data, with the standard template of the aforementioned non-compliant high-frequency features, and calculate the matching value. The matching value is assumed to range from 0 to 100, where 0 represents complete dissimilarity and 100 represents a perfect match. If the rail surface in the new data is smooth and without cracks, the matching value with the "rail crack" feature may only be 10. If the gap between the fastener and the sleeper in the new data reaches 3mm, and the position matches the "fastener loose" template highly, the matching value may reach 80, showing a high degree of similarity.

[0043] Then, threshold division and labeling rules are applied. Assuming the first preset threshold is 30 and the second preset threshold is 70, the matching values ​​are divided into three intervals using these two thresholds to achieve automated classification. Match value < first preset threshold 30: marked as "pre-detection normal".

[0044] Example: In the new data, the track gauge deviation is 1mm, which is within the standard range, and the matching value with the "track gauge exceeds the standard" feature is 20 < 30, which is determined to be normal and does not require manual intervention.

[0045] Matching value > second preset threshold 70: marked as "pre-inspection unqualified".

[0046] Example: In the new data, an 8mm long crack appears on the surface of the rail, and the matching value with the "rail crack" feature is 85 > 70, which is determined to be a suspected high-frequency defect and is directly pushed to the maintenance system for priority investigation.

[0047] First preset threshold 30 ≤ matching value ≤ second preset threshold 70: marked as "to be re-inspected" Example: In the new data, the gap between the fastener and the sleeper is 1.5mm, close to the loose critical value of 2mm, and the matching value with the "fastener loosening" feature is 50, between 30-70. Due to the uncertainty, manual review is required to confirm whether it is a defect.

[0048] The relay node serves as an intermediate link in data processing, and completes similarity comparison locally, which can quickly screen out highly suspected high-frequency defects and potential defects that require manual confirmation, avoiding uploading all data without distinction to the terminal, and greatly improving the investigation efficiency of high-frequency defects. For example, there is no need for manual inspection of inspection images, the system automatically marks 80% of normal data and 10% of highly suspected defects, and manual inspection only needs to focus on 10% of the data to be re-inspected, which not only reduces the workload, but also prioritizes the processing of high-frequency risks. In this way, track inspection is upgraded from full manual investigation to system pre-screening + accurate investigation, especially the response speed of high-frequency defects can be improved several times, ensuring track safety.

[0049] The center server is configured with a data comparison strategy, and the data comparison strategy of the center server realizes deep analysis through four-level verification. The data comparison strategy is used for hierarchical processing of pre-inspection qualified data uploaded by the relay node, including a basic verification step, a trend verification step, an abnormal feature extraction step, and a fault association step.

[0050] The pre-inspection qualified data is compared with the preset equipment standard parameter library to verify whether the pre-inspection qualified data is in the safe operation interval; the basic verification step compares the data with the equipment standard parameter library, which contains the safety threshold of each equipment, such as rail top surface wear ≤ 0.3mm, fastener torque 35 to 40N・m.

[0051] The trend model is constructed by extracting historical inspection data in a preset time interval, and the deviation rate of the current pre-inspection qualified data is compared with the trend model. The trend verification step extracts historical data of the last three months to construct a trend model. A linear regression algorithm is used, and the linear regression model formula is y=kx+b, where y is the predicted value, x is the time variable, k is the slope, and b is the intercept. For example, the trend model of the interval track settlement data shows that the average settlement is 0.02 mm per week. If the current data shows that the single-week settlement is 0.1 mm, the deviation rate is equal to 0.1 minus 0.02 divided by 0.02 multiplied by 100%, which is equal to 400%. It is determined that the trend is abnormal.

[0052] Based on the comparison results of the basic verification step and the deviation analysis of the trend verification step, potential abnormal features are extracted from the pre-inspection qualified data. The abnormal feature extraction step mines potential risks from qualified data using an abnormal feature recognition model. For example, although the data is within the standard of 35 to 40 N・m for the fastener torque of 38 N・m in the safe interval, the trend verification shows that it has decreased from 40 N・m to 38 N・m in the last three weeks, showing a downward trend. Therefore, the continuous decrease in torque is extracted as a potential abnormal feature.

