Train dynamic scheduling method and resource allocation system

Abstract

The invention relates to the technical field of data processing, and provides a train dynamic scheduling method and a resource allocation system, through real-time acquisition and processing of heterogeneous data, generation of train priorities, and generation and verification execution of a scheduling scheme, fine management of a whole train scheduling process is realized, all steps cooperate with one another and are progressive layer by layer, and the scheduling efficiency is improved. Various complex factors in the train running process can be comprehensively considered, the limitation that a traditional dispatching mode depends on experience or a single index is changed, the scientificity, accuracy and timeliness of train dispatching decisions are remarkably improved, resources can be dynamically and efficiently allocated, train running arrangement can be optimized, and the train dispatching efficiency is improved. The transportation efficiency and the service quality are improved, and the ever-increasing diversified requirements of modern railway transportation are effectively met.

Description

Technical Field

[0001] This application relates to the technical field of data processing, and particularly to a train dynamic scheduling method and a resource allocation system. Background Art

[0002] With the rapid development of the railway transportation industry, the running density of trains has been increasing continuously, and the transportation environment has become increasingly complex, which puts forward higher requirements for the accuracy, efficiency and flexibility of train scheduling. Early train scheduling mainly relied on the experience of dispatchers and manual operations, and the train operation was commanded by means of telephones and signal equipment. This method was inefficient, error-prone, and difficult to cope with complex transportation scenarios. With the progress of information technology, an automated scheduling system based on a fixed train timetable came into being. This scheduling system schedules trains through a pre-established train operation plan, which improves the scheduling efficiency and accuracy to a certain extent. However, this method lacks adaptability to real-time changing factors. When encountering emergencies, it is unable to adjust the scheduling plan in time, which easily leads to problems such as train delays and line congestion.

[0003] At the same time, most of the existing technologies have a single data type, a simple evaluation method for resource status, a single optimization goal for generating a scheduling plan, and a verification method for the scheduling plan that cannot effectively evaluate and adjust according to real-time changing environmental factors and operating states. Most of them do not solve how to achieve dynamic optimization of train scheduling and reasonable allocation of resources in a complex operating environment. Summary of the Invention

[0004] Aiming at the deficiencies of the prior art, this application provides a train dynamic scheduling method and a resource allocation system.

[0005] In a first aspect, this application provides a train dynamic scheduling method, which includes: obtaining heterogeneous data in real time, where the heterogeneous data includes track circuit signals, platform passenger flow images, and meteorological environment data, aligning the heterogeneous data in space and time through a spatio-temporal attention mechanism at an on-vehicle edge node, and processing the heterogeneous data. At the same time, the power supply network, spare trains, and signal machines are monitored to determine the resource status, so as to output a train data matrix, where the train data matrix includes signal feature vectors, passenger flow density coefficients, meteorological correction parameters, and resource status; Construct an evaluation vector for train scheduling according to the train data matrix and the scheduling knowledge graph. The evaluation vector includes an urgency factor, an economic value factor, and a line impact factor. Based on the fuzzy analytic hierarchy process, the dimension weights of the evaluation vector are dynamically adjusted to generate a train priority; Generate a scheduling plan through reinforcement learning. The state space of the reinforcement learning represents the train priority and the resource status, and the reward function of the reinforcement learning represents the time punctuality rate, the energy consumption reduction rate, and the conflict risk value; Perform Monte Carlo simulation for each scheduling scheme in parallel through the GPU, verify the conflict probability, energy consumption deviation, and delay margin of each scheduling scheme based on the threshold, so as to output the scheduling instruction set after passing the verification. The scheduling instruction set includes train path adjustment and resource allocation instructions. Execute the scheduling instruction set and monitor the feedback data during the execution process of the scheduling instruction set in real time to determine whether to resample heterogeneous data and update the resource status.

[0006] As an alternative implementation, the determination sub-strategy of the resource status includes: Deploy sensors to monitor the power supply network, standby trains, and signal lights in real time to obtain resource data. The resource data includes the load rate of the power supply network, the equipment health and location of standby trains, and the failure probability of signal lights; Taking the power supply network, standby trains, and signal lights as nodes, establish a topology graph. The node attributes represent resource data, and the edges represent the coupling relationship between resource data. Extract the node feature vectors of the topology graph through a graph convolutional network; Embed a Bayesian network on the topology graph, input the node feature vectors into the Bayesian network, and train the Bayesian network through historical operation data to determine the state transition probability between nodes, and integrate the state transition probability between nodes to determine the resource status.

[0007] As an alternative implementation, the output strategy of the train data matrix includes: Obtain heterogeneous data in real time. The heterogeneous data includes track circuit signals, platform passenger flow images, and meteorological environment data. Align the heterogeneous data in space and time through a spatio-temporal attention mechanism at the on-vehicle edge node, and perform standardization processing on the heterogeneous data; Extract features from the track circuit signals through a residual convolutional neural network, and map the extracted features into signal feature vectors. The signal feature vectors include signal strength, signal frequency, and signal phase; Perform semantic segmentation on the platform passenger flow image, calculate the ratio of the passenger pixel area to the total platform area to obtain the initial passenger flow density, and predict the passenger flow change trend through a long short-term memory network in combination with historical passenger flow data to dynamically weight and correct the initial passenger flow density to generate a passenger flow density coefficient; Analyze the influence degree of different meteorological environment data on train operation data, braking distance, and energy consumption through a random forest algorithm, assign influence weights to each meteorological environment data according to the analysis results, and perform weighted summation of the standardized meteorological environment data and the influence weights to generate a meteorological correction parameter; Monitor the power supply network, standby trains, and signal lights to determine the resource status, integrate the signal feature vectors, passenger flow density coefficients, meteorological correction parameters, and resource status, and output the train data matrix.

[0008] As an alternative implementation, the construction sub-strategy of the evaluation vector includes: Normalize the train data matrix and associate the train data matrix with the scheduling knowledge graph; Combine the signal feature vector and resource status in the train data matrix to analyze the train operation status, extract the historical scheduling data of the scheduling knowledge graph to evaluate the late arrival risk value under the train operation status, and evaluate the passenger impact value under train late arrival according to the passenger flow density coefficient and platform carrying capacity. Weighted sum the late arrival risk value and the passenger impact value to obtain the urgency factor of the train; According to the energy consumption and the meteorological correction parameters in the train data matrix, determine the energy consumption cost of the train in combination with the scheduling knowledge graph, obtain the operation cost of the train in combination with the resource cost, determine the operation revenue value according to the passenger flow density coefficient and the scheduling knowledge graph, and determine the economic value in combination with the operation revenue value and the operation cost. Normalize the economic value to obtain the economic value factor of the train; Combine the passenger flow density coefficient and resource status in the train data matrix to analyze the train line load, evaluate the line load value in combination with the historical load data of the scheduling knowledge graph, and perform a weighted combination of the line load value and the line importance in the scheduling knowledge graph to obtain the line impact factor; Integrate the urgency factor, economic value factor and line impact factor to construct an evaluation vector for train scheduling.

[0009] As an alternative implementation, the generation strategy of the train priority includes: Use the fuzzy analytic hierarchy process to take the train priority as the target layer and the evaluation vector as the criterion layer. Characterize the importance between the criterion layers with fuzzy numbers and determine the fuzzy judgment matrix; Conduct a consistency test and correction on the fuzzy judgment matrix, solve the fuzzy characteristic equation to obtain the fuzzy weight vector of the evaluation vector, and perform a fuzzification process on the fuzzy weight vector to obtain the dimension weight of the evaluation vector; Perform a weighted sum of the evaluation vector and the dimension weight to obtain the comprehensive evaluation value of the train, and generate the train priority according to the comprehensive evaluation value of the train.

[0010] As an alternative implementation, the generation strategy of the scheduling plan includes: Construct reinforcement learning. The state space of the reinforcement learning represents the train priority and resource status, the action space of the reinforcement learning represents train speed adjustment, train path selection and resource allocation, and the reward function of the reinforcement learning represents the time punctuality rate, energy consumption reduction rate and conflict risk value; Train the deep Q network through the state space and action space, calculate the target Q value, and update the parameters of the deep Q network based on the mean square error loss function; Select the action with the largest target Q value through the deep Q network, generate a scheduling plan according to the action, and evaluate the generated scheduling plan through the on-time rate, energy consumption reduction rate, and conflict risk value in the reward function to determine whether to adjust the reinforcement learning.

[0011] As an optional implementation manner, the verification sub-strategy of the scheduling instruction set includes: Construct a GPU parallel computing architecture and decompose the tasks of Monte Carlo simulation into three layers of sub-tasks; Determine the threshold based on historical operation data, and perform time series analysis on the conflict probability, energy consumption deviation, and delay margin of each scheduling plan through a long short-term memory network; Dynamically adjust the threshold through the attention mechanism in combination with meteorological environment data, passenger flow density coefficient, and resource status; Perform fuzzy matching between the conflict probability, energy consumption deviation, and delay margin of each scheduling plan obtained by Monte Carlo simulation and the threshold through the fuzzy comprehensive evaluation method; Set membership functions for the conflict probability, energy consumption deviation, and delay margin, determine the evaluation result of the scheduling plan through the fuzzy composition operator, compare the evaluation result with the evaluation threshold to judge whether the scheduling plan passes the verification, and output the scheduling instruction set after passing the verification.

