Rail transit intelligent inspection robot machine control method based on digital twinning

Abstract

The invention discloses a rail transit intelligent inspection robot machine control method based on digital twinning. The method comprises the steps that a vertical rail digital twinning model is constructed based on collected rail transit data; generating an inspection path based on an A * algorithm, and obtaining a global inspection path of the inspection robot in combination with a real-time obstacle avoidance trajectory output by the digital twin model; dividing routing inspection priorities, carrying out dynamic routing inspection path allocation, and adjusting routing inspection paths when detecting path conflicts; simulating to obtain the use frequency of each track section, starting a reinforced inspection mechanism for the track with high use frequency, and executing cross validation inspection; and on the basis of the collected data, performing anomaly detection by comparing the data of the fault feature library, and triggering an anomaly alarm when an anomaly is detected. The method has the advantages that the inspection path is optimized through the digital twinborn technology and the A * algorithm, the inspection task is intelligently distributed in combination with real-time data and dynamic scheduling, and the rail transit inspection efficiency, reliability and safety are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to path planning, in particular to a rail transit intelligent inspection robot machine control method based on digital twinning. BACKGROUND

[0002] With the continuous expansion of the rail transit network, intelligent inspection robots can realize remote monitoring and real-time data uploading, enabling maintenance personnel to keep abreast of the running state of the facilities at any time. Intelligent inspection robots not only improve inspection efficiency and extend the service life of facilities, but also reduce operating costs and manpower investment, and are widely used in rail transit systems such as subways, light rails, high-speed railways, etc., promoting the intelligent development of the rail transit industry.

[0003] Current rail transit intelligent inspection robot machine control methods on the market usually lack the integration of comprehensive track geometry, kinematics and environmental dynamic models, resulting in poor flexibility in path planning and task allocation. Most methods rely on static preset paths or simple local path planning, failing to fully consider real-time environmental changes and dynamic obstacles of the track, and cannot effectively avoid sudden path conflicts or obstacles. The optimization of inspection paths usually lacks fine global scheduling mechanisms, and the task allocation and path adjustment of multi-robot collaborative work are relatively rough, which may lead to low efficiency and resource waste. In addition, existing methods are not highly intelligent in terms of track usage frequency analysis and inspection priority division, making it difficult to conduct differentiated inspection according to the usage of different track sections, especially in the reinforcement inspection mechanism of high-frequency tracks, which is weak and difficult to achieve precise inspection coverage and efficient fault warning. SUMMARY

[0004] To improve existing methods, a rail transit intelligent inspection robot machine control method based on digital twinning is provided, which optimizes inspection paths through digital twinning technology and A* algorithm, and intelligently allocates inspection tasks by combining real-time data and dynamic scheduling, improving rail transit inspection efficiency, reliability and safety.

[0005] To achieve the above purposes, the technical solution adopted by the present application is: The rail transit intelligent inspection robot machine control method based on digital twinning includes: Based on the collected rail transit data, a rail transit geometry model, a kinematics model and an environmental dynamic model are constructed, and a rail digital twinning model is established to generate running simulation data containing track data and environmental data; Based on the A* algorithm, an inspection path is generated, and combined with the real-time obstacle avoidance trajectory output by the digital twinning model, a global inspection path of the inspection robot is obtained; The inspection priority is divided based on the real-time positioning data of each inspection robot and the task amount of the track section to be inspected, and the dynamic inspection path is allocated. When a path conflict is detected, the inspection path is dynamically adjusted based on the inspection priority; Based on the use frequency of each track section simulated by the track digital twin model, a frequency threshold is set, and the enhanced inspection mechanism is started for the track with high use frequency, and multiple inspection robots are dispatched to perform cross-validation inspection; Based on the collected track geometric parameters, surface defects and environmental data, abnormality detection is performed by comparing with the fault feature library data, and an abnormality alarm is triggered when an abnormality is detected.

