A high-speed train assisted driving control method and system

Abstract

The present invention discloses a high-speed train assisted driving control method and system. The method includes establishing an end-edge-cloud collaborative system for the high-speed train operation process, constructing a speed and level mechanism characteristic model based on dynamic analysis of the speed change mechanism characteristics and level change mechanism characteristics of the high-speed train operation process; determining an LSTM-based high-speed train assisted driving control strategy prediction model based on train and line data obtained from the collaborative system, the speed and level mechanism characteristic model, and an LSTM neural network structure; and determining the high-speed train assisted driving control strategy using an LSTM-based high-speed train assisted driving control strategy prediction model downloaded from the cloud layer and established offline based on a large amount of historical data, combining the real-time collected data with the data obtained by the end layer. The LSTM-based high-speed train assisted driving control strategy prediction model is then used to determine the high-speed train assisted driving control strategy, and the strategy is sent to an on-board edge computing node via a 5G network for assisted driving control. The present invention can improve the accuracy of assisted driving control strategy prediction during high-speed train operation.

Description

Technical Field

[0001] The present invention relates to the technical field of assisted driving control, and in particular to a high-speed train assisted driving control method and system. Background Art

[0002] High-speed train assisted driving systems are crucial for ensuring smooth and high-speed train operation and are a core technology for the future development of intelligent high-speed railways. Real-time and accurate prediction of control strategies is paramount in high-speed train driving tasks, whether for assisted driving systems or higher-level intelligent driving systems. However, with the rapid development of high-speed trains, the information available to drivers through the human-machine interface (HMI) during operation is limited, making it difficult to effectively respond to unknown internal and external disturbances. Furthermore, due to the complex and ever-changing operating environment of high-speed trains, the operation process largely depends on the driver's personal experience, expertise, and work ability. If drivers have varying levels of skill, their performance will inevitably vary significantly when faced with the same operational task, impacting train safety, punctuality, comfort, and stopping accuracy. Without considering the train's own state and the driver's state, and without accurate prediction of control strategies, train control is impossible. Enabling high-speed train assisted driving systems to accurately understand the driver's state and intelligently predict control strategies is a major challenge in the development of future train control systems.

[0003] To model the high-speed train operating process, researchers have proposed mechanism analysis and data-driven prediction methods. Because the driver's traction and brake lever position changes are highly nonlinear, current control input models often ignore certain factors in the train's operation, making it difficult to accurately and reliably describe the complex operating characteristics of high-speed trains. To address the current challenges in high-speed train operation, some researchers have proposed using data-driven modeling methods for edge computing to build more accurate models. Although these studies all employ edge-cloud collaboration, their edge computing-based algorithm models are relatively simple, which to some extent limits their practical application. With the development of artificial intelligence, deep learning has gained attention, and long-short-term memory (LSTM) networks are particularly well-suited for processing complex time series data. Researchers have used LSTM to analyze big data features, leveraged edge computing for parallel computing, and improved the efficiency of industrial electrical equipment identification. They have also used LSTM networks to predict short-term subway passenger and traffic flows. However, due to the diverse and complex nature of high-speed train operating data, these traditional LSTM network models are difficult to directly apply to predicting assisted driving control strategies during high-speed train operation. Summary of the Invention

[0004] The purpose of the present invention is to provide a high-speed train assisted driving control method and system, which can improve the accuracy of assisted driving control strategy prediction during high-speed train operation.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] A high-speed train assisted driving control method, comprising:

[0007] Construct an end-edge-cloud collaborative system for the high-speed train operation process; the end-edge-cloud collaborative system for the high-speed train operation process includes an end layer for collecting, storing and transmitting mileage and slope; an edge layer for collecting, storing and transmitting train traction braking force, train traction braking level and train running speed, and online determining the high-speed train assisted driving control strategy based on the acquired data and a downloaded high-speed train operation process prediction model; and a cloud layer for offline constructing a high-speed train operation process model based on a large amount of historical data for download by the edge layer, and receiving and backing up working condition information and processing results from the end layer and the edge layer for later use; the working condition information includes: train running status information and line status information; the processing results include train running speed prediction results and train traction braking level prediction results;

