A simulation driving training method and system for urban rail vehicles based on scenario simulation

Abstract

The present invention discloses a method and system for urban rail vehicle simulation driving training based on scene simulation, belonging to the technical field of driving simulation. The method specifically comprises: real-time collection of track line parameters, weather data, passenger flow density, vehicle equipment status and driving data, and preprocessing thereof to construct a dynamic personalized simulation scene, simulate driving operations, identify micro-features of the driver's eyebrow area according to the driver's operation and preset simulation scene, analyze the driver's psychological stress in combination with historical data, dynamically adjust the simulation scene according to the psychological stress analysis results, score the driver's performance in this training according to preset evaluation criteria, and generate a detailed training report; the present invention can judge the driver's stress in the simulation scene in real time by performing pressure analysis on the eyebrow area and performing pressure assessment in combination with physiological dimensions, provide different simulation scenes for different drivers, and greatly improve the training effect.

Description

Technical Field

[0001] The present invention belongs to the technical field of driving simulation, and in particular relates to a method and system for training urban rail vehicle driving simulation based on scenario simulation. Background Art

[0002] Numerous rail transit vehicle driving simulation training systems already exist, but their advantages and disadvantages are becoming increasingly apparent. Advantages include significantly reducing the shortage of training resources, avoiding safety incidents that can easily occur during actual training, and, to a certain extent, alleviating the pressure on instructors. However, disadvantages include: first, the lack of grading of dynamic scenarios, which is disconnected from actual operational dynamics; second, the lack of multimodal feedback, resulting in limited feedback; third, low stress assessment accuracy, making it impossible to switch scenarios in real time; and fourth, manual assessments make it difficult to determine whether drivers have mastered all training content.

[0003] In order to quickly train qualified urban rail vehicle drivers, a large amount of driving training is necessary, but the level of training that each person can accept is different. For example, in complex driving scenarios, the pressure on novice drivers will inevitably increase, and operational errors will occur. It is necessary to switch to a lower-level scenario for training. The current technology for identifying psychological stress cannot quickly identify it in complex driving scenarios, which will affect the training effect.

[0004] For example, the Chinese patent with authorization announcement number CN112102681B discloses a standard EMU driving simulation training system and method based on an adaptive strategy. The method includes: an external environment simulation device obtains the train line type, train signal type, driver job type, train section type, train course type, train operation type, and driver evaluation level information under the standard EMU driving application scenario; the obtained information is input into the trainee model to obtain the trainee's course training evaluation value; matching is performed based on the trainee's course training evaluation value, matching the course that the trainee needs to take, and inputting it into the EMU driving simulation device; the EMU driving simulation device matches the corresponding train simulation logic based on the course that the trainee needs to take; the trainee performs train simulation operations according to the corresponding simulation logic. This invention automatically matches training courses for trainees and provides a complete standard EMU driving simulation training process.

[0005] The defect of the above technical solution is that when the driver is under great pressure due to complex simulation scenes, the simulation scenes cannot be adjusted in real time, resulting in poor training results. Summary of the Invention

[0006] In response to the shortcomings of the existing technology, the present invention proposes a scenario-based simulation-based urban rail vehicle simulation driving training method and system. By combining dynamic scenario generation, five-sense immersive feedback, stress assessment, and causal chain tracing, the simulation scenario can be adjusted in real time according to the pressure faced by the driver during training, thereby improving the trainees' training effect, adaptability to complex scenarios, emergency decision-making ability, and psychological stability.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A simulation driving training method based on scenario simulation, comprising:

[0009] Collect simulated driving data in real time, perform pre-processing, and build dynamic and personalized simulation scenarios;

[0010] Simulate driving operations, identify the driver's eyebrow area's micro-features based on the driver's operations and preset simulation scenarios, analyze the driver's psychological stress based on historical data, and dynamically adjust the simulation scenario based on the psychological stress analysis results;

[0011] Based on the preset evaluation criteria, the driver's performance in training is scored and a training report is generated.

[0012] Specifically, the construction of a dynamic personalized simulation scenario includes:

[0013] Based on the pre-processed simulated driving data, a sliding time window is used to classify driving errors and calculate their frequencies to construct a driver capability profile.

[0014] Based on the state space and action space, a state-action reward model is constructed to dynamically adjust the driver's ability profile;

[0015] A three-layer architecture is used to generate dynamic personalized hierarchical simulation scenarios, including a basic scenario layer, a disturbance enhancement layer, and an event chain weaving layer. The basic scenario layer includes a vehicle dynamics model and an environment interaction model.