[0006] Preferably, the track transportation data collected is used to construct a track transportation geometric model, a kinematic model and an environmental dynamic model, and a track digital twin model is established to generate running simulation data containing track data and environmental data, which specifically includes: Based on the collected track geometric data, a track transportation geometric model is established, including a track plane model, a track longitudinal section model and a track transverse section model; Based on the principles of physics, a kinematic model of the train is constructed, including a dynamics equation, a vehicle speed and an acceleration; Based on the collected environmental related data and historical meteorological data, an environmental dynamic model is established; The track transportation geometric model, the kinematic model and the environmental dynamic model are integrated to generate a track digital twin model, and real-time collected data is input into the model for digital twin simulation; Based on the digital twin simulation results of the track digital twin model, track running simulation data containing track data and environmental data is generated.

[0007] Preferably, the inspection path is generated based on the A* algorithm, and the real-time obstacle avoidance trajectory output by the digital twin model is combined to obtain the global inspection path of the inspection robot, which specifically includes: Based on the track digital twin model, the inspection target area is determined, and the inspection area is divided into grids to form a two-dimensional grid map; Each grid in the grid map corresponds to a specific spatial area of the inspection target area, and the corresponding cost is given according to the track number density in the spatial area; Based on the A* algorithm, a heuristic function is designed, and the Euclidean distance from the starting position of the inspection task of each inspection robot to the target position is taken as the heuristic value to obtain the estimated shortest distance; Based on the real-time obstacle data in the environmental data in the track digital twin model, the inspection robot is simulated to recognize through sensors by the track digital twin model, and the original estimated shortest distance path is adjusted by combining the A* algorithm to generate a new path through local path planning; The local paths of the track sections are aggregated to generate global path planning data.

[0008] Preferably, the inspection priority is divided based on the real-time positioning data of each inspection robot and the task amount of each track section to be inspected, and the dynamic inspection path is allocated. When a path conflict is detected, the inspection path is dynamically adjusted based on the inspection priority, and the specific method comprises: The real-time positioning data of each inspection robot and the task amount of each track section to be inspected are obtained, and the task amount includes the length of the track section, the equipment condition, the maintenance requirement, and the historical fault record. The weight is divided based on the real-time positioning data of each inspection robot and the task amount of each track section to be inspected, an inspection priority scoring model is constructed, and the inspection priority of each inspection task is obtained according to the scoring result of each inspection task. Based on the global path planning data simulated by the track digital twin model, the inspection priorities of each inspection task are combined with the task amount, and the inspection tasks are allocated to each inspection robot. After the path allocation, it is detected whether there is a path conflict. If there is, the conflicting tasks are adjusted according to the priority of the inspection task. If the priorities are the same, the local path is re-planned.

[0009] Preferably, the usage frequency of each track section is simulated based on the track digital twin model, and a frequency threshold is set. The enhanced inspection mechanism is started for the track with high usage frequency, and multiple inspection robots are dispatched to perform cross-validation inspection, and the specific method comprises: Based on the simulation data of the track digital twin model, the historical usage frequency and the usage frequency in the future time period of each track section are obtained. Based on the historical usage frequency of each track section, a track usage frequency threshold is set, and each track section is divided into a low usage frequency section and a high usage frequency section. Based on the low usage frequency section, a regular inspection is performed. Based on the high usage frequency section, the number of inspections is increased, and multiple inspection robots are dispatched to perform task allocation for the same high usage frequency section. Multiple robots simultaneously perform inspection tasks on different high-frequency track sections to work cooperatively and perform cross-validation inspection.

[0010] Preferably, based on the collected track geometric parameters, surface defects, and environmental data, abnormality detection is performed by comparing with the fault feature library data, and an abnormality alarm is triggered when an abnormality is detected, and the specific method comprises: Based on the track geometric parameters, surface defects, and environmental data obtained by inspection and collection, data preprocessing is performed, and data features are extracted. Based on the historical fault data and track defect features, the fault feature library is constructed, the data features of the real-time collected data are compared with the data features in the fault feature library, and the fault matching probability distribution data is obtained. Based on the fault matching probability distribution data, an abnormality is judged, and the type of abnormality, including local fault and global fault, is judged. Based on the judgment of abnormality, a multi-level abnormality alarm is triggered, and real-time abnormality data is reported.