[0053] The historical fault database is called to obtain fault precursor features, which are specific to the occurrence rate of associated features being higher than a preset occurrence threshold in a preset period before the fault occurs. The potential abnormal features are matched with the fault precursor features. The fault association step matches the precursor features through the historical fault database, and uses a feature matching model. The fault precursor feature is a typical early warning signal before the fault occurs, such as the precursor feature of track cracking, which includes three consecutive detections of decreasing fastener torque and gradually increasing track gauge deviation. The occurrence rate of such features in historical data is 70%, and the preset occurrence threshold is 50%. If the matching degree of the potential abnormal features and the precursor features reaches 85% higher than the 80% threshold, it is determined that there is a high correlation risk, triggering a fault warning. The matching degree is calculated using a cosine similarity model, and the formula is cos θ=(A・B) / (|A|・|B|), where A is the potential abnormal feature vector and B is the fault precursor feature vector.

[0054] When the sample pre-inspection comparison module performs sample comparison, the historical pre-inspection error rate corresponding to the inspection area or task type is used to dynamically determine the number of sample comparisons. The higher the error rate, the more the initial number of samples. This mechanism dynamically associates error rate and sample size, solving the problem of low efficiency in low-risk scenarios and insufficient accuracy in high-risk scenarios caused by fixed sample size: it avoids excessive comparison of low-error-rate areas, reduces relay node computing resource consumption, and increases the number of samples in high-error-rate areas to improve pre-inspection reliability, optimizing resource utilization and improving data quality.

[0055] When the basic verification step finds that the parameters are out of limits, or the trend verification step finds that the trend is abnormal, or the fault association step finds that the association risk is high, a fault warning message is generated, and a key review task is pushed to the corresponding relay node, the key review task including a review area and a detection accuracy requirement, and at the same time, the priority of the next inspection task in the review area is increased.

[0056] The center server adjusts the data processing priority according to the load state. The center server dynamically adjusts the receiving priority and processing priority of the pre-inspection qualified data from the relay node according to the current system load state, and when the center server load is high, high-priority task data is preferentially received and processed, and low-priority task data is delayed. A load priority adjustment model is adopted, and the load is high when the CPU occupancy rate is > 80% or the memory usage rate is > 75%, at which time high-priority task data such as region data with a risk index of 8 points or more is preferentially processed, and low-priority data such as region data with a risk index of 3 points or less is delayed. The delay time is calculated as delay time seconds equal to the current CPU occupancy rate minus 80% multiplied by 10, for example, when the CPU occupancy rate is 85%, the delay is 5 seconds.

[0057] The relay node is configured with a task reallocation strategy, when the relay node detects that the inspection device is offline or does not respond for more than a preset waiting time, the inspection device is defined as an offline device and the task reallocation strategy is triggered, the task reallocation strategy includes defining the task unit currently executed by the offline device as an offline task, evaluating the priority of the offline task, and re-allocating the offline task and unexecuted task units based on the priority of the task unit, if the corresponding relay node cannot be reallocated, a task rollback signal is generated to the center server and a sub-task is redefined.

[0058] The task reallocation strategy ensures the continuity of the task in the event of a sudden situation, and a task reallocation model is adopted, when the inspection device is offline and the communication is interrupted for more than 30 seconds or does not respond for more than 10 minutes of the expected completion time, the relay node marks it as an offline device. After evaluating the priority of the offline task, it is re-allocated, for example, detecting a turnout point rail as a high-priority task with a priority of 8 points, and preferentially allocating it to a device with a remaining power > 60% and a current load < 30%. If the load of all available devices is > 70%, a task rollback signal is sent to the center server, and the center server integrates the task into the sub-task of the adjacent region and re-allocates it.

[0059] The following is combined with actual application scenarios: The morning inspection of Metro Line 2 in a certain city is taken as an example to illustrate the platform operation process. The line is 15 kilometers long with 10 stations, and the comprehensive inspection of track structure, power supply equipment and communication facilities needs to be completed. The main task covering the whole line is split into 13 sub-tasks by the center server according to "3 intervals + 10 stations", such as "station 5 to station 8 interval inspection" as one of the sub-tasks. Through the deployment scoring mechanism: position matching degree, processing capacity adaptation degree, load redundancy degree weighted calculation, the sub-task matches the station 6 relay node, with a total score of 8.2, and the sub-task is issued to it.