[0016] Specifically, the simulated driving operation identifies micro-features of the driver's eyebrow area based on the driver's operation and the preset simulation scene, analyzes the driver's psychological stress in combination with historical data, and dynamically adjusts the simulation scene based on the psychological stress analysis results, including:

[0017] Dynamically couple the six-degree-of-freedom motion platform;

[0018] Build a force feedback model and associate it with odor events to provide real-time feedback on force and odor;

[0019] Identify the micro features of the driver's eyebrow area and analyze the driver's psychological stress by combining historical data;

[0020] Every t time period, the difficulty of the simulation scene is automatically increased to simulate passenger behavior and provide dynamic feedback. The simulation scene is dynamically adjusted by combining force, smell and pressure feedback.

[0021] Specifically, the dynamic coupling of the six-degree-of-freedom motion platform includes:

[0022] According to the vehicle dynamics model output and actual operation data, the relationship equation between the hydraulic cylinder length and posture is established and solved;

[0023] Predict vehicle posture changes in the future, generate motion commands in advance, and compensate for response delays;

[0024] Load the track irregularity spectrum to generate random vibration signals, and adjust the stiffness according to the number of passengers.

[0025] Specifically, the identifying micro features of the driver's eyebrow area and analyzing the driver's psychological stress in combination with historical data include:

[0026] During a period of T in the driver's historical training, facial images of the driver within the period of T in the driver's historical training are collected through a first face collection window at a preset frame interval, and are set as first collected data;

[0027] The eyes and eyebrows are cropped in the first collected data, and the areas outside the eyes and eyebrows are removed to obtain the second collected data;

[0028] Reconstruct the eyebrow muscle deformation based on the 3D deformable model to capture the characteristics of the eyebrow region, including: orbicularis oculi muscle contraction rate, brow peak displacement, eyeball displacement, and blink duration;

[0029] Simultaneously collect steering wheel grip and heart rate variability to build a triangular verification of eyebrow-physiology-operation, inject training time tags in real time, and calculate the temporal correlation between events and eyebrow-physiology-operation;

[0030] Based on the temporal correlation between events and eyebrow-physiology-operation, the driver's real-time stress is calculated and a stress-time-scenario heat map is generated.

[0031] Specifically, the driver's performance in training is scored according to the preset evaluation criteria, and a training report is generated, including:

[0032] Real-time collection of driver evaluation data during historical training time period T, including operational specification data, physiological data, and emergency decision-making data, and pre-processing of the collected evaluation data;

[0033] Evaluate the driver's training results based on the pre-processed evaluation data and preset evaluation criteria;

[0034] Using the analytic hierarchy process, weights are assigned to pre-set evaluation criteria and the overall score of the driver training is calculated.

[0035] Generate visual reports on driver training through 3D radar charts, pressure heat maps and causal chain images.

[0036] Specifically, the construction of a force feedback model and association with odor events to provide real-time force and odor feedback includes:

[0037] Calculate the vehicle steering resistance torque based on the wheel-rail adhesion coefficient and dynamic disturbance;

[0038] Adaptive fuzzy PID control is used to dynamically adjust the damping according to the driver's operation smoothness;

[0039] Correlate odor events and analyze the source of odors;

[0040] Airflow guidance technology is used to diffuse the generated odor.

[0041] A scenario-based urban rail vehicle simulation driving training system is used to implement the scenario-based urban rail vehicle simulation driving training method, comprising: a simulation scenario construction module, a driving training module, and a training evaluation module;

[0042] The simulation scene construction module is used to collect simulated driving data in real time, perform preprocessing, and construct dynamic personalized simulation scenes;

[0043] The driving training module is used to simulate driving operations, identify micro-features of the driver's eyebrow area based on the driver's operations and preset simulation scenarios, analyze the driver's psychological stress in combination with historical data, and dynamically adjust the simulation scenarios based on the psychological stress analysis results;

[0044] The training evaluation module scores the driver's performance during training according to preset evaluation criteria and generates a training report.

[0045] Specifically, the driving training module includes: a dynamic coupling unit, a pressure analysis unit and a scene adjustment unit;

[0046] The dynamic coupling unit is used to dynamically couple the six-degree-of-freedom motion platform;

[0047] The pressure analysis unit is used to build a force feedback model and associate it with odor events, providing real-time feedback on force and odor;

[0048] The scene adjustment unit is used to identify the micro features of the driver's eyebrow area, analyze the driver's psychological pressure in combination with historical data, and provide dynamic feedback, combining force, smell and pressure feedback to dynamically adjust the simulation scene.