[0012] As an optional implementation manner, the update strategy of the resource status includes: Execute the scheduling instruction set and monitor the feedback data during the execution of the scheduling instruction set in real time. The feedback data includes the operation data of the power supply network, standby trains, and signal lights, as well as the execution status data of the scheduling instruction set; Extract the index features of the feedback data through a convolutional neural network; Learn the index features of the feedback data through a long short-term memory network to predict the change trend of the resource status, update the resource status, and at the same time update the node attributes of the topology graph, the parameters of the Bayesian network, and the state transition probability.

[0013] As an optional implementation manner, the resampling strategy of the heterogeneous data includes: Analyze the correlation relationship between track circuit signals, platform passenger flow images, and meteorological environment data through the association rule mining algorithm, perform anomaly detection on the correlation relationship based on statistical methods, and analyze the reasons for the detected abnormal heterogeneous data; Determine the adjustment mechanism of the sampling frequency according to the results of anomaly detection and cause analysis, and select the sampling method according to the type of abnormal heterogeneous data; Evaluate the quality of the resampled heterogeneous data to judge whether to resample the heterogeneous data, and fuse the resampled heterogeneous data with the train data matrix to update the train data matrix.

[0014] In a second aspect, the present application provides a train dynamic resource allocation system, which includes: obtaining heterogeneous data in real time. The heterogeneous data includes track circuit signals, platform passenger flow images, and meteorological environment data. The heterogeneous data is aligned in space and time through an on-vehicle edge node, and the heterogeneous data is processed. At the same time, the load rate of the power supply network, the equipment health and location of standby trains, and the failure probability of signal lights are obtained to determine the resource status, and a train data matrix is output. An evaluation vector for train scheduling is constructed based on the train data matrix and the scheduling knowledge graph. The dimension weights of the evaluation vector are dynamically adjusted based on the fuzzy analytic hierarchy process to generate the train priority. A scheduling plan is generated through reinforcement learning. According to the train priority in the scheduling plan, the power quotas of each section in the power supply network are dynamically allocated, and a resource allocation instruction is generated in combination with the resource status. The Monte Carlo simulation is performed on each scheduling plan in parallel through the GPU. Based on the threshold, the conflict probability, energy consumption deviation, and delay margin of each scheduling plan are verified to output a set of scheduling instructions after passing the verification. The scheduling instruction set is executed and the execution effect of the scheduling instruction set is monitored in real time to determine whether to resample the heterogeneous data and update the resource status.

[0015] Compared with the prior art, the beneficial effects of the present application are as follows: Through the real-time acquisition and processing of heterogeneous data, to the generation of train priorities, and then to the generation and verification execution of the scheduling plan, the refined management of the entire train scheduling process is realized. Each step cooperates with each other and progresses layer by layer, and can comprehensively consider various complex factors in the train operation process, changing the limitations of traditional scheduling methods that rely on experience or single indicators, significantly improving the scientificity, accuracy, and timeliness of train scheduling decisions, enabling dynamic and efficient resource allocation in the face of complex and changing operation scenarios, optimizing train operation arrangements, improving transportation efficiency and service quality, and effectively meeting the growing diverse needs of modern railway transportation.

[0016] Obtain and process heterogeneous data in real time, determine the resource status and output the train data matrix. By obtaining heterogeneous data such as track circuit signals, platform passenger flow images, and meteorological environment data, the key factors affecting train operation are comprehensively covered, enabling scheduling decisions to be based on a richer and more comprehensive information basis and avoiding decision-making biases caused by data missing; by monitoring the power supply network, standby trains, and signal lights, the resource status can be accurately determined, and the potential relationships and change trends between resources can be deeply explored, providing an accurate basis for reasonable resource allocation; the integrated and processed data forms a train data matrix, presenting various types of information in a structured form, facilitating subsequent analysis and application, and providing a standardized and normalized data input for constructing the evaluation vector and generating the scheduling plan, etc.

[0017] Construct an evaluation vector based on the train data matrix and the dispatching knowledge graph to generate the train priority. The evaluation vector is constructed from multiple dimensions such as the urgency level, economic value, and line impact, comprehensively considering key factors such as the safety, economy, and line resource utilization of train operation, making the generation of train priority more comprehensive and objective, capable of balancing the needs of different aspects and optimizing resource allocation; dynamically adjust the dimension weights of the evaluation vector based on the fuzzy analytic hierarchy process, which can flexibly reflect the relative importance of each evaluation factor according to different operation scenarios and requirements, making the train priority more in line with the actual situation and enhancing the flexibility and adaptability of dispatching decisions.

[0018] Generate a dispatching plan through reinforcement learning. Using reinforcement learning, based on the train priority and resource status, through continuous trial and error and learning, starting from multiple objectives such as the time punctuality rate, energy consumption reduction rate, and conflict risk value, automatically search for the optimal dispatching plan, realizing the intelligence and automation of dispatching plan generation. Compared with traditional methods, it can find better solutions more efficiently and improve the quality of the dispatching plan; the reward function covers multiple key indicators such as the time punctuality rate, energy consumption reduction rate, and conflict risk value, prompting reinforcement learning to comprehensively consider multiple important aspects of train operation when generating the dispatching plan, effectively balancing the safety, punctuality, and economy of train operation, and enhancing the overall efficiency of railway transportation.

[0019] Perform Monte Carlo simulation on the dispatching plan through GPU parallelism, verify and output the dispatching instruction set, and monitor the execution feedback. By means of the GPU parallel computing architecture to perform Monte Carlo simulation on the dispatching plan, the simulation calculation efficiency is greatly improved, and a large number of dispatching plans can be quickly verified in a short time to ensure that feasible dispatching plans can be screened out within a limited time to meet the real-time requirements of train dispatching; verify the conflict probability, energy consumption deviation, and time delay margin of the dispatching plan based on thresholds, and comprehensively evaluate the dispatching plan by means of the fuzzy comprehensive evaluation method, etc., effectively excluding plans with potential risks or non-compliance requirements, ensuring that the output dispatching instruction set has high feasibility and reliability; real-time monitor the feedback data during the execution of the dispatching instruction set, and judge whether to resample heterogeneous data and update the resource status according to the execution effect, so that the changes in actual operation can be sensed in a timely manner, and the dispatching decision can be dynamically adjusted and optimized based on new information to form a closed-loop management, continuously improving the accuracy and adaptability of train dispatching and ensuring the stability and efficiency of train operation. Description of the Drawings

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts. Among them: Figure 1 It is the flowchart of the train dynamic scheduling method provided by the embodiments of the present application; Figure 2 It is the determination sub-strategy diagram of the resource status of the train dynamic scheduling method provided by the embodiments of the present application; Figure 3 It is the verification sub-strategy diagram of the scheduling instruction set of the train dynamic scheduling method provided by the embodiments of the present application. Detailed implementation manners

[0021] To make the objectives, technical solutions and advantages of the embodiments of the present application more obvious and understandable, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings of the specification. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments.

[0022] Embodiment 1 As Figure 1 shown, it is the flowchart of the train dynamic scheduling method provided by the embodiments of the present application. The train dynamic scheduling method includes: S1. Heterogeneous data is acquired in real time. The heterogeneous data includes track circuit signals, platform passenger flow images, and meteorological environment data. The heterogeneous data is spatially and temporally aligned through a spatio-temporal attention mechanism at the on-vehicle edge node, and the heterogeneous data is processed. Meanwhile, the power supply network, spare trains, and signal lights are monitored to determine the resource status, so as to output a train data matrix. The train data matrix includes signal feature vectors, passenger flow density coefficients, meteorological correction parameters, and resource status.

[0023] As Figure 2 shown, the determination sub-strategy of the resource status includes: Sensors are deployed to monitor the power supply network, spare trains, and signal lights in real time to obtain resource data. The resource data includes the load rate of the power supply network, the equipment health and location of the spare trains, and the failure probability of the signal lights; Taking the power supply network, spare trains, and signal lights as nodes, a topology graph is established. The node attributes represent the resource data, and the edges represent the coupling relationship between the resource data. The node feature vectors of the topology graph are extracted through a graph convolutional network; A Bayesian network is embedded in the topology graph. The node feature vectors are input into the Bayesian network, and the Bayesian network is trained through historical operation data to determine the state transition probability between the nodes. The state transition probabilities between the nodes are integrated to determine the resource status.

[0024] The operation of trains depends on the normal operation of resources such as the power supply network, spare trains and signal lights. Real-time acquisition of their status data is the basis for understanding the resource status and providing a basis for subsequent scheduling decisions. Voltage sensors, current sensors and load sensors are deployed at key nodes of the power supply network to obtain data such as the load rate of the power supply network in real time. Equipment health monitoring sensors are installed on spare trains to monitor equipment operating parameters to obtain equipment health. At the same time, positioning devices are used to obtain the position of the train. Fault detection modules are set at signal lights to monitor the failure probability of signal lights in real time. These sensors and monitoring modules transmit data to the on-board edge nodes through wireless networks. This realizes the comprehensive and real-time collection of resource data, ensures the timeliness and accuracy of data, enables the resource allocation system to perceive changes in resource status in a timely manner, provides original data for building a resource topology map, and ensures that the node feature vectors extracted by the graph convolutional network can truly reflect the resource status.