[0008] Based on the dynamic analysis of the speed change mechanism characteristics and the level change mechanism characteristics during the operation of high-speed trains, a speed and level change mechanism characteristic model is constructed;

[0009] Based on the train and line data, speed and level mechanism characteristic model, and LSTM neural network structure obtained from the end-edge-cloud collaborative system of the high-speed train operation process, a LSTM-based high-speed train assisted driving control strategy prediction model is determined;

[0010] The edge layer combines the data collected in real time with the data obtained by the terminal layer, and uses the LSTM-based high-speed train assisted driving control strategy prediction model downloaded from the cloud layer based on a large amount of historical data and established offline to determine the high-speed train assisted driving control strategy, and sends it to the on-board edge computing node through the 5G network for assisted driving control.

[0011] Optionally, the speed and level change mechanism characteristic model is constructed based on the dynamic analysis of the speed change mechanism characteristic and the level change mechanism characteristic during the high-speed train operation process, specifically including the following formula:

[0012] ;

[0013] In the formula, s, e, t, and m are the train speed, train traction braking level, time, and line mileage, respectively. is the total mass of the train, and It is the train traction handle level and brake handle level, and are the marker coefficients of train traction and braking, and The train's traction and braking forces, and , s and related, Indicates the bending resistance; represents air resistance; The train handle level characteristics, , , and related.

[0014] Optionally, determining a high-speed train assisted driving control strategy prediction model based on LSTM based on train and line data, speed and level mechanism characteristic model, and LSTM neural network structure obtained from the end-edge-cloud collaborative system of the high-speed train operation process specifically includes:

[0015] Using the formula Determine the forget gate information update process;

[0016] Using the formula Determine the input gate information update process;

[0017] Using the formula Determine the current cell renewal process;

[0018] Using the formula Determine the output gate information update process;

[0019] Where, Output of the forget gate; is the activation function; is the previous hidden state, is the current input; and are the weight and bias from the input to the forget gate, Output of the input gate sigmoid function; Output of the input gate tanh function; and is the weight of the input to the input gate, and is the bias from the input to the input gate, is the current cell state; is the previous cell state, Output of the output gate sigmoid function; is the current hidden state; and are the weight and bias of the input to output gate respectively.

[0020] Optionally, the LSTM-based high-speed train assisted driving control strategy prediction model is:

[0021] ;

[0022] Where, The output is the train running speed and the train traction brake level. For input Same output The nonlinear relationship between the input It is the train running speed, train traction braking level, line slope, line mileage, train traction force and braking force.

[0023] Optionally, the method further includes determining an LSTM-based high-speed train assisted driving control strategy prediction model based on train and line data, speed and level mechanism characteristic models, and an LSTM neural network structure obtained from the end-edge-cloud collaborative system of the high-speed train operation process, and further includes:

[0024] Using the formula Determine the root mean square error (RMSE) of the LSTM-based high-speed train assisted driving control strategy prediction model under different inputs;

[0025] Where, is the total number of test data; and They are The predicted value at the moment and its corresponding actual value.

[0026] A high-speed train auxiliary driving control system, comprising:

[0027] A collaborative system construction module is used to build an end-edge-cloud collaborative system for the high-speed train operation process; the end-edge-cloud collaborative system for the high-speed train operation process includes an end layer for collecting, storing and transmitting mileage and slope, an edge layer for collecting, storing and transmitting train traction braking force, train traction braking level and train running speed, and online determining the high-speed train assisted driving control strategy based on the acquired data and the downloaded high-speed train operation process prediction model; and a cloud layer for offline building a high-speed train operation process model based on a large amount of historical data for download by the edge layer, and receiving and backing up working condition information and processing results from the end layer and the edge layer for later call; the working condition information includes: train running status information and line status information; the processing results include train running speed prediction results and train traction braking level prediction results;