[0049] Compared with the prior art, the present invention has the following beneficial effects:

[0050] The present invention proposes a scenario-based simulation-based urban rail vehicle driving training method with significant practical value and technical advantages. The method constructs dynamic personalized training scenarios by real-time collection of multi-source data such as track lines, weather, passenger flow, equipment status and driving behavior, combined with sliding window analysis and state-action-reward model, to achieve precise training tailored to each individual. By introducing multimodal physiological signals such as eyebrow micro-expression recognition, heart rate variability, and grip strength, the driver's psychological stress is accurately assessed, and multi-sensory interaction methods such as a six-degree-of-freedom motion platform, force feedback and odor stimulation are used to dynamically adjust the simulation scenario, effectively enhancing the immersion and authenticity of training. In addition, the training process is quantitatively evaluated based on the hierarchical analysis method, and a visual report containing a three-dimensional radar chart, a pressure heat map and a causal chain map is generated to help managers fully grasp the driver's ability profile and development trend, thereby improving urban rail transit safety and driver training efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 A flow chart of a scenario-based urban rail vehicle driving simulation training method provided by the present invention;

[0052] Figure 2 A schematic diagram of dynamic adjustment of the simulation scene provided by the present invention;

[0053] Figure 3 This is a schematic diagram of eyebrow recognition provided by the present invention;

[0054] Figure 4 This is an architecture diagram of a scenario-based urban rail vehicle driving simulation training system provided by the present invention. DETAILED DESCRIPTION

[0055] The present application is described in detail below with reference to specific embodiments. The following embodiments will help those skilled in the art to further understand the present application, but are not intended to limit the present application in any form. It should be noted that those skilled in the art may make several variations and improvements without departing from the scope of the present application. These all fall within the scope of protection of the present application.

[0056] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0057] It should be noted that, if there is no conflict, the various features in the embodiments of the present application can be combined with each other and are all within the scope of protection of the present application. In addition, although the functional modules are divided in the device schematic and the logical order is shown in the flow chart, in some cases, the steps shown or described can be performed in a different order than the module division in the device or the order in the flow chart. In addition, the words "first", "second", "third", etc. used in this application do not limit the data and execution order, but only distinguish between the same items or similar items with basically the same functions and effects.

[0058] Unless otherwise defined, all technical and scientific terms used in this specification have the same meanings as those commonly understood by those skilled in the art to which this application belongs. The terms used in this specification and in the specification of this application are only for the purpose of describing specific embodiments and are not intended to limit this application. The term "and / or" as used in this specification includes any and all combinations of one or more of the relevant listed items.

[0059] Example 1

[0060] See also Figure 1-Figure 3 The present invention provides an embodiment of a scenario-based urban rail vehicle driving simulation training method, comprising the following specific steps:

[0061] Step S1: Through on-board sensors, roadside equipment and cloud layers, track line parameters, weather data, passenger flow density, vehicle equipment status and driving data are collected in real time, and pre-processed to build a dynamic and personalized simulation scenario.

[0062] The on-board sensors include: a laser radar for collecting track profiles; an inertial measurement sensor (IMU) for monitoring vehicle posture; and a brake cylinder pressure sensor for capturing mechanical response delays.

[0063] Roadside equipment includes: 5G base stations, which transmit real-time signal status, including yellow / red light phases, contact network voltage fluctuations, etc.; trackside cameras, which identify foreign objects on the roadbed and are used to distinguish between fallen leaves and metal parts.

[0064] The cloud layer includes: access to the city traffic brain API to obtain real-time passenger flow heat maps, weather radar data, etc.

[0065] In step S1, a dynamic personalized simulation scenario is constructed, and the specific steps include:

[0066] Step S101: Based on the pre-processed simulated driving data, a sliding time window is used to classify driving errors and calculate their frequencies to construct a driver capability profile.

[0067] In this embodiment, the pre-processed simulated driving data includes: operational data: braking deceleration smoothness, number of speeding, and lookout time; decision-making data: fault response time and type of misoperation; physiological data: pressure value and pupil diameter increase; driving error categories include: operational data: sudden braking, speeding, misoperation, etc.; decision-making data: response delay, misjudgment of fault, etc.; psychological data: high stress fluctuation, distraction, etc.