[0025] There are mutual correlations and influences between resources. The relationship between resources can be intuitively displayed through the topological map. The graph convolutional network can effectively extract the feature vectors that imply the relationship, laying the foundation for accurately evaluating the resource status; the obtained resource data is used as the node attribute, and the edges are determined according to the actual connection and mutual influence relationship of the resources, and a directed weighted topological graph is constructed. The topological graph is processed by the graph convolutional network. Through multi-layer convolution operations, the nodes and their neighborhood information are aggregated, and the node feature vectors containing the coupling relationship between resources are extracted; thereby, the complex resource relationship is converted into a topological structure and feature vector that is easy for computers to process, which improves the efficiency and accuracy of resource status analysis, can mine the potential influence patterns between resources, and provide effective input data for the Bayesian network, so that the Bayesian network can more accurately determine the state transition probability between nodes based on these feature vectors.

[0026] Bayesian networks are good at dealing with uncertainty problems. After training with historical data, they can predict resource status changes based on current feature vectors, providing more reliable resource status information for scheduling decisions. The Bayesian network is embedded in the constructed topology map, and the extracted node feature vectors are used as input. A large amount of historical operation data is collected, and the Bayesian network is trained using methods such as maximum likelihood estimation to determine the state transition probability between nodes. Finally, these state transition probabilities are integrated to comprehensively judge the current state of resources and future change trends. This enables dynamic prediction of resource status, early detection of potential problems, and the purchase of time for scheduling decisions, improving the foresight and accuracy of scheduling. The determined resource status, as an important part of the train data matrix, affects the construction of subsequent train scheduling evaluation vectors and the generation of scheduling plans.

[0027] The output strategies of the train data matrix include: Obtain heterogeneous data in real time. The heterogeneous data includes track circuit signals, platform passenger flow images, and meteorological environment data. At the on-vehicle edge node, the heterogeneous data is aligned in space and time through a spatio-temporal attention mechanism, and the heterogeneous data is normalized; Extract features from the track circuit signals through a residual convolutional neural network, and map the extracted features to signal feature vectors. The signal feature vectors include signal strength, signal frequency, and signal phase; Perform semantic segmentation on the platform passenger flow images, calculate the ratio of the passenger pixel area to the total platform area to obtain the initial passenger flow density, and predict the passenger flow change trend through a long short-term memory network in combination with historical passenger flow data to dynamically weight and correct the initial passenger flow density, generating a passenger flow density coefficient; Analyze the influence degree of different meteorological environment data on train operation data, braking distance, and energy consumption through a random forest algorithm, assign influence weights to each meteorological environment data according to the analysis results, and perform weighted summation on the normalized meteorological environment data and the influence weights to generate a meteorological correction parameter; Monitor the power supply network, standby trains, and signal lights to determine the resource status, integrate the signal feature vectors, passenger flow density coefficients, meteorological correction parameters, and resource status, and output a train data matrix.

[0028] Train operation is affected by a variety of heterogeneous data. These heterogeneous data have differences in space and time and need to be aligned and normalized to provide a unified and accurate data basis for subsequent analysis; Through sensors and cameras installed on trains and along the track, track circuit signals, platform passenger flow images, and meteorological environment data are obtained in real time. A spatio-temporal attention mechanism is deployed at the on-vehicle edge node. This spatio-temporal attention mechanism automatically learns the correlation features of different heterogeneous data in the spatio-temporal dimension according to the timestamp and spatial position information of the data, realizes the spatio-temporal alignment of heterogeneous data, and then uses methods such as normalization to normalize the heterogeneous data to eliminate the influence of dimensions; Thus, the problems of spatio-temporal inconsistency and different dimensions of heterogeneous data are solved, making the heterogeneous data comparable and compatible, improving the efficiency and accuracy of heterogeneous data processing, and providing high-quality data for subsequent extraction of signal feature vectors, calculation of passenger flow density coefficients, and meteorological correction parameters, ensuring the accuracy of these parameter calculations.

[0029] Track circuit signals contain key information about train operation. However, the original track circuit signals are complex. The residual convolutional neural network can effectively extract their key features, form signal feature vectors, and be used to analyze the train operation status. The standardized track circuit signals are input into the residual convolutional neural network. Through a multi-layer residual block structure, the residual convolutional neural network automatically extracts the deep features of the track circuit signals in the time series, including key information such as signal strength, frequency, and phase, and maps them into signal feature vectors. Thus, it can accurately extract key features from complex track circuit signals, avoid information loss, and provide an accurate signal analysis basis for train dispatching. The signal feature vectors, as part of the train data matrix, participate in the construction of subsequent train dispatching evaluation vectors and affect the determination of train priorities.

[0030] Passenger flow density affects train dispatching strategies. Through semantic segmentation and combined with historical data prediction, it can more accurately reflect the current and future passenger flow conditions and provide a basis for dispatching decisions. The semantic segmentation algorithm is used to process the passenger flow images on the platform, distinguish passengers from the background, calculate the ratio of the pixel area of passengers (i.e., the number of pixels of passengers) to the total area of the platform to obtain the initial passenger flow density. At the same time, historical passenger flow data is collected, and the long short-term memory network is used to predict the trend of passenger flow changes. According to the prediction results, the initial passenger flow density is dynamically weighted and corrected to generate the passenger flow density coefficient. It not only considers the current passenger flow situation but also combines historical data to predict future trends, making the passenger flow density coefficient more forward-looking and accurate, providing support for the reasonable arrangement of train capacity. The passenger flow density coefficient, as a component of the train data matrix, is used to construct the train dispatching evaluation vector and affects the determination of train priorities and the formulation of dispatching plans.

[0031] Meteorological environment has an important impact on train operation. By analyzing the relationship between it and train operation parameters, meteorological correction parameters are generated, which can consider the influence of meteorological factors in dispatching decisions and optimize dispatching plans. The standardized meteorological environment data and train operation data are input into the random forest algorithm to analyze the influence degree of different meteorological factors on train operation parameters such as braking distance and energy consumption. According to the analysis results, influence weights are assigned to each meteorological environment data, and the standardized meteorological environment data and weights are weighted and summed to obtain the meteorological correction parameters. Thus, the influence of the meteorological environment on train operation is quantified, enabling dispatching decisions to fully consider meteorological factors, improving the safety and efficiency of train operation, and reducing operating costs. The meteorological correction parameters, as part of the train data matrix, participate in the construction of subsequent train dispatching evaluation vectors and affect the generation of train priorities and dispatching plans.

[0032] Train dispatching needs to comprehensively consider various factors, integrate these key data to form a train data matrix, and provide comprehensive data support for subsequent dispatching decisions; determine the resource status according to the logical steps determined by the above resource status, and splice and integrate the extracted signal feature vectors, calculated passenger flow density coefficients, generated meteorological correction parameters, and determined resource status according to specific dimensions and orders to form a train data matrix; thus integrating various key data related to train operation into a matrix, making the data structure clear, facilitating the subsequent dispatching decision-making system to quickly obtain and analyze, improving the efficiency and accuracy of dispatching decisions, and the output train data matrix is used to construct an evaluation vector for train dispatching, which in turn affects the generation of train priorities and the formulation of dispatching plans.

[0033] S2. Construct an evaluation vector for train dispatching based on the train data matrix and the dispatching knowledge graph. The evaluation vector includes an urgency factor, an economic value factor, and a line impact factor, and dynamically adjust the dimension weights of the evaluation vector based on the fuzzy analytic hierarchy process to generate train priorities.

[0034] The construction sub-strategies of the evaluation vector include: Normalize the train data matrix and associate the train data matrix with the dispatching knowledge graph; Combine the signal feature vector and resource status in the train data matrix to analyze the train operation status, extract the historical dispatching data of the dispatching knowledge graph to evaluate the late arrival risk value under the train operation status, and evaluate the passenger impact value under train late arrival according to the passenger flow density coefficient and platform carrying capacity. Weighted sum the late arrival risk value and the passenger impact value to obtain the urgency factor of the train; According to the energy consumption and the meteorological correction parameters in the train data matrix, combine the dispatching knowledge graph to determine the energy consumption cost of the train, combine the resource cost to obtain the operation cost of the train, and determine the operation revenue value according to the passenger flow density coefficient and the dispatching knowledge graph. Combine the operation revenue value and the operation cost to determine the economic value, and normalize the economic value to obtain the economic value factor of the train; Combine the passenger flow density coefficient and resource status in the train data matrix to analyze the train line load, combine the historical load data of the dispatching knowledge graph to evaluate the line load value, and perform a weighted combination of the line load value and the line importance in the dispatching knowledge graph to obtain the line impact factor; Integrate the urgency factor, the economic value factor, and the line impact factor to construct an evaluation vector for train dispatching.

[0035] The data dimensions and dimensions in the train data matrix are different. Normalization processing can eliminate the differences, facilitating comparison and analysis. The dispatching knowledge graph contains information such as historical dispatching experience and rules. Associating the two can provide more comprehensive data support and knowledge reference for the construction of the evaluation vector. The min-max normalization method is adopted to map data such as signal feature vectors, passenger flow density coefficients, meteorological correction parameters, and resource status in the train data matrix to a unified interval. Using knowledge graph embedding technology, semantic matching is performed on the key data in the train data matrix with the entities and relationships in the dispatching knowledge graph, establishing a data association index to achieve rapid interactive call of data and knowledge; thereby ensuring data consistency and comparability, improving data analysis efficiency, integrating historical experience and real-time data, mining potential information, providing a richer and more accurate basis for the subsequent calculation of evaluation factors, and providing a standardized and associated knowledge data basis for calculating the emergency degree factor, economic value factor, and line impact factor, making the calculation of each factor more accurate, and further ensuring the reliability of the evaluation vector.