[0028] The speed and level mechanism characteristic model construction module is used to construct the speed and level mechanism characteristic model based on the dynamic analysis of the speed change mechanism characteristics and level change mechanism characteristics during the high-speed train operation process;

[0029] An LSTM-based high-speed train assisted driving control strategy prediction model construction module is used to determine the LSTM-based high-speed train assisted driving control strategy prediction model based on train and line data obtained from the end-edge-cloud collaborative system of the high-speed train operation process, the speed and level mechanism characteristic model, and the LSTM neural network structure;

[0030] The assisted driving control module is used to combine the real-time collected data at the edge layer with the data obtained by the terminal layer. It uses the LSTM-based high-speed train assisted driving control strategy prediction model established offline based on a large amount of historical data downloaded from the cloud layer to determine the high-speed train assisted driving control strategy, and sends it to the on-board edge computing node through the 5G network for assisted driving control.

[0031] A high-speed train assisted driving control system comprises: at least one processor, at least one memory and computer program instructions stored in the memory, and the method described is implemented when the computer program instructions are executed by the processor.

[0032] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0033] This paper addresses the issues of limited access to operational instructions for high-speed train drivers, significant influence of the driver's individual skills and experience during operation, and traditional predictions that only consider train speed. By developing an end-to-end, edge-to-cloud collaborative system for high-speed train operation, this paper also combines mechanism-based and data-driven modeling techniques to construct a LSTM-based prediction model for high-speed train assisted driving control strategies. Furthermore, the paper analyzes the effectiveness and feasibility of this LSTM-based prediction model based on field data, overcoming the blindness of traditional assisted driving control strategy settings. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0035] Figure 1 A schematic flow chart of a high-speed train assisted driving control method provided by the present invention;

[0036] Figure 2 This is a framework diagram of the end-edge-cloud collaborative system for the high-speed train operation process;

[0037] Figure 3 This is the internal structure diagram of the LSTM network;

[0038] Figure 4 Schematic diagram of data preprocessing process;

[0039] Figure 5 This is a diagram of the training and testing process of the LSTM high-speed train assisted driving control strategy prediction model;

[0040] Figure 6 This is a comparison chart of the speed prediction value and the actual value of Scheme 1;

[0041] Figure 7 This is a comparison chart of the speed prediction value and the actual value of Scheme 2;

[0042] Figure 8 This is a comparison chart of the speed prediction value and the actual value of Scheme 3;

[0043] Figure 9 This is a comparison chart of the speed prediction value and the actual value of Scheme 4;

[0044] Figure 10 This is a comparison chart of the speed prediction value and the actual value of Scheme 5;

[0045] Figure 11 Comparison of speed prediction errors under different input schemes;

[0046] Figure 12 This is a comparison chart between the predicted and actual values of the traction brake level in Scheme 1;

[0047] Figure 13 This is a comparison chart between the predicted and actual values of the traction brake level in Scheme 2;

[0048] Figure 14 Comparison chart of traction brake level prediction error under different input schemes. DETAILED DESCRIPTION

[0049] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0050] The purpose of the present invention is to provide a high-speed train assisted driving control method and system to solve the problem that the driver has limited information available during the operation of the high-speed train and cannot effectively deal with the influence of internal and external unknown disturbances, which may affect the safety, punctuality and comfort of the train operation process; and the high-speed train operation environment is complex and changeable, and the operation process depends largely on the driver's personal experience, professional knowledge, and work ability. Most existing models can only predict the system output, and the real-time prediction of the assisted driving control strategy has not yet been solved.

