Train auxiliary driving method and device
By identifying train operation scenarios and dynamically switching parameters through a gated cyclic unit model, the problem of insufficient adaptability of the ATO system to dynamic environments is solved, enabling more precise assisted driving and improving the safety and efficiency of train operation.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-03
- Publication Date
- 2026-03-17
AI Technical Summary
The existing ATO system lacks the ability to adapt to dynamic changes in the actual operating environment, resulting in a mismatch between control commands and on-site needs. Drivers need to intervene manually, but there are problems such as delayed response or over-control, which affects train operation safety, especially during peak passenger flow or on complex lines.
A train assisted driving method based on a gated cyclic unit model is adopted. By acquiring the current train operation data, the operating scenario is identified, and the corresponding model parameter group is dynamically switched to generate real-time assisted driving output results, thereby reducing human intervention.
It improves the safety of train operation and its ability to adapt to complex environments, the output results are more in line with the actual situation, the response lag and over-control are reduced, and the safety, comfort and energy consumption performance are improved.
Smart Images
Figure CN121671685A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of train assisted driving technology, and in particular to a train assisted driving method and device. Background Technology
[0002] Automatic Train Operation (ATO) systems are widely used in rail transit, capable of automatically controlling train traction, cruise, coasting, and braking, significantly reducing the driver's workload. However, existing ATO systems generally employ rule-based control strategies, relying on pre-set standard operating curves and fixed control parameters, lacking the ability to adapt to dynamic changes in the actual operating environment and making it difficult to flexibly adjust according to real-time operating conditions.
[0003] Therefore, in actual operation, the control commands generated by the ATO may not be completely matched with the on-site requirements. At this time, the driver needs to intervene manually. However, when the driver intervenes, he relies on experience to adjust the operation, which may result in problems such as delayed response or over-control. This is especially obvious during peak passenger flow, temporary speed limits, or complex slope sections, affecting the safety of train operation, especially during peak passenger flow or on complex lines. Summary of the Invention
[0004] In view of the above problems, the present invention provides a train assisted driving method and device, the main purpose of which is to improve the operational safety of trains.
[0005] To solve the above-mentioned technical problems, the present invention proposes the following solution: In a first aspect, the present invention provides a train assisted driving method, the method comprising: Obtain current train operation data; The current operating scenario is determined based on the current train operation data. Based on the current operating scenario, the corresponding target parameter set is selected from multiple sets of gated loop unit model parameter sets for different operating scenarios; Switch the existing parameter set of the gated loop unit model to the target parameter set; The current train operation data is processed using a gated loop unit model with the target parameter set to obtain the train's assisted driving output results; The assisted driving output results are output to the display device to assist train driving.
[0006] In a second aspect, the present invention provides a train auxiliary driving device, the device comprising: The data acquisition unit is used to acquire current train operation data; A scenario determination unit is used to determine the current operating scenario based on the current train operation data obtained by the data acquisition unit; The parameter filtering unit is used to filter out the corresponding target parameter group from multiple sets of gated loop unit model parameter groups for different operating scenarios based on the current operating scenario determined by the scenario determination unit. The parameter switching unit is used to switch the existing parameter group of the gated loop unit model to the target parameter group filtered by the parameter filtering unit; The model processing unit is used to process the current train operation data using a gated loop unit model with a target parameter group that is switched by a parameter switching unit, and to obtain the train's assisted driving output results. The result output unit is used to output the assisted driving output results obtained by the model processing unit to the display device to assist train driving.
[0007] To achieve the above objectives, according to a third aspect of the present invention, a storage medium is provided, the storage medium including a stored program, wherein, when the program is executed, the device on which the storage medium is located controls the execution of the train assisted driving method of the first aspect.
[0008] To achieve the above objectives, according to a fourth aspect of the present invention, a processor is provided for running a program, wherein the program executes the train assisted driving method of the first aspect described above.
[0009] By employing the above technical solution, this invention provides a train assisted driving method and device. Based on a gated cyclic unit model, this invention pre-configures corresponding model parameter combinations for different train operation scenarios. When the system acquires current train operation data, it first identifies the current operation scenario based on this data and then selects a target parameter group matching the operation scenario from a set of preset parameter combinations. Subsequently, the system switches the existing parameter group of the gated cyclic unit model to this target parameter group and uses the model configured with the new parameter group to perform real-time inference on the current train operation data, generating assisted driving output results. Finally, this result can be output to a display device to provide the driver with operational suggestions, realizing driving assistance functions. Compared with the prior art where driver intervention results in response lag or over-control, this invention, through dynamic inference based on real-time operation data, makes the output results more consistent with the current actual situation, eliminating the need for manual intervention, suppressing response lag and over-control, and significantly improving the safety of train operation. More importantly, this invention introduces an adaptive operation scenario mechanism, which can dynamically load targeted optimized model parameters according to the characteristics of different operation stages, thereby making the control strategy more closely aligned with actual operating conditions. This design significantly improves the adaptability of the driver assistance system to complex operating environments, effectively enhances the accuracy and rationality of the output results, and further improves the safety, comfort, and energy consumption of train operation.