[0036] Combine the signal feature vector and resource status in the train data matrix to analyze the train operation status, extract the historical dispatching data of the dispatching knowledge graph to evaluate the late arrival risk value under the train operation status, and evaluate the passenger impact value under train late arrival according to the passenger flow density coefficient and platform carrying capacity. The late arrival risk value and the passenger impact value are weighted and summed to obtain the emergency degree factor of the train. Train late arrival will affect the operation order and passenger experience. Comprehensively considering factors such as operation status, historical data, and passenger flow to evaluate the late arrival risk and passenger impact can accurately quantify the train's emergency degree, providing a basis for dispatching decisions to prioritize the handling of emergency trains. By analyzing the signal feature vector to determine whether the track circuit signal is abnormal, and combining the resource status to determine whether equipment such as power supply and signal machines is normal, comprehensively evaluate the train operation status, extract historical late arrival data under similar operation status from the dispatching knowledge graph, use case-based reasoning technology to predict the late arrival risk value of the current train, determine the impact of passengers staying on the platform on the platform itself according to the passenger flow density coefficient and platform carrying capacity, evaluate the passenger impact value when the train is late, and finally determine the weight according to the historical data statistics, and perform weighted summation on the late arrival risk value and the passenger impact value to obtain the emergency degree factor; thus comprehensively considering various factors, accurately evaluating the train's emergency degree, enabling dispatching decisions to prioritize handling emergencies, reducing the adverse impact of late arrival on operation and passengers, and improving the emergency handling ability of railway transportation. The emergency degree factor, as an important part of the evaluation vector, directly affects the generation of train priority. Trains with a high emergency degree will be given a higher priority in dispatching.

[0037] Railway operation needs to consider costs and revenues. Accurately evaluating the economic value of trains helps optimize resource allocation, improve operational efficiency, and achieve the economic goals of railway transportation. Based on energy consumption and meteorological correction parameters, combined with the empirical consumption relationship of energy consumption under different meteorological conditions in the dispatching knowledge graph, estimate the energy consumption cost of trains, aggregate resource costs such as energy consumption cost, equipment maintenance cost, and labor cost to obtain the train operation cost. According to the passenger flow density coefficient, combined with the ticket sales law under different passenger flow conditions in the dispatching knowledge graph, predict the operation revenue value, and subtract the operation cost from the operation revenue value to get the economic value. Then, use the normalization method to map the economic value to a specific interval to obtain the economic value factor. Thus, quantify the economic indicators of train operation, provide an economic reference for railway operation decision-making, help arrange train operation reasonably, reduce costs, increase revenues, and achieve the maximization of economic benefits. The economic value factor is one of the key elements of the evaluation vector, participates in the calculation of train priority, enables dispatching decisions to take into account both urgency and economic factors, and optimizes resource allocation.

[0038] Train operation generates loads on the lines. Different lines have different importance levels. Considering the line load and importance comprehensively can accurately evaluate the impact of train operation on the lines, plan train operation reasonably, and ensure the efficient operation of the lines. Judge the pressure of passenger flow on the lines according to the passenger flow density coefficient, analyze the equipment carrying capacity in combination with the resource status, evaluate the train line load, extract historical load data from the dispatching knowledge graph, use the time series analysis method to predict the current line load value, obtain relevant indicators of line importance in the dispatching knowledge graph, such as the economic status of the cities connected by the line and the proportion of passenger flow, determine the weight according to the analytic hierarchy process, and perform a weighted combination of the line load value and line importance to get the line impact factor. Thus, comprehensively evaluate the impact of trains on the lines, enable dispatching decisions to arrange trains reasonably according to the actual situation of the lines, avoid line congestion, improve the overall operation efficiency and reliability of the lines. The line impact factor, as a component of the evaluation vector, plays a role in the generation of train priority and affects the formulation of the dispatching plan, making train operation more in line with the actual needs of the lines.

[0039] A single factor cannot comprehensively reflect the train dispatching requirements. Integrating multiple factors to construct an evaluation vector can comprehensively evaluate the train dispatching situation from multiple dimensions and provide a comprehensive basis for generating reasonable train priorities. Combine the calculated urgency factor, economic value factor, and line impact factor in the preset order and format to form a multi-dimensional vector, that is, the evaluation vector of train dispatching. Thus, achieve a comprehensive quantitative evaluation of the train dispatching situation, provide comprehensive and accurate data support for the subsequent generation of train priorities, make dispatching decisions more scientific and reasonable. The evaluation vector is the direct basis for generating train priorities, and its accuracy and comprehensiveness determine the rationality of train priorities, which in turn affect the formulation and implementation of the dispatching plan.

[0040] The generation strategy of train priority includes: Using the fuzzy analytic hierarchy process, taking the train priority as the target layer and the evaluation vector as the criterion layer, representing the importance between criterion layers through fuzzy numbers and determining the fuzzy judgment matrix; Conducting consistency test and correction on the fuzzy judgment matrix, obtaining the fuzzy weight vector of the evaluation vector by solving the fuzzy characteristic equation, and defuzzifying the fuzzy weight vector to obtain the dimensional weight of the evaluation vector; Performing weighted summation of the evaluation vector and the dimensional weight to obtain the comprehensive evaluation value of the train, and generating the train priority according to the comprehensive evaluation value of the train.

[0041] The influencing factors of train scheduling are complex and fuzzy, and it is difficult for traditional methods to handle them accurately. The fuzzy analytic hierarchy process can effectively process fuzzy information. By constructing a hierarchical structure and a fuzzy judgment matrix, the relative importance of each evaluation factor is quantified, laying a foundation for generating accurate train priorities; setting the train priority as the target layer, taking the urgency factor, economic value factor, and line influence factor in the evaluation vector as the criterion layer, and using triangular fuzzy numbers or trapezoidal fuzzy numbers to compare the importance between each factor in the criterion layer pairwise according to experience and actual situations to construct a fuzzy judgment matrix; thus fully considering the fuzzy nature and uncertainty in train scheduling decisions, more truly reflecting the actual situation, making the quantification of the importance of each evaluation factor more accurate, improving the scientific nature of train priority generation. The fuzzy judgment matrix is the basis for calculating the dimensional weight of the evaluation vector, and its accuracy directly affects the weight calculation result, and further affects the rationality of the train priority.

[0042] Ensuring the consistency of the fuzzy judgment matrix, avoiding logical contradictions, guaranteeing the accuracy of weight calculation. Solving the fuzzy characteristic equation to obtain the fuzzy weight vector and processing it can convert fuzzy information into precise weights available for calculation; using the fuzzy consistency index and the fuzzy random consistency index to conduct consistency test on the fuzzy judgment matrix. If the consistency requirement is not met, the matrix is corrected through expert feedback. Using fuzzy mathematics theory, solving the fuzzy characteristic equation to obtain the fuzzy weight vector of the evaluation vector, and then through defuzzification methods such as the centroid method and the maximum membership degree method, converting the fuzzy weight vector into precise dimensional weights; thus ensuring the rationality of the fuzzy judgment matrix and the accuracy of weight calculation, making the train priority generation based on reliable weight allocation, improving the reliability and effectiveness of scheduling decisions. The determined dimensional weights are used to calculate the comprehensive evaluation value of the train, directly affecting the train priority ranking. Reasonable weights can make the priority more in line with the actual scheduling requirements.

[0043] The evaluation vector and dimension weights are combined through weighted summation to obtain a comprehensive evaluation value. Based on this, the train priority is generated, realizing the transformation from multi-dimensional evaluation to specific priority ranking, providing a clear priority reference for dispatching decisions. Multiply the urgency factor, economic value factor, and line impact factor in the evaluation vector by their corresponding dimension weights respectively, and then sum the products to obtain the comprehensive evaluation value of the train. Sort the trains in descending order of the comprehensive evaluation value to generate the train priority. Thus, the complex multi-dimensional evaluation results are transformed into an intuitive priority ranking, providing a clear decision-making basis for the dispatcher, making the dispatching decision more efficient and accurate, optimizing the allocation of train dispatching resources. The generated train priority is an important basis for generating the dispatching plan through reinforcement learning. Trains with higher priorities will be given priority consideration in the generation of the dispatching plan, affecting the overall planning and implementation of the dispatching plan.

[0044] S3. Generate the dispatching plan through reinforcement learning. The state space of reinforcement learning represents the train priority and resource status, and the reward function of reinforcement learning represents the time punctuality rate, energy consumption reduction rate, and conflict risk value.

[0045] The generation strategy of the dispatching plan includes: Construct reinforcement learning. The state space of reinforcement learning represents the train priority and resource status, the action space of reinforcement learning represents train speed adjustment, train path selection, and resource allocation, and the reward function of reinforcement learning represents the time punctuality rate, energy consumption reduction rate, and conflict risk value; Train the deep Q-network through the state space and action space, calculate the target Q value, and update the parameters of the deep Q-network based on the mean square error loss function; Select the action with the largest target Q value through the deep Q-network, generate the dispatching plan according to the action, and evaluate the generated dispatching plan through the time punctuality rate, energy consumption reduction rate, and conflict risk value in the reward function to determine whether to adjust the reinforcement learning.