[0051] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0052] Figure 1 This is a flow chart of a high-speed train assisted driving control method provided by the present invention, as shown in FIG. Figure 1 As shown, the present invention provides a high-speed train assisted driving control method, comprising:

[0053] S101, constructing an end-edge-cloud collaborative system for the high-speed train operation process; the end-edge-cloud collaborative system for the high-speed train operation process includes an end layer for collecting, storing and transmitting mileage and slope, an edge layer for collecting, storing and transmitting train traction braking force, train traction braking level and train running speed, and online determining the high-speed train assisted driving control strategy based on the acquired data and the downloaded high-speed train operation process prediction model, and a cloud layer for offline building a high-speed train operation process model based on a large amount of historical data for download by the edge layer, and receiving and backing up working condition information and processing results from the end layer and the edge layer for later call; the working condition information includes: train running status information and line status information; the processing results include train running speed prediction results and train traction braking level prediction results;

[0054] Among them, by analyzing the multivariable time-varying nonlinear hysteresis characteristics of the high-speed train operation process, a mechanism combining edge computing and cloud big data preprocessing is adopted to construct an end-edge-cloud collaborative system for the high-speed train operation process, realizing real-time transmission and efficient processing of effective information on the control input side.

[0055] like Figure 2As shown, the end layer includes temporary speed limit nodes (i.e., transponder transmission modules) and track section nodes (including line mileage and line slope). Due to the static nature of mileage and slope, the slope and mileage data for different sections can be pre-stored in the central cloud and the edge cloud covering that section. Furthermore, information such as line slope and mileage is also typically stored in the transponder. In the event of an emergency, when the train passes over the transponder, the transponder antenna and transponder transmission module transmit this information to the train, which then sends the relevant line information at the time of the emergency to the vehicle-edge cloud.

[0056] The edge layer includes vehicle-edge cloud nodes, onboard edge computing nodes, and train operation control dynamics nodes (including train traction braking force, train traction braking level, and train speed). Vehicle-edge cloud nodes are located near 5G base stations or trackside, transmitting and processing data in real time, reducing the burden on cloud computing. Onboard edge computing nodes analyze data in real time and provide critical instructions to the driver in real time.

[0057] The cloud layer includes multiple cloud servers.

[0058] S102: Based on the dynamic analysis of the speed change mechanism characteristics and the level change mechanism characteristics during the high-speed train operation process, a speed and level change mechanism characteristic model is constructed;

[0059] S102 specifically includes the following formula:

[0060] ;

[0061] In the formula, s, e, t, and m are the train speed, train traction braking level, time, and line mileage, respectively. is the total mass of the train, and It is the train traction handle level and brake handle level, and are the marker coefficients of train traction and braking, and The train's traction and braking forces, and , s and related, Indicates the curve resistance; represents air resistance; The train handle level characteristics, , , and related.

[0062] S103, according to the train and line data, speed and level mechanism characteristic model and LSTM neural network structure obtained from the end-edge-cloud collaborative system of the high-speed train operation process, determine the LSTM-based high-speed train assisted driving control strategy prediction model, and Figure 3 As shown;

[0063] S103 specifically includes:

[0064] Using the formula Determine the forget gate information update process;

[0065] Using the formula Determine the input gate information update process;

[0066] Using the formula Determine the current cell renewal process;

[0067] Using the formula Determine the output gate information update process;

[0068] Where, Output of the forget gate; is the activation function; is the previous hidden state, is the current input; and are the weight and bias from the input to the forget gate, Output of the input gate sigmoid function; Output of the input gate tanh function; and is the weight of the input to the input gate, and is the bias from the input to the input gate, is the current cell state; is the previous cell state, Output of the output gate sigmoid function; is the current hidden state; and are the weight and bias of the input to output gate respectively.

[0069] The LSTM-based high-speed train assisted driving control strategy prediction model is:

[0070] ;

[0071] Where, The output is the train running speed and the train traction brake level. For input Same output The nonlinear relationship between the input It is the train running speed, train traction braking level, line slope, line mileage, train traction force and braking force.