[0010] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and in order to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description
[0011] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 A flowchart of a train assisted driving method provided by an embodiment of the present invention is shown; Figure 2 This invention provides a flowchart of another train assisted driving method according to an embodiment of the invention. Figure 3 This diagram illustrates a block diagram of a train auxiliary driving device provided in an embodiment of the present invention. Figure 4 A block diagram of another train auxiliary driving device provided by an embodiment of the present invention is shown. Detailed Implementation
[0012] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.
[0013] To address the problems of existing technologies, this invention proposes a novel train assisted driving method through continuous creative effort. This method pre-trains a Gated Recurrent Unit (GRU) model based on historical operational data and generates differentiated model parameter sets for different operating scenarios. After training, the parameter sets are deployed to the train's onboard equipment boards. During operation, the onboard system collects train status data in real time, identifies the current operating scenario, and dynamically loads matching model parameters. Subsequently, the model performs real-time inference calculations based on the current train operation data, generates traction and braking assistance suggestions, and outputs the results to the driver's cab information display device, providing the driver with real-time, adaptive driving prompts without driver intervention and avoiding response lag.
[0014] The training process is executed by a ground training system, and the specific steps are as follows: First, historical operating data covering all train operation scenarios is acquired, including scenarios such as interval operation, station entry braking, and station exit traction. Using this full-scenario historical operating data, the gated cyclic unit (GRU) model is pre-trained to obtain the basic gated cyclic unit model and its corresponding full-scenario parameter combination, which is then saved as a full-scenario parameter file.
[0015] Subsequently, based on subsets of each sub-scenario from the full-scene data, the base model was trained with targeted adjustments to obtain specific parameter combinations applicable to each operating scenario. During training, only the model's weight parameters were updated, keeping the model structure unchanged. After training, the parameter combinations corresponding to each sub-scenario were packaged into independent sub-scenario parameter files, which were then incorporated into the parameter management system along with the full-scene parameter file.
[0016] The training process described above is completed offline on the ground system. After verification, the generated parameter files are distributed to the train's onboard equipment boards via the vehicle-to-ground communication system or local import and stored in non-volatile storage units for use during train operation, thereby enabling the pre-deployment of model parameters and adaptive switching of scenarios.
[0017] This training strategy, which pre-trains a basic model based on full-scene data and then fine-tunes it using data from each sub-scene, enables collaborative optimization of global feature extraction and local scene specialization in the model parameter space.
[0018] Specifically, full-scenario pre-training enables the model to learn the common laws of train operation (such as the basic evolution trend of speed-time and traction and braking response characteristics), and builds a robust initial parameter space. Subsequent scenario-specific adjustment training, while keeping the model structure unchanged, finely tunes the weight parameters for different operating conditions such as interval operation, station entry braking, and station exit traction, so that the model has stronger adaptability and prediction accuracy in each specific scenario.
[0019] This strategy effectively avoids the generalization dilemma of a single model "failing to address one aspect while addressing another" in multiple scenarios, and achieves efficient modeling of shared basic features and independent optimization behaviors, significantly improving the model's scene recognition matching degree and the rationality of control suggestions in complex operating environments.
[0020] Next, combined Figure 1 The present invention provides a train assisted driving method, the specific execution steps of which are as follows: Figure 1 As shown, steps 101-106 are included: 101. Obtain current train operation data.
[0021] 102. Determine the current operating scenario based on the current train operation data.
[0022] 103. Based on the current operating scenario, select the corresponding target parameter group from multiple sets of gated loop unit model parameter groups for different operating scenarios.
[0023] In this invention, the current train operation data is first acquired, including real-time train operation status information and external dispatching instruction information. The real-time train status information includes: current speed, acceleration, traction force level, and braking force level; the external dispatching information includes: operation plan data from the centralized dispatching system (CTC) or the automatic train monitoring system (ATS), such as planned arrival and departure times, station dwell times, speed-limited sections of the line, route numbers, and passenger flow levels.