[0046] In order to enable train dispatching to adapt to complex and ever-changing operating environments, the train dispatching process is simulated through reinforcement learning. Taking train priority and resource status as the core decision-making basis, starting from key indicators such as time punctuality rate, energy consumption reduction rate, and conflict risk value, it automatically learns and generates the optimal dispatching plan; a reinforcement learning model is constructed. The train priority and resource status information are integrated into the state space, and the action space clearly covers actual dispatching actions such as train speed adjustment (such as acceleration, deceleration, and constant speed), train route selection (choosing among multiple optional lines), and resource allocation (dispatching spare trains and adjusting power supply network resources, etc.). At the same time, a reward function is defined, with the time punctuality rate, energy consumption reduction rate, and conflict risk value as the reward evaluation indicators, and the corresponding reward or punishment rules for each indicator in different situations are clarified to guide the reinforcement learning process; thus, a reinforcement learning model that conforms to the actual needs of train dispatching is constructed, providing a structured learning framework for subsequent automatic generation of efficient dispatching plans, enabling the model to learn and optimize from actual operation data and target indicators. At the same time, the determined state space, action space, and reward function provide clear rules and goals for the training of the deep Q-network, enabling the deep Q-network to effectively learn for train dispatching problems and then generate reasonable dispatching plans.

[0047] The deep Q-network needs to continuously learn in the given state space and action space, evaluate the value of different actions by calculating the target Q-value, and adjust the network parameters according to the mean squared error loss function to gradually optimize the dispatching strategy to adapt to complex train dispatching scenarios; the constructed state space is input into the deep Q-network in sequence. The deep Q-network evaluates each action in the action space according to the current state and outputs the corresponding Q-value. At the same time, an experience replay mechanism is adopted to store the state, action, reward, and next state in each decision-making process in the experience pool. A batch of data is randomly sampled from the experience pool regularly, and the target Q-value is calculated according to the Bellman equation. Then, the error between the predicted Q-value and the target Q-value is calculated using the mean squared error loss function, and the parameters of the deep Q-network are updated through the backpropagation algorithm to continuously optimize the network's ability to evaluate the value of actions; through this training method, the deep Q-network can learn from a large number of historical experiences, avoid learning biases caused by the correlation of consecutive decisions, effectively improve the learning efficiency and accuracy of the network for train dispatching strategies, and gradually approach the optimal dispatching plan. After training and optimization, the deep Q-network can more accurately evaluate the value of different actions in various states, provide a reliable basis for selecting the action with the largest target Q-value, and thus generate more reasonable dispatching plans.

[0048] After the deep Q-network is trained, it is necessary to select the optimal action from the evaluated actions to generate a specific scheduling plan. At the same time, the plan is evaluated through the reward function to determine whether the current reinforcement learning strategy is effective, so as to adjust and optimize in a timely manner. For the state space composed of the current train priority and resource status, the deep Q-network outputs the Q value corresponding to each action, and selects the action with the largest Q value. If the action corresponding to the largest Q value is to adjust the running path of a certain train, a corresponding train path adjustment plan is generated according to this action, and a complete scheduling plan is generated in combination with the resource allocation requirements. Then, the scheduling plan is simulated and executed. According to the time punctuality rate, energy consumption reduction rate and conflict risk value generated during the actual operation process, the corresponding reward value is calculated based on the reward function. If the reward value does not meet the expectation or there is a large gap compared with the historical excellent plan, it is considered that the current reinforcement learning needs to be adjusted, and the parameters of the deep Q-network are re-optimized or the rules of the reward function are adjusted; thus, a complete closed-loop from learning to decision-making is realized. By selecting the action with the largest Q value to generate a scheduling plan and evaluating it based on the reward function, problems existing in the scheduling plan and learning strategy can be found in a timely manner, ensuring that the generated scheduling plan has high practicability and optimization. The evaluated scheduling plan will enter the simulation as a candidate plan. Through links such as GPU parallel Monte Carlo simulation and threshold verification, the final scheduling instruction set that meets the requirements is further screened to ensure the safety and efficiency of the actually executed scheduling plan.

[0049] S4. Perform Monte Carlo simulation on each scheduling plan in parallel through GPU, and verify the conflict probability, energy consumption deviation and time delay margin of each scheduling plan based on the threshold, so as to output the scheduling instruction set after passing the verification. The scheduling instruction set includes train path adjustment and resource allocation instructions. Execute the scheduling instruction set and monitor the feedback data during the execution process of the scheduling instruction set in real time to determine whether to resample heterogeneous data and update the resource status.

[0050] As Figure 3 shown, the verification sub-strategy of the scheduling instruction set includes: Construct a GPU parallel computing architecture and decompose the tasks of Monte Carlo simulation into three layers of subtasks; Determine the threshold based on historical operation data, and perform time series analysis on the conflict probability, energy consumption deviation and time delay margin of each scheduling plan through a long short-term memory network; Dynamically adjust the threshold through the attention mechanism in combination with meteorological environment data, passenger flow density coefficient and resource status; Perform fuzzy matching between the conflict probability, energy consumption deviation and time delay margin of each scheduling plan obtained by Monte Carlo simulation and the threshold through the fuzzy comprehensive evaluation method; Set membership functions for the conflict probability, energy consumption deviation, and time delay margin. Through the fuzzy composition operator, comprehensively determine the evaluation results of the scheduling scheme by combining the membership functions. Compare the evaluation results with the evaluation threshold to determine whether the scheduling scheme passes the verification, and output the scheduling instruction set after passing the verification.

[0051] There are a large number of train scheduling schemes, and the Monte Carlo simulation has a large amount of calculation. The traditional calculation method is inefficient. By constructing a GPU parallel computing architecture and decomposing tasks, the parallel computing power of the GPU can be fully utilized, significantly shortening the simulation time and meeting the real-time requirements of train scheduling; use the CUDA programming model to build a GPU parallel computing architecture, and divide the Monte Carlo simulation tasks into three levels of sub-tasks: the scheme level, the scenario level, and the calculation level according to functions. The scheme-level tasks are responsible for distributing different scheduling schemes to different GPU thread blocks. The scenario-level tasks start multiple threads in parallel for simulation for different operation scenarios such as different meteorological conditions and passenger flows within each thread block. The calculation-level tasks are executed by each thread for specific calculations, such as train operation trajectory and resource consumption calculations, etc. Through shared memory and synchronization mechanisms, efficient data interaction and collaboration between threads are achieved; thus, significantly improving the efficiency of the Monte Carlo simulation, ensuring that a large number of scheduling scheme simulations can be processed in a short time, providing support for quickly screening out feasible scheduling schemes, improving the response speed of train scheduling, quickly completing the simulation calculation, providing timely data for the subsequent verification of the scheduling scheme based on the threshold, ensuring the smooth progress of the verification process, and accelerating the generation speed of the entire scheduling instruction set.

[0052] Historical operation data contains the laws and experiences of train scheduling under different conditions. Determining the threshold based on this can be used as a benchmark for evaluating the feasibility of the scheduling scheme. The long short-term memory network can effectively process time series data and analyze the changing trends of relevant indicators of the scheduling scheme, providing a basis for accurately evaluating the scheme; collect a large amount of historical train operation data, including conflict probability, energy consumption deviation, and time delay margin data under different time periods, weather conditions, and passenger flows, etc. Use statistical analysis methods and combine expert experience to determine the reasonable threshold range for each indicator. Organize the conflict probability, energy consumption deviation, and time delay margin data of each scheduling scheme into a sequence in chronological order and input it into the long short-term memory network. The long short-term memory network automatically learns the time-dependent relationships and trend characteristics in the data; the threshold determined based on historical data conforms to the actual operation situation. The long short-term memory network can accurately capture the changing trends of the scheduling scheme indicators, improving the accuracy and reliability of the evaluation of the scheduling scheme, preventing unreasonable schemes from entering the actual scheduling, and providing basic data and analysis results for dynamically adjusting the threshold and fuzzy matching, making the subsequent verification process more scientific and reasonable, and ensuring that the selected scheduling scheme meets the actual operation requirements.

[0053] The train operation environment is complex and changeable, with real-time changes in meteorology, passenger flow, and resource status. Fixed thresholds cannot adapt to different situations. By combining the attention mechanism with multi-source data to dynamically adjust the thresholds, the thresholds can be better adapted to the actual operating conditions, ensuring the accuracy of the dispatching plan evaluation. The meteorological environment data, passenger flow density coefficient, and resource status obtained in real time are fused with the historical threshold data. Using the attention mechanism, according to the different degrees of influence of each data on train operation, weights are automatically assigned to highlight the key influencing factors. For example, in extreme weather, higher weights are given to the meteorological environment data, and the thresholds of indicators such as conflict probability and energy consumption deviation are adjusted accordingly. Thus, the dynamic adaptive adjustment of the thresholds is realized, enabling the dispatching plan evaluation to better adapt to different operating environments, improving the flexibility and accuracy of the evaluation, ensuring that the dispatching plan can meet the operation requirements in various situations. The adjusted thresholds provide a more accurate standard for fuzzy matching and plan evaluation, making the subsequent verification results more reliable and helping to screen out better dispatching plans.

[0054] There is a certain degree of ambiguity in the evaluation indicators of the dispatching plan. Precise matching is difficult to reflect the actual situation. The fuzzy comprehensive evaluation method can effectively handle fuzzy information and more reasonably evaluate the compliance degree of the dispatching plan with the thresholds through fuzzy matching. Fuzzy sets and membership functions are defined for indicators such as conflict probability, energy consumption deviation, and time delay margin. For example, the conflict probability is divided into fuzzy levels such as "low", "medium", and "high", and the corresponding membership functions for each level are determined. The indicator values of the plan obtained from the Monte Carlo simulation are substituted into the membership function to calculate the membership degrees belonging to different fuzzy levels, and then fuzzy matching is carried out according to the fuzzy range set by the threshold to judge the compliance degree of the plan indicators with the thresholds. Thus, it more truly reflects the ambiguity and uncertainty in the dispatching plan evaluation, improves the rationality and credibility of the evaluation results, avoids misjudgment caused by precise matching, ensures that the evaluation of the dispatching plan is more in line with the actual situation, and provides a basis for determining the evaluation results of the dispatching plan, making it more scientific to judge whether the plan passes the verification by setting the evaluation threshold and ensuring that the output dispatching instruction set has high feasibility.