[0011] Compared with the prior art, the advantages of the present application are: By integrating track geometry model, kinematics model, environment dynamic model and real-time data, an efficient and accurate track inspection solution is provided. Through A* algorithm optimization of inspection path, combined with real-time obstacle avoidance technology, the inspection robot path can be dynamically adjusted to effectively avoid path conflict and improve inspection efficiency. In addition, this method intelligently allocates inspection tasks based on the frequency of track use, and starts the reinforcement inspection mechanism to ensure that high-frequency tracks are checked first. Through the cooperative work of multiple robots, cross-validation inspection is performed to further improve the reliability of inspection. Combined with the fault feature library for abnormality detection, potential problems of the track can be quickly identified and an alarm is triggered to ensure early fault warning and timely maintenance. BRIEF DESCRIPTION OF DRAWINGS

[0012] Figure 1 The method proposed by the present application is shown in the schematic diagram; Figure 2 The track digital twin model proposed by the present application is shown in the schematic diagram; Figure 3 The global inspection path proposed by the present application is shown in the schematic diagram; Figure 4 The inspection path optimization proposed by the present application is shown in the schematic diagram; Figure 5 The reinforcement inspection mechanism proposed by the present application is shown in the schematic diagram; Figure 6 The abnormality alarm proposed by the present application is shown in the schematic diagram; Figure 7 The architecture of the electronic device in the present scheme is shown in the schematic diagram; Figure 8 The structure of the computer readable storage medium in the present scheme is shown in the schematic diagram. DETAILED DESCRIPTION

[0013] The following description is used to disclose the present application so that those skilled in the art can implement the present application. The preferred embodiments in the following description are only as examples, and other obvious modifications can be made by those skilled in the art.

[0014] Referring to Figure 1 As shown, the track transportation intelligent inspection robot machine control method based on digital twin includes: Step 1: Based on the collected rail transit data, construct the rail transit geometric model, kinematic model and environmental dynamic model, and establish a digital twin model of the track to generate operational simulation data containing track data and environmental data; Step 2: Generate the inspection path based on the A* algorithm, and combine it with the real-time obstacle avoidance trajectory output by the digital twin model to obtain the global inspection path of the inspection robot; Step 3: Based on the real-time positioning data of each inspection robot and the workload of the track segment to be inspected, the inspection priority is divided, and the inspection path is dynamically allocated. When a path conflict is detected, the inspection path is dynamically adjusted based on the inspection priority. Step 4: Based on the simulated usage frequency of each track segment using the digital twin model of the track, and setting frequency thresholds, activate the enhanced inspection mechanism for tracks with high usage frequency, and dispatch multiple inspection robots to perform cross-validation inspections. Step 5: Based on the collected track geometry parameters, surface defects and environmental data, anomaly detection is performed by comparing with the fault feature database data, and an anomaly alarm is triggered when an anomaly is detected.

[0015] See Figure 2 As shown, based on the collected rail transit data, a geometric model, kinematic model, and environmental dynamic model of the rail transit system are constructed, and a digital twin model of the track is established to generate operational simulation data containing track data and environmental data. Specifically, this includes: Based on the collected track geometry data, a track transportation geometric model is established, including a track plane model, a track longitudinal profile model, and a track cross profile model. Based on the principles of physics, a kinematic model of a train is constructed, including the dynamic equations, speed, and acceleration. Based on collected environmental data and historical meteorological data, an environmental dynamic model is established. Based on the above-mentioned geometric model, kinematic model and environmental dynamic model of rail transit, a digital twin model of the track is generated, and the real-time collected data is input into the model for digital twin simulation. Based on the digital twin simulation results of the track digital twin model, track operation simulation data containing track data and environmental data is generated.