[0024] The current train operation data can be in two forms: (1) Instantaneous data at a single moment, i.e., real-time sampled values within the current control cycle; (2) Short time series data, i.e., multiple frames of data continuously collected within the most recent preset time window (e.g., 1 to 5 seconds).
[0025] When using single-moment data, the system directly determines the current operating scenario based on all parameters at that moment. When using short-time series data, it can more accurately capture temporal features such as velocity change trends and acceleration dynamics, which helps improve the robustness of scene recognition. To ensure real-time performance, the data acquisition period should be as short as possible, typically set to 100ms~500ms.
[0026] It is important to note that if the data collection time window is too long, or happens to cover the transition phase between scenarios (such as when acceleration after exiting a station has just ended and cruise begins in a section), the input data may simultaneously contain features of multiple operating conditions, leading to fuzzy or unclear scene recognition results. To avoid this problem, the time window should be set reasonably, prioritizing shorter time windows (such as 1-3 seconds) to ensure that the analyzed data segment has clear dominant operating condition features, thereby improving the accuracy and reliability of scene determination.
[0027] In rare cases, if the system still determines that it is in a transitional phase and cannot be clearly assigned to a specific operational scenario, it will default to the full operational scenario. This scenario does not distinguish between specific operational scenarios and can serve as a safety fallback strategy during transitional periods to ensure the stability of assisted driving outputs.
[0028] Subsequently, in step 103, based on the above results, the system selects the target parameter group corresponding to the current running scenario (or transition state) from the pre-deployed multiple sets of gated loop unit model parameter groups to complete parameter matching.
[0029] 104. Switch the existing parameter set of the gated cyclic unit model to the target parameter set.
[0030] 105. Use a gated loop unit model with target parameter set to process the current train operation data and obtain the train's auxiliary driving output results.
[0031] 106. Output the auxiliary driving results to the display device to assist train driving.
[0032] After determining the target parameter set, the on-board equipment replaces the GRU model parameters (i.e., the "existing parameter set") in the current running memory with the target parameter set through its built-in model parameter management interface. Specifically, it can read the sub-scenario parameter file (such as the approach braking parameter file) corresponding to the current running scenario from the local non-volatile storage unit, or read the full-scenario parameter file when it is determined to be in a transition state, and load the parameter file into the weight register of the GRU model.
[0033] The parameter switching process is completed within a single control cycle, typically taking less than 50 milliseconds. After loading, an integrity check is performed to ensure that the parameter data has not been corrupted or transmitted abnormally. If the check passes, the model status is marked as parameter-ready, and it enters the inference mode, ready to execute the next calculation.
[0034] Next, the current train operation data obtained in step 101 is input into the GRU model that has completed parameter switching, and forward inference calculation is initiated in step 105. At this time, if the input is single-moment data, the model will perform single-step forward propagation based on the current state and the hidden state retained in the previous inference cycle; if the input is short-time series data, the model will perform cyclic calculations sequentially at multiple time steps, and continuously update and propagate the hidden state to maintain the consistency of the temporal dynamics. After the inference is completed, the auxiliary driving output results for the next control cycle are generated. The output content may include recommended traction / braking levels, target speed, energy-saving operation prompts (such as coasting or regenerative braking timing), etc.
[0035] Subsequently, the onboard equipment sends the assisted driving output results to the human-machine interface or train information display device in the driver's cab via the onboard communication bus (such as MVB, CAN, or Ethernet). The output information is presented in real time in digital or graphical form, such as dynamically displaying prompts like "Apply Level 3 traction", "500 meters from the station, prepare to brake", and "Recommended speed: 62 km / h".
[0036] Based on the above Figure 1 As can be seen from the implementation, the train assisted driving method provided by this invention is based on a gated cyclic unit model, which pre-configures corresponding model parameter combinations for different train operation scenarios. When the system acquires the current train operation data, it first identifies the current operation scenario based on this data and then selects a target parameter group matching the operation scenario from a set of preset parameter combinations. Subsequently, the system switches the existing parameter group of the gated cyclic unit model to the target parameter group and uses the model configured with the new parameter group to perform real-time inference on the current train operation data, generating assisted driving output results. Finally, the results can be output to a display device to provide the driver with operation suggestions, realizing driving assistance functions. Compared with the existing technology where driver intervention results in response lag or over-control, this invention, through dynamic inference based on real-time operation data, makes the output results more consistent with the current actual situation, eliminating the need for manual intervention, suppressing response lag and over-control, and significantly improving the safety of train operation. More importantly, this invention introduces an operation scenario adaptive mechanism, which can dynamically load targeted optimized model parameters according to the characteristics of different operation stages, thereby making the control strategy more consistent with actual working conditions. This design significantly improves the adaptability of the driver assistance system to complex operating environments, effectively enhances the accuracy and rationality of the output results, and further improves the safety, comfort, and energy consumption of train operation.