[0055] Considering multiple evaluation indicators comprehensively, by setting membership functions and fuzzy composition operators, the fuzzy evaluation results of each indicator are synthesized to obtain a comprehensive evaluation result of the scheme. By comparing it with the evaluation threshold, it is determined whether the scheme is feasible to ensure the safety and reliability of the output scheduling instruction set. According to the actual operation requirements and experience, appropriate membership functions are set for the conflict probability, energy consumption deviation, and time delay margin respectively, and appropriate fuzzy composition operators are selected, such as the maximum-minimum composition operator and the weighted average composition operator. The membership degrees of each indicator are comprehensively calculated to obtain the comprehensive evaluation result of the scheduling scheme. An evaluation threshold is set, and the comprehensive evaluation result is compared with the evaluation threshold. If the evaluation result is greater than or equal to the evaluation threshold, it is determined that the scheduling scheme passes the verification, and a scheduling instruction set including train path adjustment and resource allocation instructions is output. Thus, the scheduling scheme is comprehensively evaluated from multiple dimensions, considering the impacts of each indicator comprehensively. Through a scientific evaluation method, it is ensured that the selected scheduling scheme meets the actual operation requirements, improving the safety and efficiency of train scheduling. The verified scheduling instruction set after output will enter the execution stage, and at the same time, the feedback data during the execution process will be used to determine whether it is necessary to resample heterogeneous data and update the resource status, providing a reference for subsequent scheduling decisions.

[0056] The update strategy of the resource status includes: Execute the scheduling instruction set and monitor the feedback data during the execution process of the scheduling instruction set in real time. The feedback data includes the operation data of the power supply network, standby trains, and signal lights, as well as the execution status data of the scheduling instruction set. Extract the index features of the feedback data through a convolutional neural network. Learn the index features of the feedback data through a long short-term memory network to predict the change trend of the resource status, update the resource status, and at the same time update the node attributes of the topology graph, the parameters of the Bayesian network, and the state transition probability.

[0057] During the execution process of the scheduling instruction set, the resource status will change. Monitoring the feedback data in real time can timely grasp the resource operation situation, provide an accurate basis for resource status update, and ensure the effectiveness of scheduling decisions. Deploy various sensors and monitoring devices in the train operation system to collect the operation data such as the voltage, current, and load rate of the power supply network, the equipment status and location of standby trains, and the working status of signal lights in real time, as well as the execution status data such as the execution progress and completion status of scheduling instructions. The data is transmitted to the scheduling center server in real time through a wireless network. Thus, the real-time monitoring of the execution process of scheduling instructions and the resource operation status is realized, accurate data is obtained in a timely manner, providing reliable information for subsequent resource status analysis and update, ensuring the safety of train operation and the smooth progress of scheduling. The obtained feedback data provides the original data for subsequent extraction of index features and prediction of the change trend of resource status, which is the basis for resource status update and determines the accuracy and timeliness of the update.

[0058] The feedback data is large in volume and complex, making it difficult to directly process. The convolutional neural network has a powerful feature extraction ability and can automatically extract key index features from the data, providing effective information for analyzing the changes in resource status. The obtained feedback data is preprocessed and converted into a format suitable for the input of the convolutional neural network. A convolutional neural network is constructed, and appropriate convolutional layers, pooling layers, and fully connected layer structures are designed. The preprocessed data is input into the network. The convolutional kernels in the convolutional layer extract local features from the data, the pooling layer reduces the dimension, and the fully connected layer integrates the features to output the extracted index feature vectors, such as the abnormal features of the power supply network and the fault features of standby trains, etc. Thus, key index features can be efficiently and accurately extracted from complex feedback data, reducing the data processing volume, highlighting the important information in the data, providing strong support for subsequent predicting the changing trend of resource status, improving the efficiency of resource status analysis. The extracted index feature vectors are used as the input of the long short-term memory network, providing a basis for learning data features and predicting the changing trend of resource status, and affecting the accuracy of the prediction results.

[0059] The changes in resource status have time series characteristics. The long short-term memory network can effectively process time series data, learn data features, and predict the changing trend. According to the prediction results, the resource status and related model parameters are updated, which can keep the resource status information real-time and accurate, providing a reliable basis for scheduling decisions. The index feature vectors extracted by the convolutional neural network are organized into a sequence in chronological order and input into the long short-term memory network. The long short-term memory network automatically learns the long-term and short-term dependence relationships in the data through the forget gate, input gate, and output gate, and predicts the changing trend of resource status in the next period of time. According to the prediction results, the resource status information is updated. For example, the device status predicted to have a fault is marked as abnormal, and at the same time, the attributes of the corresponding nodes in the resource topology map are updated, and the Bayesian network is retrained to adjust the state transition probability between nodes. Thus, the changing trend of resource status can be accurately predicted, the resource status and related model parameters can be updated in a timely manner, enabling early response to resource status changes, optimizing scheduling decisions, and improving the reliability and stability of train operation. The updated resource status information will affect subsequent train scheduling decisions, including the sub-strategy for determining resource status and the sub-strategy for constructing the evaluation vector, ensuring that the entire scheduling process is based on the latest and accurate resource status information.

[0060] The resampling strategy for heterogeneous data includes: Analyze the correlation relationships among track circuit signals, platform passenger flow images, and meteorological environment data through the association rule mining algorithm, perform anomaly detection on the correlation relationships based on statistical methods, and analyze the causes of the detected abnormal heterogeneous data. According to the results of anomaly detection and cause analysis, determine the adjustment mechanism of the sampling frequency, and select the sampling method according to the type of abnormal heterogeneous data. Perform quality assessment on the resampled heterogeneous data to determine whether to resample the heterogeneous data, and fuse the resampled heterogeneous data with the train data matrix to update the train data matrix.

[0061] There are internal correlations among heterogeneous data. Analyzing the correlation relationships helps to discover data patterns. Abnormal heterogeneous data will affect the accuracy of scheduling decisions. Through anomaly detection and cause analysis, problem data can be discovered in a timely manner and the reasons can be found to ensure data quality. Use association rule mining algorithms such as Apriori to analyze track circuit signals, platform passenger flow images, and meteorological environment data, mine frequent item sets and association rules among the data, and use statistical methods (such as the 3σ principle and box plot method) to perform anomaly detection on the correlation relationships to determine whether the data deviates from the normal pattern. For the detected abnormal heterogeneous data, combine information such as data collection time, equipment status, and external environment to conduct cause analysis to determine whether it is caused by data collection errors, equipment failures, or actual situation changes; thus, gain an in-depth understanding of the correlation relationships among heterogeneous data, discover abnormal data in a timely manner and clarify the reasons, avoid abnormal data from misleading scheduling decisions, improve the reliability and effectiveness of the data, and provide guarantee for accurate scheduling. According to the results of anomaly detection and cause analysis, determine whether it is necessary to adjust the sampling strategy, provide a basis for determining the sampling frequency adjustment mechanism and selecting the sampling method in the follow-up, and ensure the pertinence and effectiveness of the resampling operation.

[0062] Different abnormal situations and data types have different requirements for sampling frequency and method. Reasonably adjusting the sampling frequency and selecting the sampling method can improve the data collection efficiency and reduce the data processing cost while ensuring data quality. If the anomaly is caused by large data fluctuations or important events, increase the sampling frequency. If it is a data collection error, adjust the sampling frequency according to the actual situation after correcting the error. For continuous data, such as track circuit signal strength, use equal-interval sampling or stratified sampling. For discrete data, such as the number of platform passengers, use random sampling or systematic sampling. For image data, such as platform passenger flow images, use feature-based sampling methods; thus, achieve the adaptive adjustment of the sampling strategy, select the optimal sampling method and frequency for different situations, improve the accuracy and efficiency of data acquisition, obtain more valuable data, and meet the data requirements of train scheduling. The quality of the resampled data directly affects subsequent quality assessment and data fusion. Appropriate sampling methods and frequencies can ensure the reliability of the resampled data and provide high-quality data for updating the train data matrix.

[0063] There are still quality problems with the resampled data. Conducting quality assessment can ensure that the data meets the requirements. Fusing the qualified resampled data with the train data matrix and updating the matrix information can make the train data matrix always reflect the latest and accurate data, providing a reliable basis for dispatching decisions. Establish a data quality assessment index system to evaluate the resampled data from aspects such as data integrity, accuracy, consistency, and timeliness. Use methods such as the entropy weight method to determine the weights of each index, calculate the comprehensive data quality score, and set a quality threshold. If the comprehensive score is less than the quality threshold, perform the resampling operation again. If it is greater than the quality threshold, fuse the resampled data with the train data matrix and update the corresponding data in the matrix. Thus, ensure the quality of the resampled data, prevent low-quality data from entering the train data matrix and affecting dispatching decisions, update the train data matrix in a timely manner, make dispatching decisions based on the latest and accurate data, improve the scientificity and accuracy of dispatching decisions. The updated train data matrix will be used for the construction of subsequent evaluation vectors and the generation of dispatching plans, affecting the accuracy and effectiveness of the entire train dispatching process.