[0060] After receiving the sub-task, the station 6 relay node processes based on the node task splitting strategy: the interval is equipped with 3 track robots (A, B, C), among which A is equipped with a high-definition camera and a laser radar, B is equipped with an infrared sensor, and C is a backup device. Combined with the fault history density, the fastener failure rate of station 6 to station 7 interval is high, 4 times per kilometer, and the 20 fastener detection tasks in this area are allocated to the state optimal A with remaining power 90%, load 10%, and the power cable temperature detection task is allocated to B.

[0061] When A performs the fastener detection task, it collects images through the high-definition camera and transmits them back to the relay node. The sample pre-inspection comparison module of the relay node calls the local sample set, which contains 50 normal fastener image samples and 30 typical loose fastener image samples in the interval, and compares the key features of the returned images with the samples, such as the similarity of fastener bolt position and spring strip angle: 3 images of fastener outline and local sample set "unqualified sample" such as bolt missing sample with high frequency feature matching value of 85 points, pre-set second threshold value of 80 points, marked as "pre-inspection unqualified", only recorded to the center server; 7 images of fastener features and local sample set matching value is 40 points, between first threshold value 20 points and second threshold value 80 points, marked as "to be re-inspected", the relay node pushes the re-inspection instruction to A to clearly need to focus on shooting the bolt and spring strip connection part; The matching value of the remaining 10 image features and the "qualified sample" in the local sample set is 15 points, which is lower than the first threshold value of 20 points, that is, the similarity with unqualified features is very low, marked as "pre-inspection qualified", transmitted to the center server.

[0062] After receiving the re-inspection instruction, A re-collects images for the area to be re-inspected, and after returning, the matching value is reduced to 18 points after comparison by the relay node, which meets the qualified standard and is determined as pre-inspection qualified and uploaded.

[0063] When the central server compares the qualified data, it finds that the track fastener torque data near station 7 is in the safe interval 36 N·m, but the trend model shows that the deviation rate has been continuously decreasing from 39 N·m for nearly 3 weeks, with a deviation rate of 7.7%. It is extracted as a potential abnormal feature. The fault correlation step matches the pre-feature matching degree of 82% from track fastener loosening to track cracking. A fault warning is generated, and the relay node pushes the fastener key review task from station 7 to station 8 interval to the relay node, requiring the detection accuracy to be improved to ±1 N·m, and the priority of the afternoon inspection task in this area is increased from 5 minutes to 8 minutes.

[0064] During operation, robot A suddenly fails and the battery is depleted. The priority of the 5 unfinished fastener detection tasks of robot A is 7, which is redistributed by the relay node. Using the task allocation model, it is found that robot C has 70% of the remaining battery and a load of 20%, which meets the allocation conditions. Robot C takes over the task and completes it within 15 minutes, ensuring that all inspections are completed before the morning operation.

Claims

1. A rail transit inspection task scheduling and data management platform, characterized in that, The utility model relates to a kind of distributed task allocation method and system, including A center server is configured with center task distribution strategy to split main task into several subtasks; At least one relay node is set in the inspection site, and the relay node communicates with the center server, receives the subtask distributed by the center server, and splits the subtask into several task units based on the preset node task splitting strategy; One or more groups of inspection equipment communicate with the relay node, receive and execute the task unit, and the inspection equipment executes the task unit and returns the inspection data obtained by inspection to the corresponding relay node, and the relay node receives the inspection data and carries out data pre-checking, and if the inspection data meets the basic inspection requirements, the inspection data is defined as pre-check qualified data and transmitted to the center server for complete comparison;If the inspection data does not meet the basic inspection requirements, the inspection data is defined as pre-check unqualified data and transmitted to the center server for recording. 2.The rail transit inspection task scheduling and data management platform according to claim 1, characterized in that, The center task distribution strategy includes Divide the inspection range into several inspection areas, divide the main task into several pending subtasks according to the geographical location attribute of the inspection range, each pending subtask corresponds to a different inspection area, and each pending subtask is associated with a group of spatial location parameters; Obtain the processing capacity and resource configuration of each relay node, match each relay node with the pending subtask, construct a deployment score relationship, select the appropriate relay node and deliver the corresponding pending subtask as the subtask. 3.The rail transit inspection task scheduling and data management platform according to claim 2, characterized in that, The center task distribution strategy is configured with a preset priority adjustment sub-strategy, which includes Obtain the historical fault record of the inspection, evaluate the risk index, and task time limit for rescission, Sort the execution priority of the subtask to obtain a priority label, Assign the execution order of the subtask based on the order of the priority label.