[0037] Furthermore, as a response to Figure 1 Further refinement and extension of the illustrated embodiment, this invention also provides another train assisted driving method, such as... Figure 2 As shown, the specific steps are as follows: 201. Obtain current train operation data.
[0038] The implementation method of step 201 is the same as that of step 101, and can achieve the same technical effect and solve the same technical problem, so it will not be repeated here.
[0039] 202. Determine the current operating scenario based on the current train operation data.
[0040] 203. Based on the current operating scenario, select the corresponding target parameter group from multiple sets of gated loop unit model parameter groups for different operating scenarios.
[0041] 204. Switch the existing parameter set of the gated cyclic unit model to the target parameter set.
[0042] In step 202 of this embodiment, in order to avoid being in a transitional situation, the current train operation data is preferentially selected as single-moment data. Specifically, during the execution process, the speed value, acceleration value, braking force level, traction force level, planned arrival and departure times of each station, and speed limit sections in the current train operation data can be obtained.
[0043] If the acceleration value is negative and the braking force level is greater than zero, and the speed limit value of the speed-limited section is greater than the speed value, and the current time is close to the arrival time in the planned arrival and departure time, the current operating scenario is determined to be the station entry braking scenario. If the acceleration is positive, the traction force level is greater than zero, and the speed value is lower than the preset threshold, and the current time is close to the departure time in the planned arrival and departure time, the current operating scenario is determined to be the departure traction scenario; if the conditions for the arrival braking scenario and the departure traction scenario are not met, the current operating scenario is determined to be the section operating scenario.
[0044] In addition, to improve the accuracy and robustness of operational scenario determination, the distance between the train and the next station can be introduced as an auxiliary spatial criterion, based on speed, acceleration, traction / braking force level and planned arrival and departure time. Furthermore, the route status and departure permission instructions issued by the signal system can be integrated as external coordination basis to enhance the reliability of the determination.
[0045] Furthermore, it's important to note that if the scenario identification module fails to identify the current operating scenario within a preset response time threshold (e.g., 200ms), it is considered a scenario identification timeout. In this case, the system immediately loads the full-scenario parameter file and calls the basic gated cyclic unit model based on that file to process the current train operation data and generate assisted driving output results. This mechanism serves as a first fallback strategy, ensuring that even with scenario identification delays or anomalies, the model can still perform inference based on general parameters, preventing interruptions to the assisted driving function.
[0046] Subsequently, in step 203, the onboard equipment receives and stores the pre-trained full-scene parameter file and multiple sub-scene parameter files—each corresponding to the interval operation, station entry braking, and station exit traction scenarios, respectively. Each parameter file contains the weight parameter combination of the Gated Cyclic Unit (GRU) model in the corresponding scenario. A mapping table is then established between the sub-scene parameter files and their corresponding operating scenarios, between the full-scene parameter files and the full scenario, and between unknown or fault scenarios, forming GRU model parameter sets for different operating scenarios. This mapping table will quickly locate the corresponding parameter file based on the operating scenario identification results during operation, ensuring the accuracy of parameter retrieval.
[0047] Based on this mechanism, after the operation scenario identification is completed, the target parameter group can be selected from multiple sets of scenario-based GRU parameter groups according to the mapping relationship table; then in step 204, the current parameter group of the model is switched to the target parameter group to achieve adaptive optimization of the working condition.
[0048] However, another abnormal situation needs to be considered during parameter switching: if the loading and verification from the existing parameter group to the target parameter group cannot be completed within the preset parameter switching time (e.g., 50ms), the parameter switching is considered a failure. In this case, the system immediately terminates the switching operation, maintains the original parameter group, and uses the current GRU model with the existing parameter group to process the input data and generate assisted driving suggestions. Simultaneously, a "parameter loading timeout" notification is sent to the driver's cab display device to alert the system to the abnormal status. This is the second fallback strategy, used to ensure the continuous availability of the model and prevent inference blocking due to parameter switching lag.
[0049] These two fallback mechanisms address perception anomalies and execution anomalies respectively, forming a complete fault-tolerant closed loop. Together, they ensure the high availability and operational continuity of the driver assistance system in dynamic operating environments, which is a key design feature that enables the engineering robustness of this invention in actual vehicle deployment.