[0068] The purpose of setting up the sliding time window is to retain the temporal context characteristics of the behavior. Each window contains driving data of a fixed length (such as 30 seconds). The sliding window can capture dynamic changes in the time series and the recurrence of behavioral patterns. For example: every 30 seconds is a window, and the sliding step is 10 seconds. The first window analyzes 0-30 seconds of data, the second window analyzes 10-40 seconds of data, and so on. After analyzing multiple sliding windows, the frequency of occurrence of different types of errors in the entire training process is counted, and the driver's ability profile is constructed through the dimensional projection of the error behavior, including operational stability, rule compliance, decision-making ability and attention maintenance.

[0069] Step S102: Based on the state space and action space, a state-action reward model is constructed to dynamically adjust the driver's ability profile.

[0070] In this embodiment, the state space is composed of a combination of multiple variables and is hierarchically divided into: environmental state variables, driver state variables, and vehicle state variables. The action space includes common vehicle operation actions, such as acceleration and deceleration. The state-action reward model is constructed based on the Markov decision process. The mapping relationship between the frequency of driving errors and the driver's ability profile is established through the Q-learning algorithm. The driver's ability profile is dynamically adjusted according to the reward value of each state and action.

[0071] Step S103: Generate a dynamic personalized hierarchical simulation scenario using a three-layer architecture, including a basic scenario layer, a disturbance enhancement layer, and an event chain weaving layer. The basic scenario layer includes a vehicle dynamics model and an environment interaction model.

[0072] In this embodiment, the basic scene layer is a dynamic skeleton, which is based on the realistic rendering of the scene of the physics engine. It includes: a vehicle dynamics model and an environmental interaction model. The vehicle dynamics model uses multi-body dynamics software to build an urban rail vehicle model; the environmental interaction model includes weather-track coupling and passenger flow-driving feedback.

[0073] The perturbation enhancement layer is for personalized pain point attack, including weak point targeted perturbation and pressure gradient design;

[0074] The event chain weaving layer is causal complexity, including: first-level events, single disturbances; second-level events, causal chain disturbances; third-level events, nonlinear chain disturbances;

[0075] The levels are graded according to the difficulty of the basic scenario layer, disturbance enhancement layer, event chain weaving layer, pressure standard and evaluation standard. For example: L1 basic level: single causal chain; L3 advanced level: 3-level causal chain; L5 expert level: 8-level causal chain. By splitting the scenario construction into three layers, the degree of intervention of the current training can be automatically determined according to the driver's ability profile.

[0076] Step S2: Simulate driving operations, identify the micro features of the driver's eyebrow area based on the driver's operations and the preset simulation scene, analyze the driver's psychological pressure in combination with historical data, and dynamically adjust the simulation scene based on the psychological pressure analysis results.

[0077] The specific steps of step S2 are:

[0078] Step S201: Dynamically couple the six-degree-of-freedom motion platform.

[0079] The six-degree-of-freedom platform includes translation along the XYZ axis (X / Y / Z) and rotation around the XYZ axis (roll / pitch / yaw).

[0080] The specific steps of step S201 are:

[0081] Step S2011: Based on the vehicle dynamics model output and actual operation data, a relationship equation between the length and posture of the hydraulic cylinder is established, and the relationship equation is solved.

[0082] In this embodiment, the vehicle dynamics model outputs: pitch angle, roll angle, vertical acceleration, real-time operation data: braking deceleration, steering angle, and the relationship equation between the hydraulic cylinder length and posture is constructed based on the Stewart platform structure and then solved using the Newton iteration method; Principle: The posture is converted into the extension and contraction amount of each hydraulic cylinder through inverse kinematics to ensure that the motion posture is consistent with the vehicle dynamics model. The benefits are: improved motion accuracy, support for complex posture combinations, such as simultaneous pitch + roll + translation, and realistic reproduction of the vehicle's serpentine motion.

[0083] Step S2012: Predict vehicle posture changes in future time periods, generate motion instructions in advance, and compensate for response delays.

[0084] In this embodiment, key parameters such as vehicle speed, acceleration, and steering wheel angle sampled by the sliding window method are used to predict the possible vehicle posture change trends in the future. Pre-adjusted motion instructions are generated based on the prediction results and sent to the six-degree-of-freedom platform. If there is a response delay between the actual vehicle posture and the predicted value, dynamic correction is performed through the feedback channel.

[0085] Step S2013: Load the track irregularity spectrum, generate a random vibration signal, and adjust the stiffness according to the number of passengers.