[0064] Embodiment 2 The embodiment of the present application provides a train dynamic resource allocation system, which includes a resource perception module, an elastic allocation module, and a feedback control module.

[0065] The resource perception module is used to obtain heterogeneous data in real time. The heterogeneous data includes track circuit signals, platform passenger flow images, and meteorological environment data. The heterogeneous data is aligned in time and space through an on-vehicle edge node and processed. At the same time, the load rate of the power supply network, the equipment health status and location of standby trains, and the failure probability of signal lights are obtained to determine the resource status, and a train data matrix is output at the same time.

[0066] In a large railway hub, sensors deployed along the track and on-vehicle train equipment work continuously. The track circuit signal sensors collect signal data multiple times per second, high-definition cameras capture platform passenger flow images in real time, meteorological monitoring equipment synchronously collects meteorological environment data such as wind speed, rainfall, and temperature. At the same time, current and voltage sensors installed at key nodes of the power supply network monitor the load rate in real time, various sensors on standby trains collect equipment health status data, the positioning system provides real-time feedback on the train's location, and the built-in detection module of the signal light monitors the failure probability.

[0067] After these data are transmitted to the on-vehicle edge node, the spatio-temporal attention mechanism within the node starts to operate. Taking a train approaching a station as an example, the on-vehicle edge node receives the circuit signal of the current track of the train, the real-time passenger flow image of the platform, and the meteorological data at that time. The spatio-temporal attention mechanism quickly analyzes the timestamps and spatial positions of the data, precisely aligns different types of data in the spatio-temporal dimension, and normalizes the data, laying a foundation for subsequent in-depth data analysis. At the same time, it outputs a train data matrix containing resource status.

[0068] The elastic allocation module is used to construct an evaluation vector for train scheduling based on the train data matrix and the scheduling knowledge graph, dynamically adjust the dimension weights of the evaluation vector based on the fuzzy analytic hierarchy process, generate the train priority, generate a scheduling plan through reinforcement learning, dynamically allocate the power quotas of each section in the power supply network according to the train priority in the scheduling plan, and generate a resource allocation instruction in combination with the resource status.

[0069] After the resource allocation system obtains the train data matrix output by the resource perception module, it first normalizes it to ensure the comparability of different types of data within the matrix. For example, it maps the signal strength, frequency, and phase data in the signal feature vector, as well as the passenger flow density coefficient and meteorological correction parameters, etc., to a specific interval. Then, using data association technology, it establishes connections between the various data in the train data matrix and the relevant entities and relationships in the scheduling knowledge graph. For example, according to the train operation line, it associates information such as the historical scheduling cases and equipment maintenance records of this line in the knowledge graph.

[0070] Combining the signal feature vector and the resource status in the train data matrix, the resource allocation system deeply analyzes the train operation status. If the signal feature vector shows that a certain track circuit signal is abnormal, and the resource status indicates that the probability of a fault in the signal machine in this area is increasing, the resource allocation system extracts similar historical scheduling data from the scheduling knowledge graph, evaluates the late arrival risk value of the train in the current state, and at the same time calculates the passenger impact value caused by the train's late arrival based on the passenger flow density coefficient and the platform carrying capacity. For example, when the passenger flow density coefficient of a certain platform shows a peak state and the platform carrying capacity is close to saturation, if the train is late, the passenger impact value will increase significantly. Finally, the late arrival risk value and the passenger impact value are weighted and summed according to the preset weights to obtain the emergency degree factor of the train.

[0071] Based on the energy consumption and meteorological correction parameters, combined with the empirical consumption relationship of energy consumption under different meteorological conditions in the scheduling knowledge graph, the resource allocation system estimates the energy consumption cost of the train. For example, in strong wind weather, according to the meteorological correction parameters and energy consumption, the additional energy consumption cost is calculated. At the same time, the resource costs such as equipment maintenance and manpower are summarized to obtain the operating cost of the train. Then, according to the passenger flow density coefficient and the ticket sales law under different passenger flow conditions in the scheduling knowledge graph, the operating revenue value of the train for this trip is predicted. The economic value is obtained by subtracting the operating cost from the operating revenue value and is normalized to generate an economic value factor.

[0072] The resource allocation system combines the passenger flow density coefficient and resource status in the train data matrix to analyze the load situation of the line where the train is located. If the passenger flow density coefficient is high and the resource status shows that the equipment load on the line is large, the line load increases. At the same time, referring to the historical load data in the scheduling knowledge graph, the current line load value is evaluated. Then, according to the line importance index in the knowledge graph, such as the economic status of the cities connected by the line and the proportion of passenger flow, the line load value and line importance are weighted and combined to obtain a line impact factor.

[0073] Integrate the calculated emergency degree factor, economic value factor and line impact factor to construct an evaluation vector for train scheduling. Then, use the fuzzy analytic hierarchy process. Take the train priority as the target layer and the evaluation vector as the criterion layer. According to experience and actual situation, the importance between each factor in the criterion layer is compared pairwise, and the comparison results are characterized by fuzzy numbers to determine the fuzzy judgment matrix. After performing consistency test and correction on the fuzzy judgment matrix, solve the fuzzy characteristic equation to obtain the fuzzy weight vector of the evaluation vector, and then perform defuzzification to obtain the dimensional weight of the evaluation vector. Finally, perform weighted summation of the evaluation vector and the dimensional weight to obtain the comprehensive evaluation value of the train. Generate the train priority according to the comprehensive evaluation value. For example, through calculation, the comprehensive evaluation value of a certain train is relatively high, and its priority is set to a higher level in the current scheduling task.

[0074] Take the generated train priority and resource status as the state space of reinforcement learning, define operations such as train speed adjustment, train path selection and resource allocation as the action space, and use the time punctuality rate, energy consumption reduction rate and conflict risk value as the reward function to construct reinforcement learning. For example, when a certain train has a high priority and the resource status of the front line permits, the reinforcement learning can choose to adjust the train speed to improve the punctuality rate as an action and give corresponding rewards according to the actual effect.

[0075] The deep Q-network is trained using the constructed state space and action space. During the training process, the resource allocation system stores the states, actions, rewards, and next states in each train scheduling decision-making process in the experience pool, randomly extracts a batch of data from the experience pool at regular intervals, calculates the target Q-value according to the Bellman equation, and then updates the parameters of the deep Q-network based on the mean squared error loss function. After multiple iterative trainings, the deep Q-network can accurately evaluate the values of different actions in various states.

[0076] For the current train operation state, the deep Q-network selects the action with the largest target Q-value from the action space. If the action corresponding to the largest Q-value is to adjust the running path of a certain train to avoid congested sections, a corresponding scheduling plan is generated according to this action, including train path adjustment and resource allocation instructions. In terms of resource allocation, according to the train priorities in the scheduling plan, the power quotas of each section in the power supply network are dynamically allocated. For trains with high priorities, the power supply of the line sections where they are located is preferentially guaranteed. At the same time, resource allocation instructions such as spare train deployment and signal setting are generated according to the resource status to ensure the reasonable allocation of train operation resources.

[0077] The feedback control module is used to perform Monte Carlo simulations on each scheduling plan in parallel through the GPU, verify the conflict probability, energy consumption deviation, and time delay margin of each scheduling plan based on the threshold, so as to output the scheduling instruction set after passing the verification, execute the scheduling instruction set and monitor the execution effect of the scheduling instruction set in real time to determine whether to resample heterogeneous data and update the resource status.

[0078] The resource allocation system performs Monte Carlo simulations on each scheduling plan generated by the elastic allocation module through the GPU parallel computing architecture, determines the initial thresholds of the conflict probability, energy consumption deviation, and time delay margin based on historical operation data, and uses the long short-term memory network to perform time series analysis on the relevant indicators of each scheduling plan. At the same time, combined with real-time meteorological environment data, passenger flow density coefficient, and resource status, the thresholds are dynamically adjusted through the attention mechanism.

[0079] During the simulation process, the fuzzy comprehensive evaluation method is used to perform fuzzy matching on the conflict probability, energy consumption deviation, and time delay margin of each scheduling plan obtained from the Monte Carlo simulation with the threshold, set membership functions for each indicator, determine the evaluation results of the scheduling plan through the fuzzy composition operator to synthesize the membership functions, and compare with the evaluation threshold to judge whether the scheduling plan passes the verification and output the scheduling instruction set after passing the verification.

[0080] After executing the scheduling instruction set, the system for real-time monitoring of resources during the execution process feeds back data, including the operation data of the power supply network, standby trains, and signal lights, as well as the execution status data of the scheduling instruction set. The convolutional neural network is used to extract the index features of the feedback data, and then the long short-term memory network is used to learn these features to predict the change trend of the resource status. According to the prediction results, the resource status is updated, and at the same time, the node attributes of the topology map, the parameters of the Bayesian network, and the state transition probability are updated.

[0081] For the implementation principle of the above train dynamic resource allocation system, reference can be made to the specific embodiments of the train dynamic scheduling method, which will not be elaborated here one by one.