[0050] Despite the aforementioned failures, the fundamental reason why the on-board equipment can still execute the fallback strategy is that, for the first type of failure, since the full-scenario parameter file is based on general model parameters obtained from mixed training of all operating conditions, it has been pre-set in the on-board storage and verified. Even if the specific scenario cannot be determined at present, the equipment can still call the basic model to perform inference and output conservative but safe auxiliary suggestions. For the second type of failure, since the parameter switching adopts the "in-place retention + asynchronous loading" mechanism, the original parameter group will not be overwritten before the new parameter is loaded. Therefore, when the switching fails, the device does not need to perform a recovery operation and can directly continue to use the existing parameter group to maintain the model operation, ensuring that the auxiliary functions are not interrupted.
[0051] 205. Use a gated loop unit model with target parameter set to process the current train operation data and obtain the train's auxiliary driving output results.
[0052] In this step, the train operation data can be encoded into an input vector, and this input vector can be concatenated with the hidden vector generated in the previous moment in the gated cyclic unit model to obtain a concatenated vector, where the hidden vector represents the historical operation state of the train.
[0053] Considering that the embodiments of the present invention preferably use single-moment data and combine it with the scene switching process, the hidden vector at this time can be initialized to a zero vector, or reset to a zero vector through a preset strategy, so as to avoid the hidden state of the previous scene from not matching the newly loaded target parameter group, resulting in inference deviation.
[0054] Subsequently, based on the reset gate in the gated recurrent unit model, the splicing vector is processed using the reset weight parameter and reset bias term in the target parameter group to obtain the reset gate weight vector. This vector is used to scale the hidden vector to control its participation in the generation of candidate hidden vectors.
[0055] Furthermore, based on the update gate in the gated recurrent unit model, the splicing vector is processed using the update weight parameters and update bias terms in the target parameter group to obtain the update gate weight vector. This vector is used to control the fusion ratio of the hidden vector and the candidate hidden vector when generating a new hidden vector.
[0056] Finally, based on the scaled hidden vector and the input vector, candidate hidden vectors are generated through a candidate state network, and the hidden coefficients of the hidden vectors are obtained according to the updated gate weight vector. A new hidden vector is obtained by calculating the hidden coefficients and the weighted sum of the hidden vector, the updated gate weight vector, and the candidate hidden vectors. By inputting the new hidden vector into the fully connected layer of the gated recurrent unit model, the assisted driving output of the train can be obtained.
[0057] 206. Verify the output results of the assisted driving system.
[0058] 207. After the auxiliary driving output results pass verification, they are sent to the display device to assist train driving.
[0059] In steps 206-207, it can be determined whether the assisted driving output result is within the preset safety threshold range for the corresponding operating scenario. If the assisted driving output result exceeds the safety threshold range, a safety intervention mechanism is triggered. The safety intervention mechanism includes replacing the assisted driving output result with a predefined safety default value and generating an alarm message to be sent to the display device. If it does not exceed the safety threshold range, it is directly sent to the display device to assist train driving.
[0060] The preset safety threshold range is not a fixed global limit, but is determined through the following steps: Obtain the basic safety threshold range, which includes braking level threshold, traction level threshold, and speed deviation threshold determined based on industry safety standards or train operation specifications, to ensure the safety and compliance of assisted driving output under general operating conditions.
[0061] Based on the specified historical safe operation data stored locally on the vehicle equipment, the basic safety threshold range is calibrated in a scenario-based manner to obtain the safety threshold range under each operation scenario. The specified historical safe operation data is a normal operation record within a preset time period without any safety accidents, emergency braking triggers, or overspeed alarms.
[0062] The specific calibration process is as follows: For the approach braking scenario: Based on the distribution of braking level usage in key sections during historical approach processes (such as from 500 meters to a complete stop), the average value, peak value, and operational consistency index are extracted. If the statistical results show a general trend of late braking or high braking force demand, the braking level threshold is appropriately lowered to enhance the advance and controllability of the model's recommendations, thus obtaining the braking level threshold for this scenario. At the same time, combined with the historical approach speed control accuracy, the speed deviation threshold is tightened to improve speed control safety.
[0063] For outbound traction scenarios: Based on the historical trend of traction force level changes and relevant indicators of start-up smoothness during outbound processes (such as acceleration rise rate, wheel speed difference, and whether anti-slip control is triggered), the risk tendency of traction operation is analyzed; if the data shows frequent high traction accompanied by impulsiveness or slippage, the traction force level threshold is appropriately lowered to suppress aggressive suggestions, thus obtaining the traction force level threshold for this scenario; at the same time, based on the relationship between acceleration efficiency and line speed limit, the speed deviation threshold is dynamically adjusted to balance operating efficiency and safety margin.