[0086] In this embodiment, by superimposing scene characteristic vibration and load compensation, the six-degree-of-freedom platform dynamically changes the track bumpiness of the random vibration signal input in time series, thereby enhancing the authenticity of the motion feedback and being able to truly reproduce the vehicle load changes.

[0087] Step S202: Construct a force feedback model and associate it with the odor event to provide real-time force and odor feedback.

[0088] The specific steps of step S202 are:

[0089] Step S2021: Calculate the vehicle steering resistance torque based on the wheel-rail adhesion coefficient and dynamic disturbance;

[0090] Dynamic disturbances include: slippery road surfaces and fault prompts. The adhesion coefficient is corrected in real time through multi-source data fusion. At the same time, environmental changes and fault events are converted into resistance torque disturbances, enhancing the complexity and authenticity of driving feedback.

[0091] Step S2022: Adopting adaptive fuzzy PID control to dynamically adjust the damping according to the driver's operation smoothness;

[0092] In this embodiment, the damping is dynamically adjusted based on the driver's operating smoothness, while compensating for system delays to ensure real-time and accurate force feedback. This can greatly improve the accuracy of force feedback, and unskilled operators will feel additional resistance, forcing them to develop gentle operating habits.

[0093] Step S2023: Correlate the odor events and analyze the source of the odor;

[0094] In this embodiment, odor event associations include: motor overheating, releasing a burnt smell; brake pad wear, releasing a metal dust smell; in a tunnel, a mixed smell of exhaust and wet rocks; on rainy days, triggering a rain-ozone compound smell;

[0095] Step S2024: Use airflow guidance technology to diffuse the generated odor.

[0096] In this embodiment, the odor is released in a directionally controlled manner through the micropores at the bottom of the steering wheel, and the pump flow is dynamically adjusted based on the gas diffusion model to avoid olfactory fatigue. The benefits of this are: it improves the correlation between odor and events, which is much higher than traditional solutions, improves the efficiency of pressure awakening, and the burnt smell improves the efficiency of pressure awakening and enhances the sense of urgency in fault handling.

[0097] Step S203: Identify the micro features of the driver's eyebrow area and analyze the driver's psychological stress in combination with historical data.

[0098] The specific steps of step S203 are:

[0099] Step S2031: During the driver's historical training period T, facial images of the driver within the driver's historical training period T are collected through the first face collection window at a preset frame interval, and are set as first collected data.

[0100] In this embodiment, during the driver's historical training period T, the scene difficulty gradually increases. The facial images of the driver facing scenes of different difficulty levels can be collected according to the preset frame interval. It is universal. When collecting facial images, the collection is carried out after the driver's consent, and is only used for this training evaluation and not for other purposes.

[0101] Step S2032: cropping the eyes and eyebrows in the first collected data, removing areas outside the eyes and eyebrows, and obtaining second collected data.

[0102] Through facial key point detection and region of interest (ROI) segmentation, the eye and brow areas are accurately extracted, eliminating interference from other facial areas (such as the chin and cheeks), providing accurate input data for subsequent micro-expression analysis.

[0103] Step S2033: reconstructing the muscle deformation of the eyebrow region based on the three-dimensional deformable model to capture the features of the eyebrow region, including: orbicularis oculi muscle contraction rate, brow peak displacement, eyeball displacement and blink duration.

[0104] In this embodiment, facial three-dimensional point cloud data is obtained through a structured light camera, and the muscle deformation of the eyebrow area is reconstructed in combination with a three-dimensional deformable model. The three-dimensional deformable model uses an active appearance model combined with a convolutional neural network to extract the deformation features of the eyebrow area, dynamically track the muscle surface changes in each frame, and calculate the contraction rate of the orbicularis oculi muscle, the displacement of the brow peak, the displacement of the eyeball and the duration of the blink. The advantage of reconstruction is that it can distinguish between different types of labels, such as natural frowns and stressed frowns.

[0105] Specifically, the orbicularis oculi muscle contraction rate ORC: the degree of contraction is determined by calculating the grayscale variance change rate of the canthus area; the brow peak displacement MPD: the displacement is calculated by tracking the three-dimensional coordinate changes of the brow peak point; the eyeball displacement ED: pressure causes pupil dilation and changes in gaze direction; the blink duration BTD: the blink frequency is reduced but the duration is prolonged under stress. Under normal circumstances, the blink duration is 80-100ms.