[0072] like Figure 3 As shown in the figure, it has a gating mechanism that controls the storage and update of information flow, including a forget gate, an input gate, and an output gate. During its information transmission process, the forget gate first determines which information to discard or retain. The information from the previous hidden state and the current input are simultaneously passed to the activation function of the "forget gate layer", resulting in an output ranging from 0 to 1, with 0 indicating complete discard and 1 indicating complete retention. Next, the input gate determines which new information to add to the cell state. The information from the previous hidden state and the current input are passed to the activation function of the "input gate layer" to determine which information to update, with 0 indicating unimportant and 1 indicating important. This output is then generated. This information is also passed through the hyperbolic tangent function to generate a candidate value vector. This output is then multiplied by the candidate value vector to determine which information to use to update the new cell state. Finally, the output gate is used to determine the output value, which is also the value of the current hidden state. The information from the previous hidden state and the current input information are passed to the activation function of the "output gate layer" to determine which information needs to be retained, and then an output quantity is obtained; at the same time, the new cell state is passed to the hyperbolic tangent function for processing, and then the two output quantities are multiplied to finally determine the value of the new hidden state.

[0073] Therefore, it can be seen that the LSTM network can control the hidden layer output through a gating mechanism, thereby controlling the convergence of the training gradient, effectively solving the gradient problem during training. It also has long-term memory, allowing cell states to be updated and transmitted in a straight line. Furthermore, LSTM uses the backpropagation algorithm as its training algorithm. During the model training phase, it can continuously update the network weights based on the training data, effectively extracting and memorizing the characteristics of time series data. During testing, simply inputting the data into the trained model can obtain the corresponding accurate prediction value.

[0074] S103, and later also includes:

[0075] Using the formula Determine the root mean square error (RMSE) of the LSTM-based high-speed train assisted driving control strategy prediction model under different inputs;

[0076] Where, is the total number of test data; and They are The predicted value at the moment and its corresponding actual value.

[0077] Among them, the different strategies include a one-input strategy that uses the line mileage at the previous moment to predict the train speed at the next moment; a two-input strategy that uses the line mileage and slope at the previous moment to predict the train speed at the next moment; a three-input strategy that uses the line mileage and slope at the previous moment, and the train's traction braking force to predict the train speed at the next moment; a four-input strategy that uses the line mileage and slope at the previous moment, the train's traction braking force, and the traction braking level to predict the train speed at the next moment; a five-input strategy that uses the line mileage and slope at the previous moment, the train's traction braking force, the traction braking level, and the running speed to predict the train speed at the next moment. The four-input strategy uses the line mileage and slope at the previous moment, the train's traction braking force, and the speed to predict the train's traction braking level; and the five-input strategy uses the line mileage and slope at the previous moment, the train's traction braking force, the running speed, and the traction braking level to predict the train's traction braking level.

[0078] In S104, the edge layer combines the data collected in real time with the data obtained by the terminal layer, and uses the LSTM-based high-speed train assisted driving control strategy prediction model downloaded from the cloud layer and established offline based on a large amount of historical data to determine the high-speed train assisted driving control strategy, and sends it to the on-board edge computing node through the 5G network for assisted driving control.

[0079] like Figure 4-Figure 14 As shown, this embodiment provides a method for analyzing the effectiveness of an LSTM control strategy prediction model based on high-speed railway field data:

[0080] A total of 104,448 data samples were collected from a CRH380B vehicle running on the same section of road. Each data sample record contained information related to the line and train, among which the train traction and braking level information was recorded and obtained by the image sensor.

[0081] Figure 4 The data preprocessing process for edge computing is demonstrated. To build an accurate high-speed train edge computing model, collected data requires preprocessing, including data cleaning, data transformation, data partitioning, and data standardization. Data cleaning is primarily performed on the onboard edge computing nodes, while the remaining processing is performed on the onboard edge cloud nodes. To reduce communication latency and energy consumption, these nodes exchange information via 5G wireless transmission modules.

[0082] Figure 5 This article demonstrates the training and testing process of an LSTM-based prediction model for high-speed train assisted driving control strategies. After the aforementioned data preprocessing, the data is divided into training data and test data. The training data is used to train the LSTM model, and the test data is used to verify the model's prediction performance.