[0064] For interval operation scenarios: the basic braking level threshold and traction level threshold remain unchanged, preserving the model's output flexibility in long-distance operations such as cruise and coasting; the speed deviation threshold is adaptively adjusted only based on the speed stability performance of historical interval operations, ensuring that the recommended speed meets the energy-saving optimization requirements without deviating from the safety boundary.
[0065] Therefore, the safety threshold ranges for each operating scenario are as follows: Approach braking scenario: includes calibrated braking level threshold and speed deviation threshold; Departure traction scenario: includes calibrated traction force level threshold and speed deviation threshold; Sectional operation scenario: Includes calibrated speed deviation thresholds, while retaining the original traction and braking level thresholds as default constraints.
[0066] Furthermore, as a response to the above Figure 1 In addition to the implementation of the method shown, this embodiment of the invention also provides a train auxiliary driving device for the above-described method. Figure 1 The method shown is implemented accordingly. This device embodiment corresponds to the foregoing method embodiment. For ease of reading, this device embodiment will not repeat the details of the foregoing method embodiment, but it should be clear that the device in this embodiment can implement all the contents of the foregoing method embodiment. Figure 3 As shown, the device includes: Data acquisition unit 301 is used to acquire current train operation data; The scenario determination unit 302 is used to determine the current operating scenario based on the current train operation data obtained by the data acquisition unit 301; The parameter filtering unit 303 is used to filter out the corresponding target parameter group from multiple groups of gated loop unit model parameter groups for different operating scenarios based on the current operating scenario determined by the scenario determination unit 302. The parameter switching unit 304 is used to switch the existing parameter group of the gated loop unit model to the target parameter group filtered by the parameter filtering unit 303. The model processing unit 305 is used to process the current train operation data using the gated loop unit model of the target parameter group switched by the parameter switching unit 304, and obtain the train's assisted driving output result. The result output unit 306 is used to output the assisted driving output result obtained by the model processing unit 305 to the display device to assist train driving.
[0067] Furthermore, as a response to the above Figure 2 In addition to the implementation of the method shown, this embodiment of the invention also provides another train auxiliary driving device for the above-described method. Figure 2 The method shown is implemented accordingly. This device embodiment corresponds to the foregoing method embodiment. For ease of reading, this device embodiment will not repeat the details of the foregoing method embodiment, but it should be clear that the device in this embodiment can implement all the contents of the foregoing method embodiment. Figure 4 As shown, the device includes: Data acquisition unit 301 is used to acquire current train operation data; The scenario determination unit 302 is used to determine the current operating scenario based on the current train operation data obtained by the data acquisition unit 301; The parameter filtering unit 303 is used to filter out the corresponding target parameter group from multiple groups of gated loop unit model parameter groups for different operating scenarios based on the current operating scenario determined by the scenario determination unit 302. The parameter switching unit 304 is used to switch the existing parameter group of the gated loop unit model to the target parameter group filtered by the parameter filtering unit 303. The model processing unit 305 is used to process the current train operation data using the gated loop unit model of the target parameter group switched by the parameter switching unit 304, and obtain the train's assisted driving output result. The result output unit 306 is used to output the assisted driving output result obtained by the model processing unit 305 to the display device to assist train driving.
[0068] In one optional implementation, the scene determination unit 302 is specifically used for: Obtain the speed value, acceleration value, braking force level, traction force level, planned arrival and departure times of each station, and speed limit sections from the current train operation data; If the acceleration value is negative, the braking force level is greater than zero, and the speed limit value of the speed limit section is greater than the speed value, and the current time is close to the arrival time in the planned departure time, the current operating scenario is determined to be an entry braking scenario. If the acceleration is positive, the traction force level is greater than zero, and the speed value is lower than a preset threshold, and the current time is close to the departure time in the planned arrival and departure time, the current operating scenario is determined to be a departure traction scenario. If the conditions for the entry braking scenario and the exit traction scenario are not met, the current operating scenario is determined to be a section operating scenario.