[0016] Specifically, the track plan model reflects the planar layout of the track, and describes the curvature of the track planar curve through the curve radius; the track longitudinal profile model describes the longitudinal gradient change of the track, that is, the elevation change of the track, and is established by collecting track height data and using a piecewise linear interpolation method; the cross profile model describes the cross-sectional shape of the track, which usually includes the track width and the shape of the roadbed, and obtains the cross-sectional shape of the track by collecting track width data at different locations. The track digital twin model needs to integrate the track geometric model, the train kinematics model and the environment dynamic model, and input the real-time collected data into the model for simulation, to obtain the train position, speed and acceleration changes over time, track geometric characteristics such as slope change and curvature change, and the influence of the environment on the train operation such as temperature, humidity and wind speed.

[0017] Referring to Figure 3 As shown in the drawings, the global inspection path of the inspection robot is obtained based on the A* algorithm and the real-time obstacle avoidance trajectory output by the digital twin model, and specifically includes: Based on the track digital twin model, the inspection target area is determined, and the inspection area is divided into grids to form a two-dimensional grid map; Each grid in the grid map corresponds to a specific spatial area of the inspection target area, and the corresponding cost is given according to the track number density in the spatial area; Based on the A* algorithm, the Euclidean distance from the starting position of each inspection robot to the target position is designed as the heuristic value of the heuristic function to obtain the estimated shortest distance; Based on the real-time obstacle data in the environment data in the track digital twin model, the original estimated shortest distance path is adjusted through the track digital twin model to simulate the identification of the inspection robot through the sensor and combined with the A* algorithm, and a new path is generated through local path planning; The local paths of each section of the track are summarized to generate global path planning data.

[0018] Specifically, based on the track digital twin model, the track area is spatially divided, each grid corresponds to a spatial area, contains a specific track number density, and a cost function is given according to the density, and the cost function formula is:

[0019] wherein, is the cost function of position (i, j) area, k is a constant coefficient, is the track number density in the grid, reflecting the complexity or importance of the area; The heuristic function of the A* algorithm is used to estimate the shortest distance from the current node to the target node, and the commonly used heuristic function is the Euclidean distance. The total cost function of the A* algorithm consists of two parts: the walked cost and the heuristic cost. The walked cost is the actual cost from the starting position to the current node, which is the cost of the walked path, and is usually the distance to the current position, which can be calculated by the grid step. The heuristic cost is the estimated value of the heuristic function, which is used to estimate the cost from the current position to the target position; When the path encounters an obstacle, the path from the current obstacle point to the next passable point is regenerated by local path planning, and the path is updated by the A* algorithm; Each time the inspection robot encounters an obstacle, the A* algorithm recalculates the shortest path from the current position to the target position, which can be regarded as a local path planning problem. The local path planning will use a modified cost function to consider the influence of obstacles and recalculate the shortest path; The local paths of each section of the track are summarized to generate global path planning data. Global path planning is to connect all local paths according to the layout of the track to form a whole path.

[0020] Referring to Figure 4 As shown in the figure, the inspection priority is divided based on the real-time positioning data of each inspection robot and the task amount of each track section to be inspected, and the dynamic inspection path is allocated. When a path conflict is detected, the inspection path is dynamically adjusted based on the inspection priority, which specifically includes: Obtain real-time positioning data of each inspection robot and task amount of each track section to be inspected, the task amount including track section length, equipment condition, maintenance demand, historical fault record; Based on the real-time positioning data of each inspection robot and the task amount of each track section to be inspected, weight division is performed to construct an inspection priority scoring model, and the inspection priority of each inspection task is obtained according to the scoring result of each inspection task; Based on the global path planning data simulated by the track digital twin model, the inspection priority of each inspection task is combined with the task amount, and the inspection tasks are allocated to each inspection robot; Based on the path allocation, it is detected whether there is a path conflict. If there is, the conflicting tasks are adjusted according to the priority of the inspection tasks. If the priorities are the same, the local path is re-planned.