[0106] Step S2034: Simultaneously collect steering wheel grip, heart rate variability, etc., build eyebrow-physiology-operation triangle verification, inject training time tags in real time, and calculate the time correlation between events and eyebrow-physiology-operation.

[0107] In this embodiment, a triangular verification mechanism of eyebrow-physiology-operation is constructed to improve the accuracy of driver psychological state recognition. Specifically, three types of key data are collected synchronously during the training process: one is the dynamic characteristics of the muscles in the eyebrow area, which are used to reflect changes in facial micro-expressions; the second is physiological parameters; and the third is operational behavior data. All data are timestamped and correlation analysis is performed within a fixed sliding time window. The time correlation between the three types of data is calculated. If the three show a high degree of synchronization, it is inferred that there is a psychological stress event in the time period, and it is marked as a key node and injected into the training timeline for stress modeling and personalized training scenario adjustment.

[0108] Triangle verification logic: Pressure determination must meet at least two of the following conditions:

[0109] Abnormal eyebrow features: ORC>25% and MPD>1.2mm;

[0110] Abnormal physiological signals: grip force > 300N for 2 seconds or HRV < 50ms;

[0111] Abnormal operating data: braking deceleration > 1.5m / s² or steering angular velocity > 60° / s;

[0112] Finally, the time lag correlation between the event and the multimodal data changes is calculated to identify the effective pressure response window. It should be noted that the data in the judgment conditions are set by people in this field based on simulation experiments.

[0113] Step S2035: Calculate the driver's real-time stress based on the temporal correlation between the event and the eyebrow-physiology-operation, and generate a stress-time-scene heat map.

[0114] In this embodiment, the important features of the eyebrow and eye region are used instead of the overall features of the face. The advantage of this is that the feature detection area is effectively reduced. The detection speed of the eyebrow and eye region is faster than that of the face. Although the accuracy is lower than that of the face detection, in the face of emergencies, the faster the detection speed, the better the effect. For example, when the driver is in a difficult simulation scene, the pressure increases and he may faint or have stressful behavior at any time. Using the method of this application, he can react more quickly and adjust the simulation scene immediately.

[0115] like Figure 2 As shown, step S204: every t time period, automatically increase the difficulty of the simulation scene, simulate the passenger behavior, and perform dynamic feedback, combine force, smell and pressure feedback, and dynamically adjust the simulation scene.

[0116] In this embodiment, the dynamic adjustment of the difficulty of the simulation scene is automatically triggered at every preset time interval. Based on the driver's ability portrait and real-time psychological state assessment results constructed in the early stage of training, more complex disturbance factors are introduced in stages. At the same time, force feedback is achieved through the six-degree-of-freedom platform, and the odor generation module and the psychological stress model (such as blinking frequency, heart rate variability and other indicators) are combined to perceive the driver's state in real time. When it is detected that the driver has obvious stress reaction or operation deviation, the scene difficulty is dynamically adjusted back or switched to a stable state according to the feedback results, realizing real-time adaptive matching between difficulty and driver state.

[0117] refer to Figure 2 , where the facial data is the relevant data and images of each part of the entire facial area. Because the data of each part of the face is different at different times, it is presented in the form of a curve. Figure 2 The square points on the middle curve represent preset frames. The advantage of this is that data changes can be obtained intuitively and the number of frames can be preset for collection. Assuming that the current scene is level 3, the set pressure threshold is b. If the detected pressure is greater than b, it exceeds the setting of the level 3 scene. The scene is immediately switched to level 2 for training. After a period of t, it is switched to level 3 for training. Following the step-by-step rule, the dynamic adjustment simulation scene here is Figure 2 Adjust the simulation scene to reduce the difficulty by one level.

[0118] Step S3: Score the driver's performance in this training according to the preset evaluation criteria and generate a detailed training report.

[0119] The specific steps of step S3 are:

[0120] Step S301: collecting the driver's evaluation data during the historical training period T in real time, including operation specification data, physiological data, and emergency decision-making data, and pre-processing the collected evaluation data.

[0121] The preprocessing includes: data cleaning, removing redundant data, predicting the smoothness of adjacent periods, and filling in missing data.

[0122] Step S302: Evaluate the driver's training results based on the pre-processed evaluation data and preset evaluation criteria.