[0069] In one optional implementation, the model processing unit 305 is specifically used for: The train operation data is encoded into an input vector, and the input vector is concatenated with the hidden vector generated in the previous moment in the gated cyclic unit model to obtain a concatenated vector. The hidden vector represents the historical operation state of the train. Based on the reset gate in the gated loop unit model, the splicing vector is processed using the reset weight parameter and reset bias term in the target parameter group to obtain the reset gate weight vector. The reset gate weight vector is used to scale the hidden vector to control the degree of participation of the hidden vector in generating candidate hidden vectors. Based on the update gate in the gated loop unit model, the spliced vector is processed using the update weight parameter and update bias term in the target parameter group to obtain the update gate weight vector. The update gate weight vector is used to control the fusion ratio of the hidden vector and the candidate hidden vector when generating a new hidden vector. Based on the scaled hidden vector and the input vector, candidate hidden vectors are generated through a candidate state network, and the hidden coefficients of the hidden vectors are obtained according to the update gate weight vector. A new hidden vector is obtained by calculating the weighted sum of the hidden coefficients, the hidden vector, the updated gate weight vector, and the candidate hidden vector; The new hidden vector is input into the fully connected layer of the gated cyclic unit model to obtain the train's assisted driving output.
[0070] In an optional implementation, when the scene determination unit 302 determines the current operating scene based on the current train operation data, the device further includes a backup unit 307, which is specifically used for: If it is determined that the response time of the current running scenario exceeds the preset time threshold, then load the full scenario parameter file; The current train operation data is processed using a basic gated loop unit model with the full scenario parameter file to obtain the train's assisted driving output results.
[0071] In an optional implementation, when the parameter switching unit 304 switches the existing parameter group of the gated loop unit model to the target parameter group, it is further configured to: If the switch from the existing parameter group to the target parameter group is not completed within the preset parameter switching time, the switch to the target parameter group will be terminated immediately. After the switch is terminated, the current train operation data is processed using the door loop unit model with existing parameter groups to obtain the train's assisted driving output results, and a parameter loading timeout prompt is sent to the display device.
[0072] In an optional embodiment, after the result output unit 306 obtains the assisted driving output result of the train, the device further includes a result verification unit 308, which is specifically used for: Determine whether the assisted driving output is within the preset safety threshold range for the corresponding operating scenario; If the assisted driving output exceeds the safety threshold range, a safety intervention mechanism is triggered. The safety intervention mechanism includes replacing the assisted driving output with a predefined safety default value and generating an alarm message to be sent to the display device.
[0073] In one optional implementation, the preset security threshold range is determined through the following steps: Obtain a basic safety threshold range, which includes braking level threshold, traction level threshold, and speed deviation threshold determined based on industry safety standards; Based on the specified historical safe operation data stored locally on the on-board equipment, the basic safety threshold range is calibrated in a scenario-based manner to obtain the safety threshold range under each operation scenario. The specified historical safe operation data consists of train operation records without safety accidents.
[0074] Furthermore, embodiments of the present invention also provide a storage medium for storing a computer program, wherein the computer program, when running, controls the device where the storage medium is located to execute the above-described... Figure 1-2 The train assisted driving method described herein.
[0075] Furthermore, embodiments of the present invention also provide a processor for running a program, wherein the program executes the above-described... Figure 1-2 The train assisted driving method described herein.
[0076] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.
[0077] It is understood that the relevant features in the above methods and apparatus can be referenced interchangeably. Furthermore, the terms "first," "second," etc., in the above embodiments are used to distinguish between embodiments and do not represent the superiority or inferiority of any particular embodiment.
[0078] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.
[0079] The algorithms and displays provided herein are not inherently related to any particular computer, virtual system, or other device. Various general-purpose systems can also be used in conjunction with the teachings herein. The required structure for constructing such systems is apparent from the above description. Furthermore, this invention is not directed to any particular programming language. It should be understood that the contents of the invention described herein can be implemented using various programming languages, and the above description of specific languages is for the purpose of disclosing the best mode of implementation of the invention.
[0080] In addition, the memory may include non-permanent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.
[0081] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0082] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0083] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0084] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0085] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.
[0086] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.
[0087] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.
[0088] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.
[0089] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0090] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.
Claims
1. A train auxiliary driving method characterized by comprising: The method comprises: acquiring current train operation data; determining a current operation scenario according to the current train operation data; screening a corresponding target parameter group from a plurality of groups of gated recurrent unit model parameters for different operation scenarios according to the current operation scenario; switching an existing parameter group of the gated recurrent unit model to the target parameter group; processing the current train operation data by using the gated recurrent unit model with the target parameter group to obtain an auxiliary driving output result of the train; outputting the auxiliary driving output result to a display device to assist train driving.