[0021] Specifically, the task amount of the track section includes: track section length; equipment condition: can be quantified as equipment health condition score, the value range is [0, 1], 1 is the optimal condition, 0 is the worst condition; maintenance demand: according to historical maintenance records or expected maintenance period, can be quantified as maintenance priority, the value range is [0, 1]; historical fault record: fault frequency or fault type severity, which can be represented by a weighted value, the value range is [0, 1]; considering the above factors, a comprehensive task amount is calculated; Combined with the task amount of each inspection task and the real-time positioning data of the inspection robot, an inspection priority scoring model is established to calculate the priority score of each inspection task, the formula is:

[0022] Wherein, is the priority score, for the task amount, for the inspection robot currently located at a position to a task point Euclidean distance, a weight coefficient for the task urgency, an urgency score of the task; Based on the priority score calculated, the inspection tasks are sorted according to priority. The tasks with higher priority need to be assigned to the inspection robots as soon as possible; Through the greedy algorithm, the high-priority tasks are assigned first according to the priority sorting of the inspection tasks, and the nearest robot to the task is selected for task assignment according to the real-time positioning data of the robot. Considering the current task load of the inspection robot, a load balancing strategy is introduced to ensure that the task load of each inspection robot is as uniform as possible; In the process of path planning, it is checked whether the path planned by each inspection robot intersects with the path of other robots, and the inspection task time of different inspection robots is considered to ensure that the robots do not occupy the same track segment at the same time. If there is a path conflict, the following adjustment strategies are adopted based on the priority of the inspection task: priority adjustment: tasks with higher priority retain the original planned path, and tasks with lower priority need to re-plan the path to bypass the conflict area; local path re-planning: if the paths of two inspection robots conflict, re-plan the local path of one of them to ensure that they can avoid the conflict area.

[0023] Referring to Figure 5 , based on the frequency of use of each track segment simulated by the track digital twin model, a frequency threshold is set, and the enhanced inspection mechanism is started for the track with high frequency of use, and multiple inspection robots are dispatched to perform cross-validation inspection, which specifically includes: Based on the simulation data of the track digital twin model, the historical use frequency and the use frequency in the future time period of each track segment are obtained; Based on the historical use frequency of each track segment, a track use frequency threshold is set to divide each track segment into a low use frequency segment and a high use frequency segment; Based on the low use frequency segment, regular inspection is performed, based on the high use frequency segment, the number of inspections is increased, and multiple inspection robots are dispatched to perform task assignment for the same high use frequency segment; Through the collaborative work of multiple robots simultaneously inspecting different high-frequency track segments, cross-validation inspection is performed.

[0024] Specifically, according to the historical and future use frequency of the track segment, a use frequency threshold is set, and the threshold is set to the average value of the historical use frequency plus or minus the standard deviation by the standard deviation method; For low-frequency bands with less inspection tasks and low frequency, the conventional inspection strategy can be used, that is, periodic inspection of each low-frequency band, which can be set to monthly or quarterly inspection of each low-frequency track segment; For high-frequency bands, due to frequent use, more inspections may be needed, and the inspection tasks should be increased. Based on the predicted future usage frequency, the inspection frequency is determined. For high-frequency bands, multiple inspection robots need to work together. Through the distance of the track segment, the current position of the inspection robot, and the task priority, the task is allocated. In order to ensure the quality of inspection, multiple robots perform cross-validation inspection on the same high-frequency track segment. At least 2 or more robots are arranged to alternately perform inspection tasks on the high-frequency track segment. After each robot performs an inspection, other robots will recheck the same area to ensure the reliability of the task.

[0025] Referring to Figure 6 As shown, based on the collected track geometric parameters, surface defects and environmental data, abnormal detection is performed by comparing the fault feature library data, and an abnormal alarm is triggered when an abnormality is detected. Specifically, it includes: Based on the track geometric parameters, surface defects and environmental data collected by the inspection, data preprocessing is performed, and data features are extracted; Based on historical fault data and track defect features, the fault feature library is constructed, and the data features of the real-time collected data are compared with the data features in the fault feature library to obtain fault matching probability distribution data; Based on the fault matching probability distribution data, abnormality is judged, and the type of abnormality is judged, including local fault and global fault; Based on the determination of abnormality, a multi-level abnormality alarm is triggered, and real-time abnormal data is reported.