[0123] The preset evaluation criteria include: operational normative indicators, including: braking smoothness, calculating the standard deviation of deceleration; lookout time, calculating the total time the line of sight deviates from the straight line through eye tracking;

[0124] Emergency decision-making indicators include: response time, the time from event triggering to braking intervention; misjudgment rate, the proportion of incorrect fault judgments;

[0125] Psychological stability indicators include: HRV, which calculates the root mean square difference; pupil dilation rate, which analyzes changes in pupil diameter; and skin galvanic response (SCR).

[0126] Step S303: assign weights to the preset evaluation criteria through the hierarchical analysis method, and comprehensively calculate the total score of the driver training.

[0127] Step S304: Generate a visual report of the driver training through the three-dimensional radar map, pressure heat map and cause-effect chain map.

[0128] In this embodiment, a three-dimensional radar chart displays the relative strengths of operational norms, emergency decision-making, and psychological stability. By using three-dimensional spatial geometry, abstract abilities are transformed into intuitive spatial positional relationships, facilitating the rapid identification of weaknesses in abilities.

[0129] The pressure thermodynamic diagram marks the high-pressure sections on the training timeline, uses color visual channels to encode pressure intensity, and combines the timeline to display the pressure fluctuation pattern.

[0130] The causal chain graph identifies strong association rules through association rule mining and time series analysis, and visualizes abstract causal relationships through graph theory models.

[0131] Overall effect description: Pressure is judged by the eyebrow area. Although it is not as accurate as full face recognition, it eliminates many redundant discrimination features, greatly improving the efficiency of pressure calculation. Pressure can be equivalent to driving proficiency. In graded simulation scenarios, the higher the level of the scenario, the lower the pressure, and the lower the pressure, the higher the proficiency. When drivers face high-intensity training, they can use this application method to conduct targeted training and quickly master driving skills in the shortest time.

[0132] Example 2

[0133] See also Figure 4 , another embodiment provided by the present invention: a simulation driving training system for urban rail vehicles based on scenario simulation, comprising: a simulation scenario construction module, a driving training module and a training evaluation module;

[0134] The simulation scenario construction module is used to collect track line parameters, weather data, passenger flow density, vehicle equipment status and driving data in real time, and perform preprocessing to construct a dynamic and personalized simulation scenario;

[0135] The driving training module is used to simulate driving operations, identify micro-features of the driver's eyebrow area based on the driver's operations and preset simulation scenarios, analyze the driver's psychological stress in combination with historical data, and dynamically adjust the simulation scenarios based on the psychological stress analysis results;

[0136] The training evaluation module scores the driver's performance in this training according to preset evaluation criteria and generates a detailed training report.

[0137] Driving training module, including: dynamic coupling unit, pressure analysis unit and scenario adjustment unit;

[0138] The dynamic coupling unit is used to dynamically couple the six-degree-of-freedom motion platform;

[0139] The pressure analysis unit is used to build a force feedback model and associate it with odor events, providing real-time feedback on force and odor;

[0140] The scene adjustment unit is used to identify the micro features of the driver's eyebrow area, analyze the driver's psychological pressure in combination with historical data, and provide dynamic feedback, combining force, smell and pressure feedback to dynamically adjust the simulation scene.

[0141] In addition, the parts of the above technical solutions provided in the embodiments of the present application that are consistent with the implementation principles of the corresponding technical solutions in the prior art are not described in detail to avoid excessive redundancy.