[0083] Figure 6 、 Figure 7 、 Figure 8 、 Figure 9 and Figure 10 The results of speed prediction in high-speed train assisted driving control strategy under 5 different schemes are given. Figure 11 The corresponding absolute error of the prediction is shown. Figure 6 Prediction results for the common one-input LSTM model in Solution 1. Figure 7 Corresponding to Solution 2, based on Solution 1, it makes full use of the end-layer data and adopts a two-input LSTM model for prediction. Figure 8 This is the prediction result of Solution 3. Solution 3 is based on the end-edge-cloud collaborative learning architecture and uses a three-input LSTM model for prediction. Solution 4 is based on Solution 3 and considers the train traction and braking level factors. It uses a four-input LSTM model to achieve prediction. Figure 9 This is the prediction result of Scheme 4. Scheme 5 comprehensively considers the train speed and train traction braking level factors to obtain a five-input LSTM model to predict the train running speed. Figure 10 The corresponding prediction results are given. Figure 6 、 Figure 7 、 Figure 8 、 Figure 9 、 Figure 10 It can be seen that Figure 10 The training effect and fitting accuracy of the corresponding model are better than the previous four models, and the predicted value can better approach the true value, with higher prediction accuracy. Figure 11 , it can be seen that the prediction error of Scheme 5 is smaller than that of the other four prediction schemes. This shows that the LSTM-based high-speed train assisted driving control strategy prediction model proposed in this paper can comprehensively consider factors such as traction and braking level and speed during the train operation process, compared to traditional mechanism models or data-driven models, and can more accurately and effectively predict speed in the high-speed train assisted driving control strategy.

[0084] Figure 12 and Figure 13 The results of traction brake level prediction in high-speed train assisted driving control strategy under two different schemes are given. Figure 14 The corresponding absolute error of the prediction is shown. Figure 12 This is the prediction result of Solution 1, which is based on the common four-input LSTM traction and braking level prediction model. Solution 2 is based on the end-edge-cloud collaborative auxiliary learning architecture, comprehensively considering the train speed and train traction and braking level factors, and uses a five-input LSTM model for prediction. Figure 13 The prediction results are given. Figure 12 and Figure 13 , it can be seen that at most times, Figure 13The predicted values in are closer to the true values. Figure 14 It can also be seen that compared with Scheme 1, Scheme 2 can significantly reduce the prediction error. This shows that the use of image sensors to record the control commands of the train traction and brake operating handle proposed in the present invention is of great significance for the prediction of high-speed train assisted driving control strategies. It also shows that the LSTM-based high-speed train assisted driving control strategy prediction model proposed in the present invention can comprehensively consider factors such as the traction and brake level and speed during train operation, compared with traditional mechanism models or data-driven models, and can more accurately and effectively predict the traction and brake level in the high-speed train assisted driving control strategy.

[0085] In accordance with the above method, the present invention further provides a high-speed train assisted driving control system, comprising:

[0086] A collaborative system construction module is used to build an end-edge-cloud collaborative system for the high-speed train operation process; the end-edge-cloud collaborative system for the high-speed train operation process includes an end layer for collecting, storing and transmitting mileage and slope, an edge layer for collecting, storing and transmitting train traction braking force, train traction braking level and train running speed, and online determining the high-speed train assisted driving control strategy based on the acquired data and the downloaded high-speed train operation process prediction model; and a cloud layer for offline building a high-speed train operation process model based on a large amount of historical data for download by the edge layer, and receiving and backing up working condition information and processing results from the end layer and the edge layer for later call; the working condition information includes: train running status information and line status information; the processing results include train running speed prediction results and train traction braking level prediction results;

[0087] The speed and level mechanism characteristic model construction module is used to construct the speed and level mechanism characteristic model based on the dynamic analysis of the speed change mechanism characteristics and level change mechanism characteristics during the high-speed train operation process;

[0088] An LSTM-based high-speed train assisted driving control strategy prediction model construction module is used to determine the LSTM-based high-speed train assisted driving control strategy prediction model based on train and line data obtained from the end-edge-cloud collaborative system of the high-speed train operation process, the speed and level mechanism characteristic model, and the LSTM neural network structure;

[0089] The assisted driving control module is used to combine the real-time collected data at the edge layer with the data obtained by the terminal layer. It uses the LSTM-based high-speed train assisted driving control strategy prediction model established offline based on a large amount of historical data downloaded from the cloud layer to determine the high-speed train assisted driving control strategy, and sends it to the on-board edge computing node through the 5G network for assisted driving control.