4. The rail transit inspection task scheduling and data management platform according to claim 1, characterized in that, The node task splitting strategy includes The relay node receives the subtask; Obtain the node device information of the current relay node, including the number of available inspection equipment, inspection equipment configuration and state information of the inspection equipment of the current relay node; Obtain the operating area and expected operation time of each inspection equipment; Evaluate the fault history density distribution of the current inspection area; Organize the spatial location, task duration and perception mode requirements of the task unit, and distribute the task unit to the inspection equipment that meets the conditions for execution.

5. The rail transit inspection task scheduling and data management platform according to claim 1, characterized in that, The relay node is configured with a sample pre-check comparison module for sample comparison of the returned inspection data for pre-check screening, and the sample pre-check comparison module is configured with a local database storing a local sample set distributed by the center database, and the received inspection data is compared with the same type of sample in the local sample set to obtain a similarity score, and whether it meets the basic inspection requirements is judged based on the similarity score. 6.The rail transit inspection task scheduling and data management platform according to claim 5, characterized in that, The pre-check comparison module is configured with a key feature comparison strategy, which includes Extract unqualified high-frequency features based on historical inspection data, The inspection data is compared with the unqualified high-frequency features in similarity, and an unqualified feature similarity matching value is calculated. If the unqualified feature similarity matching value is lower than a first preset threshold, the inspection data is marked as pre-inspection normal; If the unqualified feature similarity matching value is higher than a second preset threshold, the inspection data is marked as pre-inspection unqualified; If the unqualified feature similarity matching value is between the first preset threshold and the second preset threshold, the inspection data is marked as to-be-rechecked data.

7. The rail transit inspection task scheduling and data management platform according to claim 1, characterized in that, The center server is configured with a data comparison strategy, which is used for hierarchical processing of pre-inspection qualified data uploaded by the relay node, including a basic verification step, comparing the pre-inspection qualified data with a preset device standard parameter library to verify whether the pre-inspection qualified data is in a safe operation interval; a trend verification step, extracting historical inspection data in a preset time interval to construct a trend model, and comparing the deviation rate of the current pre-inspection qualified data with the trend model; an abnormal feature extraction step, based on the comparison result of the basic verification step and the deviation analysis of the trend verification step, extracting potential abnormal features from the pre-inspection qualified data; a fault association step, calling a historical fault database to obtain a fault precursor feature, the fault precursor feature being specifically an associated feature with an occurrence rate higher than a preset occurrence threshold in a preset period before the fault occurs, and matching the potential abnormal features with the fault precursor features; When the basic verification step finds that the parameters are out of limits, or the trend verification step finds that the trend is abnormal, or the fault association step finds a high association risk, a fault warning information is generated, and a key review task is pushed to the corresponding relay node, the key review task including a review area and a detection accuracy requirement, and at the same time, the priority of the next inspection task in the review area is improved. 8.The rail transit inspection task scheduling and data management platform of claim 5, characterized in that, When the sample pre-inspection comparison module performs sample comparison, the number of sample comparisons is dynamically determined according to the historical pre-inspection error rate corresponding to the inspection area or task type. The higher the error rate, the more the initial number of comparison samples.

9. The rail transit inspection task scheduling and data management platform according to claim 1, characterized in that, The center server dynamically adjusts the receiving priority and processing priority of the pre-inspection qualified data from the relay node according to the current system load state. When the center server load is high, high-priority task data is preferentially received and processed, and low-priority task data is delayed.

10. The rail transit inspection task scheduling and data management platform according to claim 1, characterized in that, The relay node is configured with a task reallocation strategy. When the relay node detects that the inspection device is offline or does not respond for more than a preset waiting time, the inspection device is defined as an offline device and the task reallocation strategy is triggered. The task reallocation strategy includes defining the task unit currently executed by the offline device as an offline task, evaluating the priority of the offline task, and reallocating the offline task and unexecuted task units based on the priority of the task unit, if the corresponding relay node cannot be reallocated, a task rollback signal is generated to the center server and a sub-task is redefined.

Images