Claims

1. A method for dynamic train dispatching, characterized in that Including: Heterogeneous data is obtained in real time. The heterogeneous data includes track circuit signals, platform passenger flow images, and meteorological environment data. At the on-vehicle edge node, the heterogeneous data is aligned in space and time through a spatio-temporal attention mechanism, and the heterogeneous data is processed. At the same time, the power supply network, spare trains, and signal lights are monitored to determine the resource status, so as to output a train data matrix. The train data matrix includes signal feature vectors, passenger flow density coefficients, meteorological correction parameters, and resource status; An evaluation vector for train scheduling is constructed based on the train data matrix and the scheduling knowledge graph. The evaluation vector includes an urgency factor, an economic value factor, and a line impact factor. The dimension weights of the evaluation vector are dynamically adjusted based on the fuzzy analytic hierarchy process to generate a train priority; A scheduling plan is generated through reinforcement learning. The state space of the reinforcement learning represents the train priority and the resource status, and the reward function of the reinforcement learning represents the time punctuality rate, the energy consumption reduction rate, and the conflict risk value; Monte Carlo simulation is performed on each scheduling plan through GPU parallelism. Based on the threshold, the conflict probability, energy consumption deviation, and delay margin of each scheduling plan are verified to output a scheduling instruction set after verification. The scheduling instruction set includes train path adjustment and resource allocation instructions, and the scheduling instruction set is executed and the feedback data during the execution of the scheduling instruction set is monitored in real time to determine whether to resample the heterogeneous data and update the resource status.

2. The train dynamic scheduling method according to claim 1, wherein The determination sub-strategy of the resource status includes: Sensors are deployed to monitor the power supply network, spare trains, and signal lights in real time to obtain resource data. The resource data includes the load rate of the power supply network, the equipment health and location of the spare trains, and the failure probability of the signal lights; Taking the power supply network, spare trains, and signal lights as nodes, a topology graph is established. The node attributes represent the resource data, and the edges represent the coupling relationship between the resource data. The node feature vectors of the topology graph are extracted through a graph convolutional network; A Bayesian network is embedded in the topology graph. The node feature vectors are input into the Bayesian network, and the Bayesian network is trained through historical operation data to determine the state transition probability between nodes, and the state transition probability between nodes is integrated to determine the resource status.

3. The train dynamic scheduling method according to claim 2, characterized in that The output strategy of the train data matrix includes: Heterogeneous data is obtained in real time. The heterogeneous data includes track circuit signals, platform passenger flow images, and meteorological environment data. At the on-vehicle edge node, the heterogeneous data is aligned in space and time through a spatio-temporal attention mechanism, and the heterogeneous data is normalized; Feature extraction is performed on the track circuit signals through a residual convolutional neural network, and the extracted features are mapped into signal feature vectors. The signal feature vectors include signal intensity, signal frequency, and signal phase; Semantic segmentation is performed on the platform passenger flow image, the ratio of the passenger pixel area to the total platform area is calculated to obtain the initial passenger flow density, and the passenger flow change trend is predicted through a long short-term memory network in combination with historical passenger flow data to dynamically weight and correct the initial passenger flow density to generate a passenger flow density coefficient; Analyze the influence degree of different meteorological environment data on train operation data, braking distance and energy consumption through the random forest algorithm, assign influence weights to each meteorological environment data according to the analysis results, and perform weighted summation of the standardized meteorological environment data and the influence weights to generate meteorological correction parameters; Monitor the power supply network, spare trains and signal lights to determine the resource status, integrate the signal feature vector, passenger flow density coefficient, meteorological correction parameters and resource status, and output the train data matrix.

4. The train dynamic scheduling method according to claim 3, wherein, The construction sub-strategy of the evaluation vector includes: Normalize the train data matrix and associate the train data matrix with the dispatching knowledge graph; Analyze the train operation status by combining the signal feature vector and resource status in the train data matrix, extract the historical dispatching data of the dispatching knowledge graph to evaluate the late arrival risk value under the train operation status, and evaluate the passenger impact value under train late arrival according to the passenger flow density coefficient and platform carrying capacity. Perform weighted summation of the late arrival risk value and the passenger impact value to obtain the urgency factor of the train; According to the energy consumption and meteorological correction parameters in the train data matrix, combine the dispatching knowledge graph to determine the energy consumption cost of the train, combine the resource cost to obtain the operation cost of the train, and determine the operation revenue value according to the passenger flow density coefficient and the dispatching knowledge graph. Combine the operation revenue value and the operation cost to determine the economic value, and normalize the economic value to obtain the economic value factor of the train; Analyze the train line load by combining the passenger flow density coefficient and resource status in the train data matrix, combine the historical load data of the dispatching knowledge graph to evaluate the line load value, and perform weighted combination of the line load value and the line importance in the dispatching knowledge graph to obtain the line influence factor; Integrate the urgency factor, economic value factor and line influence factor to construct an evaluation vector for train dispatching.

5. The train dynamic scheduling method according to claim 4, wherein The generation strategy of the train priority includes: Use the fuzzy analytic hierarchy process to take the train priority as the target layer and the evaluation vector as the criterion layer. Characterize the importance between the criterion layers by fuzzy numbers and determine the fuzzy judgment matrix; Conduct consistency test and correction on the fuzzy judgment matrix, solve the fuzzy eigen-equation to obtain the fuzzy weight vector of the evaluation vector, and perform defuzzification processing on the fuzzy weight vector to obtain the dimension weight of the evaluation vector; Perform weighted summation of the evaluation vector and the dimension weight to obtain the comprehensive evaluation value of the train, and generate the train priority according to the comprehensive evaluation value of the train.

6. The train dynamic scheduling method according to claim 5, characterized in that The generation strategy of the dispatching plan includes: Construct reinforcement learning. The state space of reinforcement learning represents the train priority and resource status, the action space of reinforcement learning represents train speed adjustment, train path selection and resource allocation, and the reward function of reinforcement learning represents the time punctuality rate, energy consumption reduction rate and conflict risk value; Train the deep Q network through the state space and action space, calculate the target Q value, and update the parameters of the deep Q network based on the mean square error loss function; Select the action with the largest target Q value through the deep Q network, generate a dispatching plan according to the action, and evaluate the generated dispatching plan through the time punctuality rate, energy consumption reduction rate and conflict risk value in the reward function to determine whether to adjust the reinforcement learning.

7. The train dynamic dispatching method according to claim 6, wherein The verification sub-strategies of the scheduling instruction set include: Construct a GPU parallel computing architecture and decompose the tasks of Monte Carlo simulation into three levels of sub-tasks; Determine the threshold based on historical operation data, and perform time series analysis on the conflict probability, energy consumption deviation, and latency margin of each scheduling scheme through a long short-term memory network; Dynamically adjust the threshold through the attention mechanism in combination with meteorological environment data, passenger flow density coefficient, and resource status; Perform fuzzy matching between the conflict probability, energy consumption deviation, and latency margin of each scheduling scheme obtained from Monte Carlo simulation and the threshold through the fuzzy comprehensive evaluation method; Set membership functions for the conflict probability, energy consumption deviation, and latency margin, determine the evaluation result of the scheduling scheme through the fuzzy composition operator to synthesize the membership functions, compare the evaluation result with the evaluation threshold to judge whether the scheduling scheme passes the verification, and output the scheduling instruction set after passing the verification.

8. The train dynamic scheduling method according to claim 7, wherein The update strategy of the resource status includes: Execute the scheduling instruction set and monitor the feedback data during the execution process of the scheduling instruction set in real time. The feedback data includes the operation data of the power supply network, standby trains, and signal lights, as well as the execution status data of the scheduling instruction set; Extract the index features of the feedback data through a convolutional neural network; Learn the index features of the feedback data through a long short-term memory network to predict the change trend of the resource status, update the resource status, and at the same time update the node attributes of the topology graph, the parameters of the Bayesian network, and the state transition probability.

9. The train dynamic scheduling method according to claim 8, wherein, The resampling strategy of the heterogeneous data includes: Analyze the correlation relationship between track circuit signals, platform passenger flow images, and meteorological environment data through the association rule mining algorithm, perform anomaly detection on the correlation relationship based on statistical methods, and analyze the reasons for the detected abnormal heterogeneous data; Determine the adjustment mechanism of the sampling frequency according to the results of anomaly detection and cause analysis, and select the sampling method according to the type of abnormal heterogeneous data; Perform quality evaluation on the resampled heterogeneous data to judge whether to resample the heterogeneous data, and fuse the resampled heterogeneous data with the train data matrix to update the train data matrix.

10. The train dynamic resource allocation system is characterized in that, Include: Obtain heterogeneous data in real time. The heterogeneous data includes track circuit signals, platform passenger flow images, and meteorological environment data. Align the heterogeneous data in time and space through in-vehicle edge nodes, process the heterogeneous data, and at the same time obtain the load rate of the power supply network, the equipment health status and location of standby trains, and the failure probability of signal lights to determine the resource status, and output the train data matrix at the same time; Construct an evaluation vector for train scheduling based on the train data matrix and the scheduling knowledge graph, dynamically adjust the dimension weights of the evaluation vector based on the fuzzy analytic hierarchy process, generate the train priority, generate a scheduling scheme through reinforcement learning, dynamically allocate the power quota of each section in the power supply network according to the train priority in the scheduling scheme, and generate a resource allocation instruction in combination with the resource status. Perform Monte Carlo simulation for each scheduling scheme in parallel through the GPU, verify the conflict probability, energy consumption deviation, and latency margin of each scheduling scheme based on the threshold, so as to output the scheduling instruction set after passing the verification, execute the scheduling instruction set and monitor the execution effect of the scheduling instruction set in real time to determine whether to resample heterogeneous data and update the resource status.

Images