[0090] In order to execute the method corresponding to the above-mentioned embodiment 1 to achieve the corresponding functions and technical effects, the present invention also provides a high-speed train assisted driving control system, including: at least one processor, at least one memory and computer program instructions stored in the memory, when the computer program instructions are executed by the processor, the described method is implemented.

[0091] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Reference can be made to the common and similar parts between the various embodiments. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method description.

[0092] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only intended to help understand the method and core concept of the present invention. At the same time, those skilled in the art will find that the specific implementation methods and application scopes may vary based on the concept of the present invention. In summary, the contents of this specification should not be construed as limiting the present invention.

Claims

1. A high-speed train assisted driving control method, characterized in that: include: Constructing an end-edge-cloud collaborative system for the high-speed train operation process; the end-edge-cloud collaborative system for the high-speed train operation process includes an end layer for collecting, storing, and transmitting mileage and grade; an edge layer for collecting, storing, and transmitting train traction braking force, train traction braking level, and train running speed, and determining the high-speed train assisted driving control strategy online based on the acquired data and a downloaded high-speed train operation process prediction model; and a cloud layer for offline constructing a high-speed train operation process model based on a large amount of historical data for download by the edge layer, and receiving and backing up working condition information and processing results from the end layer and the edge layer for later call; The working condition information includes: train running status information and line status information; the processing results include train running speed prediction results and train traction braking level prediction results; Based on the dynamic analysis of the speed change mechanism characteristics and the level change mechanism characteristics during the operation of high-speed trains, a speed and level change mechanism characteristic model is constructed; Based on the train and line data, speed and level mechanism characteristic model, and LSTM neural network structure obtained from the end-edge-cloud collaborative system of the high-speed train operation process, a LSTM-based high-speed train assisted driving control strategy prediction model is determined; The edge layer combines the data collected in real time with the data obtained by the terminal layer, and uses the LSTM-based high-speed train assisted driving control strategy prediction model established offline based on a large amount of historical data downloaded from the cloud layer to determine the high-speed train assisted driving control strategy, and sends it to the on-board edge computing node through the 5G network for assisted driving control; the high-speed train assisted driving control strategy includes the following: one input strategy uses the line mileage at the previous moment to predict the train running speed at the next moment; the second input strategy uses the line mileage and slope at the previous moment to predict the train speed at the next moment; the third input strategy uses the line mileage and slope at the previous moment, and the train traction The train speed at the next moment is predicted based on the braking force; the train speed at the next moment is predicted based on the line mileage and slope, train traction braking force and traction braking level at the previous moment under the four-input strategy; the train speed at the next moment is predicted based on the line mileage and slope, train traction braking force, traction braking level and running speed under the five-input strategy; the train speed at the next moment is predicted based on the line mileage and slope, train traction braking force and speed at the previous moment; the train traction braking level at the next moment is predicted based on the five-input strategy; The dynamic analysis of the speed change mechanism characteristics and the level change mechanism characteristics during the high-speed train operation process is used to construct a speed and level mechanism characteristic model, which specifically includes the following formulas: Where s, e, t, and m are respectively the train speed, train traction brake level, time, and line mileage. t is the total mass of the train, D r and D b k is the train traction handle level and brake handle level, r and k b The sign coefficients of train traction and braking are R and B train traction and braking force, respectively, and D r , s and D b G(m) represents the curve resistance; W(s) represents the air resistance; U is the train handle level characteristic, and k r , k b , D r and D b related.