[0142] The above-described specific embodiments further illustrate the objectives, technical solutions, and beneficial effects of the present invention. It should be understood that the above description is merely a specific embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. A scenario-based urban rail vehicle driving simulation training method, characterized in that: include: Collect simulated driving data in real time, perform pre-processing, and build dynamic and personalized simulation scenarios; Simulate driving operations, identify the driver's eyebrow area's micro-features based on the driver's operations and preset simulation scenarios, analyze the driver's psychological stress based on historical data, and dynamically adjust the simulation scenario based on the psychological stress analysis results; Based on the preset evaluation criteria, the driver's performance in training is scored and a training report is generated; The simulated driving operation identifies micro-features of the driver's eyebrow area based on the driver's operation and a preset simulation scene, analyzes the driver's psychological stress in combination with historical data, and dynamically adjusts the simulation scene based on the psychological stress analysis results, including: Dynamically couple the six-degree-of-freedom motion platform; Build a force feedback model and associate it with odor events to provide real-time feedback on force and odor; Identify the micro features of the driver's eyebrow area and analyze the driver's psychological stress by combining historical data; Every t time period, the simulation scene difficulty is automatically increased to simulate passenger behavior and provide dynamic feedback. The simulation scene is dynamically adjusted by combining force, smell and pressure feedback. The dynamic coupling of the six-degree-of-freedom motion platform includes: According to the vehicle dynamics model output and actual operation data, the relationship equation between the hydraulic cylinder length and posture is established and solved; Predict vehicle posture changes in the future, generate motion commands in advance, and compensate for response delays; Load the track irregularity spectrum to generate random vibration signals and adjust the stiffness according to the number of passengers; The method of identifying micro features of the driver's eyebrow area and analyzing the driver's psychological stress in combination with historical data includes: During a period of T in the driver's historical training, facial images of the driver within the period of T in the driver's historical training are collected through a first face collection window at a preset frame interval, and are set as first collected data; The eyes and eyebrows are cropped in the first collected data, and the areas outside the eyes and eyebrows are removed to obtain the second collected data; Reconstruct the eyebrow muscle deformation based on the 3D deformable model to capture the characteristics of the eyebrow region, including: orbicularis oculi muscle contraction rate, brow peak displacement, eyeball displacement, and blink duration; Simultaneously collect steering wheel grip and heart rate variability to build a triangular verification of eyebrow-physiology-operation, inject training time tags in real time, and calculate the temporal correlation between events and eyebrow-physiology-operation; Based on the temporal correlation between events and eyebrow-physiology-operation, the driver's real-time stress is calculated and a stress-time-scenario heat map is generated.

2. The urban rail vehicle simulation driving training method based on scenario simulation according to claim 1, characterized in that: The construction of a dynamic personalized simulation scenario includes: Based on the pre-processed simulated driving data, a sliding time window is used to classify driving errors and calculate their frequencies to construct a driver capability profile. Based on the state space and action space, a state-action reward model is constructed to dynamically adjust the driver's ability profile; A three-layer architecture is used to generate dynamic personalized hierarchical simulation scenarios, including a basic scenario layer, a disturbance enhancement layer, and an event chain weaving layer. The basic scenario layer includes a vehicle dynamics model and an environment interaction model.

3. The urban rail vehicle simulation driving training method based on scenario simulation according to claim 2, characterized in that: The driver's performance in training is scored based on the preset evaluation criteria, and a training report is generated, including: Real-time collection of driver evaluation data during historical training time period T, including operational specification data, physiological data, and emergency decision-making data, and pre-processing of the collected evaluation data; Evaluate the driver's training results based on the pre-processed evaluation data and preset evaluation criteria; Using the analytic hierarchy process, weights are assigned to pre-set evaluation criteria and the overall score of the driver training is calculated. Generate visual reports on driver training through 3D radar charts, pressure heat maps and causal chain diagrams.

4. The urban rail vehicle simulation driving training method based on scenario simulation according to claim 3 is characterized in that: The construction of the force feedback model and the association of the odor events to provide real-time feedback of force and odor includes: Calculate the vehicle steering resistance torque based on the wheel-rail adhesion coefficient and dynamic disturbance; Adaptive fuzzy PID control is used to dynamically adjust the damping according to the driver's operation smoothness; Correlate odor events and analyze the source of odors; Airflow guidance technology is used to diffuse the generated odor.

5. A scenario-based urban rail vehicle simulation driving training system, used to implement a scenario-based urban rail vehicle simulation driving training method according to any one of claims 1 to 4, characterized in that: include: Simulation scenario construction module, driving training module and training evaluation module; The simulation scene construction module is used to collect simulated driving data in real time, perform preprocessing, and construct dynamic personalized simulation scenes; The driving training module is used to simulate driving operations, identify micro-features of the driver's eyebrow area based on the driver's operations and preset simulation scenarios, analyze the driver's psychological stress in combination with historical data, and dynamically adjust the simulation scenarios based on the psychological stress analysis results; The training evaluation module scores the driver's performance during training according to preset evaluation criteria and generates a training report.

6. The urban rail vehicle simulation driving training system based on scenario simulation according to claim 5, characterized in that: The driving training module includes: a dynamic coupling unit, a pressure analysis unit and a scene adjustment unit; The dynamic coupling unit is used to dynamically couple the six-degree-of-freedom motion platform; The pressure analysis unit is used to build a force feedback model and associate it with odor events, providing real-time feedback on force and odor; The scene adjustment unit is used to identify the micro features of the driver's eyebrow area, analyze the driver's psychological pressure in combination with historical data, and provide dynamic feedback, combining force, smell and pressure feedback to dynamically adjust the simulation scene.

Images