[0026] Specifically, the features meaningful for fault diagnosis are extracted from the preprocessed data, including track flatness, track width change, track inclination, surface crack length and depth, surface pit area, temperature, humidity, and rainfall; Based on the matching of historical fault data and track defect features, a fault feature library is constructed, which is composed of historical fault types and corresponding track features. The track data features collected in real time are compared with each fault type in the fault feature library, the matching probability is calculated, and the similarity is measured; Based on the fault matching probability, abnormality is judged, and the type of abnormality is further identified, which is mainly divided into two categories: Local fault: Local fault usually refers to defects on the track surface or a small range, which has a small impact. In general, the matching probability of local fault is low, and the corresponding features mainly include features related to track surface and geometric parameters; Global failure: Global failure usually refers to a large-scale problem that affects the entire track segment, such as severe track deformation or systemic failure. Global failures usually have a high probability of matching and involve the combined characteristics of multiple track segments.

[0027] Furthermore, the method according to the embodiments of this application can also be achieved by means of... Figure 7 The architecture of the electronic device shown is used to implement this. For example... Figure 7 As shown, the electronic device 500 may include a bus 501, one or more CPUs 502, a read-only memory (ROM) 503, a random access memory (RAM) 504, a communication port 505 connected to a network, an input / output component 506, a hard disk 507, etc. The storage device in the electronic device 500, such as the ROM 503 or the hard disk 507, may store the machine control method for the intelligent inspection robot for rail transit based on digital twins provided in this application. The electronic device 500 may also include a user interface 508. Of course, Figure 7 The architecture shown is merely exemplary and can be omitted as needed when implementing different devices. Figure 7 One or more components in the illustrated electronic device.

[0028] Figure 8 This is a schematic diagram of a computer-readable storage medium structure provided in one embodiment of this application. Figure 8 The diagram illustrates a computer-readable storage medium 600 according to one embodiment of this application. The computer-readable storage medium 600 stores computer-readable instructions. When executed by a processor, the computer-readable instructions can perform the machine control method for a digital twin-based intelligent inspection robot for rail transit according to an embodiment of this application, as described with reference to the above figures. The storage medium 600 includes, but is not limited to, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0029] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0030] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0031] The above merely describes preferred embodiments of the present application, and is not used to limit the present application, any modification, equivalent replacement, improvement, etc. made within the principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for controlling a rail transit intelligent inspection robot based on digital twinning, characterized in that, The method comprises the following steps: constructing a rail transit geometric model, a kinematics model and an environment dynamic model based on collected rail transit data, and establishing a rail digital twin model to generate running simulation data containing rail data and environment data; generating an inspection path based on the A* algorithm, and obtaining a global inspection path of the inspection robot in combination with a real-time obstacle avoidance trajectory output by the digital twin model; dividing an inspection priority based on real-time positioning data of each inspection robot and a task amount of a to-be-inspected rail section, and performing dynamic inspection path allocation, when a path conflict is detected, dynamically adjusting the inspection path based on the inspection priority; based on the use frequency of each rail section simulated by the rail digital twin model, setting a frequency threshold, starting a strengthened inspection mechanism for a rail with high use frequency, and dispatching multiple inspection robots to perform cross-verification inspection; based on collected rail geometric parameters, surface defects and environment data, performing anomaly detection by comparing with fault feature library data, and triggering an abnormal alarm when an anomaly is detected.

2. The digital-twin-based rail transit intelligent inspection robot machine control method according to claim 1, characterized in that, The method of constructing a rail transit geometric model, a kinematics model and an environment dynamic model based on collected rail transit data, and establishing a rail digital twin model to generate running simulation data containing rail data and environment data comprises the following steps: based on collected rail geometric data, establishing a rail transit geometric model, including a rail plane model, a rail longitudinal section model and a rail transverse section model; based on physical principles, constructing a kinematics model of the train, including a dynamics equation, a vehicle speed and an acceleration; based on collected environment-related data and historical meteorological data, establishing an environment dynamic model; integrating the above rail transit geometric model, kinematics model and environment dynamic model to generate a rail digital twin model, and inputting real-time collected data into the model for digital twin simulation; based on the digital twin simulation result of the rail digital twin model, generating rail running simulation data containing rail data and environment data.

Images