2. The method of claim 1, wherein, Determining a current operation scenario according to the current train operation data comprises: acquiring a speed value, an acceleration value, a braking force level, a traction force level, planned arrival and departure time of each station, and a speed limit section in the current train operation data; if the acceleration value is negative, the braking force level is greater than zero, the speed limit value of the speed limit section is greater than the speed value, and the current time is close to the arrival time in the planned arrival and departure time, determining that the current operation scenario is an arrival braking scenario; if the acceleration value is positive, the traction force level is greater than zero, and the speed value is lower than a preset threshold, and the current time is close to the departure time in the planned arrival and departure time, determining that the current operation scenario is a departure traction scenario; if the conditions of the arrival braking scenario and the departure traction scenario are not met, determining that the current operation scenario is an interval operation scenario.
3. The method of claim 1, wherein, Processing the current train operation data by using the gated recurrent unit model with the target parameter group to obtain an auxiliary driving output result of the train comprises: encoding the train operation data into an input vector, splicing the input vector and a hidden vector generated at a previous time in the gated recurrent unit model to obtain a spliced vector, and the hidden vector represents a historical operation state of the train; processing the spliced vector by using a reset weight parameter and a reset bias term in the target parameter group according to a reset gate in the gated recurrent unit model to obtain a reset gate weight vector, and the reset gate weight vector is used to scale the hidden vector to control the participation degree of the hidden vector in generating a candidate hidden vector; processing the spliced vector by using an update weight parameter and an update bias term in the target parameter group according to an update gate in the gated recurrent unit model to obtain an update gate weight vector, and the update gate weight vector is used to control the fusion proportion of the hidden vector and the candidate hidden vector when generating a new hidden vector; generating a candidate hidden vector by a candidate state network based on the scaled hidden vector and the input vector, and obtaining a hidden coefficient of the hidden vector according to the update gate weight vector; obtaining a new hidden vector according to the hidden coefficient, the hidden vector, and a weighted sum value between the update gate weight vector and the candidate hidden vector; inputting the new hidden vector into a fully connected layer of the gated recurrent unit model to obtain an auxiliary driving output result of the train.
4. The method of claim 1, wherein, When determining a current operation scenario according to the current train operation data, the method further comprises: If it is determined that the response duration of the current running scenario exceeds the preset duration threshold, a full-scene parameter file is loaded; The current train operation data is processed by using a basic gated recurrent unit model with the full-scene parameter file to obtain an auxiliary driving output result of the train.
5. The method of claim 1, wherein, When the existing parameter group of the gated recurrent unit model is switched to the target parameter group, the method further comprises: If the existing parameter group is not switched to the target parameter group within a preset parameter switching duration, switching to the target parameter group is immediately terminated; After the switching is terminated, the current train operation data is processed by using the gated recurrent unit model with the existing parameter group to obtain an auxiliary driving output result of the train, and a parameter loading timeout prompt is generated to a display device.
6. The method of claim 1, wherein, After the auxiliary driving output result of the train is obtained, the method further comprises: It is determined whether the auxiliary driving output result is within a preset safety threshold range corresponding to the running scenario; If the auxiliary driving output result exceeds the safety threshold range, a safety intervention mechanism is triggered, which includes replacing the auxiliary driving output result with a predefined safety default value and generating an alarm information to a display device.
7. The method of claim 6, wherein, The preset safety threshold range is determined by the following steps: An basic safety threshold range is obtained, which includes a brake level threshold, a traction level threshold and a speed deviation threshold determined based on industry safety standards; The basic safety threshold range is calibrated based on specified historical safety operation data stored locally by the on-board device to obtain a safety threshold range for each running scenario, and the specified historical safety operation data is a train operation record without safety accidents.
8. A train auxiliary driving device characterized by comprising: The device comprises: a data acquisition unit configured to acquire current train operation data; a scenario determination unit configured to determine a current running scenario based on the current train operation data acquired by the data acquisition unit; a parameter screening unit configured to screen a corresponding target parameter group from a plurality of gated recurrent unit model parameter groups for different running scenarios based on the current running scenario determined by the scenario determination unit; a parameter switching unit configured to switch an existing parameter group of the gated recurrent unit model to the target parameter group screened by the parameter screening unit; a model processing unit configured to process the current train operation data by using a gated recurrent unit model with the target parameter group switched by the parameter switching unit to obtain an auxiliary driving output result of the train; a result output unit configured to output the auxiliary driving output result of the train obtained by the model processing unit to a display device to assist train driving.
9. A storage medium, characterized by The storage medium comprises a stored program, wherein the program controls the device where the storage medium is located to execute the train auxiliary driving method of any one of claims 1 to 7 when the program is running.
10. A processor, comprising: The processor is configured to run a program, wherein the program executes the train auxiliary driving method of any one of claims 1 to 7 when the program is running.