2. A high-speed train assisted driving control method according to claim 1, characterized in that: The LSTM-based high-speed train assisted driving control strategy prediction model is determined based on the train and line data, speed and level mechanism characteristic model, and LSTM neural network structure obtained from the end-edge-cloud collaborative system of the high-speed train operation process, specifically including: Using formula f t =σ(W f ·[h t-1 , x t ]+b f ) Determine the forget gate information update process; Using the formula Determine the input gate information update process; Using formula c t =c t-1 .f t +i t .c′ t Determine the current cell renewal process; Using the formula Determine the output gate information update process; Where, f t is the output of the forget gate; σ is the activation function; h t-1 is the previous hidden state, x t is the current input; W f and b f are the weight and bias from the input to the forget gate, i t is the output of the input gate sigmoid function; c t ′ is the output of the input gate tanh function; W i and W c is the weight of the input to the input gate, b i and b c is the bias from the input to the input gate, c t is the current cell state; c t-1 is the previous cell state, o t is the output of the sigmoid function of the output gate; h t is the current hidden state; W o and b o are the weight and bias of the input to output gate respectively.

3. The high-speed train assisted driving control method according to claim 2, characterized in that: The LSTM-based high-speed train assisted driving control strategy prediction model is: AND a =V(X train ); Where Y a The output is the train speed and the train traction brake level, V is the input X train Same as output Y a The nonlinear relationship between the input X train It is the train running speed, train traction braking level, line slope, line mileage, train traction force and braking force.

4. The high-speed train assisted driving control method according to claim 1, characterized in that: The method further includes determining an LSTM-based high-speed train assisted driving control strategy prediction model based on train and line data, speed and level mechanism characteristic models, and an LSTM neural network structure obtained from the end-edge-cloud collaborative system of the high-speed train operation process, and further includes: Using the formula Determine the root mean square error (RMSE) of the LSTM-based high-speed train assisted driving control strategy prediction model under different inputs; Where n is the total number of test data; and are the predicted value at time t and its corresponding actual value respectively.

5. A high-speed train assisted driving control system, which implements the high-speed train assisted driving control method according to any one of claims 1 to 4, characterized in that: include: A collaborative system construction module for constructing an end-edge-cloud collaborative system for the high-speed train operation process; the end-edge-cloud collaborative system for the high-speed train operation process includes an end layer for collecting, storing, and transmitting mileage and grade; an edge layer for collecting, storing, and transmitting train traction braking force, train traction braking level, and train running speed, and determining the high-speed train assisted driving control strategy online based on the acquired data and a downloaded high-speed train operation process prediction model; and a cloud layer for offline construction of a high-speed train operation process model based on a large amount of historical data for download by the edge layer, and receiving and backing up working condition information and processing results from the end layer and the edge layer for later use; The working condition information includes: train running status information and line status information; the processing results include train running speed prediction results and train traction braking level prediction results; The speed and level mechanism characteristic model construction module is used to construct the speed and level mechanism characteristic model based on the dynamic analysis of the speed change mechanism characteristics and level change mechanism characteristics during the high-speed train operation process; An LSTM-based high-speed train assisted driving control strategy prediction model construction module is used to determine the LSTM-based high-speed train assisted driving control strategy prediction model based on train and line data obtained from the end-edge-cloud collaborative system of the high-speed train operation process, the speed and level mechanism characteristic model, and the LSTM neural network structure; The assisted driving control module is used to combine the real-time collected data at the edge layer with the data obtained by the terminal layer. It uses the LSTM-based high-speed train assisted driving control strategy prediction model established offline based on a large amount of historical data downloaded from the cloud layer to determine the high-speed train assisted driving control strategy, and sends it to the on-board edge computing node through the 5G network for assisted driving control.

6. A high-speed train auxiliary driving control system, characterized in that: include: At least one processor, at least one memory, and computer program instructions stored in the memory, when the computer program instructions are executed by the processor, implement the high-speed train assisted driving control method as described in any one of claims 1-4.

Images