Speed limit curve calculation method and system based on fusion of multiple parameters

Abstract

The embodiment of the invention provides a speed limit curve calculation method and system based on multi-parameter fusion, and the method comprises the steps: obtaining train multi-source data through a GNSS / INS module, a laser radar, and / or a signal machine, carrying out the preprocessing and space-time alignment of the multi-source data, carrying out the data validity verification, and obtaining the final train multi-source data; inputting the train multi-source data into a risk prediction model LSTM, and performing risk calculation on the train multi-source data by adopting a dynamic fusion algorithm through the LSTM to obtain a risk level of the train; and generating a segmented speed limit curve based on the risk level and a multi-level braking strategy, executing a speed limit instruction through an automatic train protection (ATP) system based on the segmented speed limit curve, and feeding back a control result.

Description

Technical Field

[0001] This document relates to the field of wind turbine blade technology, and in particular to a method and system for calculating speed limit curves based on the fusion of multiple parameters. Background Technology

[0002] In existing technologies, train speed limiting protection mainly relies on the following technologies:

[0003] 1. Static speed limits based on a fixed route database: Route design parameters (such as curve radius, gradient, and bridge speed limits) are pre-stored in the onboard database, and the train matches the preset speed limit value through positioning. Typical applications include: static speed curves in high-speed rail train control systems (such as CTCS-3); and route configuration files in metro automatic driving systems (such as CBTC).

[0004] 2. Passive protection led by the signal system: This method obtains the signal status (red / yellow / green) through track circuits or wireless communication, forcing trains to adhere to signal indications for speed limits or stopping. Typical applications include: traditional railway signaling systems (such as interlocking systems); and ATP protection in fixed block sections.

[0005] 3. Manual speed limit instruction coverage: The dispatch center or on-site personnel issue temporary speed limit instructions (TSR) via wireless communication, and the trains implement fixed speed limits throughout the entire section or in certain areas. Typical applications include: temporary speed limits in construction sections; and overall speed reduction during severe weather (such as strong winds and heavy rain).

[0006] 4. Simple multi-parameter combination: A few systems attempt to integrate 2-3 types of parameters (such as positioning + signal status) and use "if-else" logic to switch speed limit modes. Typical applications include: safety protection systems for some freight trains; and multi-sensor redundancy design in early autonomous driving test vehicles.

[0007] However, the aforementioned prior art has the following technical problems:

[0008] 1. Poor environmental adaptability: Static speed limits and signaling systems cannot respond to sudden dynamic risks (such as objects encroaching on the track or signal equipment malfunctions). For example, if a rockfall occurs ahead, the signal may still show a green light, and the traditional system will continue to operate at the original speed until manual intervention is required.

[0009] 2. The contradiction between conservatism and efficiency: Manual speed limit orders usually take effect globally, leading to wasted capacity in non-risk sections. For example: Actual need: only an 80km / h speed limit 1km before and after the construction zone; Traditional method: 80km / h speed limit for the entire section, increasing transit time by 15%.

[0010] 3. Insufficient multi-parameter coordination: Existing fusion methods do not resolve parameter conflicts. For example: a green light signal + radar obstacle detection → the system prioritizes the signal, leading to collision risks. Manual speed limit instructions + curve-designed speed limits → simply take the minimum value, ignoring real-time vehicle status (such as differences in braking under no-load / heavy load).

[0011] 4. Insufficient real-time performance: The delay from risk detection to rate limiting execution is too high (≥3 seconds), failing to meet the requirements of high-speed scenarios. The root causes include: time-sharing data processing without multi-source synchronization; and control logic based on periodic scanning (e.g., 1Hz update), rather than event-driven.

[0012] 5. Limited scalability: The system is difficult to integrate with new types of sensors (such as lidar and turnout health sensors). Summary of the Invention

[0013] The purpose of this invention is to provide a method and system for calculating speed limit curves based on the fusion of multiple parameters, in order to solve the above-mentioned problems in the prior art.

[0014] This invention provides a method for calculating speed limit curves based on the fusion of multiple parameters, including:

[0015] Train multi-source data is acquired through GNSS / INS modules, lidar, and / or signal controllers. The multi-source data is preprocessed and spatiotemporally aligned, and the data validity is verified to obtain the final train multi-source data.

[0016] The train multi-source data is input into the risk prediction model LSTM. The LSTM uses a dynamic fusion algorithm to calculate the risk of the train multi-source data and obtain the risk level of the train.

[0017] Based on the risk level and a multi-level braking strategy, a segmented speed limit curve is generated. Based on the segmented speed limit curve, the speed limit command is executed through the Automatic Train Protection (ATP) system, and the control results are fed back.

[0018] This invention provides a speed limit curve calculation system based on the fusion of multiple parameters, comprising:

[0019] The perception layer is used to acquire multi-source train data through GNSS / INS modules, lidar, and / or signals;

[0020] The edge computing layer is used to preprocess and spatiotemporally align the multi-source data, perform data validity verification, and obtain the final train multi-source data.

[0021] The decision layer is used to input the multi-source train data into the risk prediction model LSTM, and use the LSTM to perform risk calculation on the multi-source train data using a dynamic fusion algorithm to obtain the risk level of the train; based on the risk level and based on a multi-level braking strategy, a segmented speed limit curve is generated.

[0022] The execution layer is used to execute speed limit commands through the Automatic Train Protection (ATP) system based on the segmented speed limit curves and to provide feedback on the control results.

[0023] This invention also provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it implements the steps of the above-described method for calculating the speed limit curve based on the fusion of multiple parameters.

[0024] This invention also provides a computer-readable storage medium storing an information transmission implementation program, which, when executed by a processor, implements the steps of the above-described method for calculating the speed limit curve based on the fusion of multiple parameters.

[0025] The embodiments of the present invention solve the core problems of existing train speed limit protection technologies, such as poor environmental adaptability, difficulty in balancing efficiency and safety, and insufficient multi-parameter coordination. By integrating multi-source parameters such as real-time positioning, signal equipment status, obstacle data, and manual speed limit commands, the dynamic risk response capability can be improved, the accuracy and efficiency of speed limit control can be optimized, the multi-parameter conflict problem can be resolved, the system scalability and compatibility can be enhanced, and the reliance on manual intervention can be reduced. Attached Figure Description

[0026] To more clearly illustrate the technical solutions in one or more embodiments of this specification or in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0027] Figure 1 This is a flowchart of a speed limit curve calculation method based on the fusion of multiple parameters according to an embodiment of the present invention;

[0028] Figure 2 This is a schematic diagram of the speed limit curve calculation method based on the fusion of multiple parameters according to an embodiment of the present invention;

[0029] Figure 3 This is a schematic diagram of a speed limit curve calculation system based on the fusion of multiple parameters according to an embodiment of the present invention;

[0030] Figure 4This is a schematic diagram of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0031] To enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the technical solutions in one or more embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this specification, and not all of the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of this document.

[0032] Method Implementation Examples

[0033] According to embodiments of the present invention, a method for calculating speed limit curves based on the fusion of multiple parameters is provided. Figure 1 This is a flowchart of a speed limit curve calculation method based on the fusion of multiple parameters according to an embodiment of the present invention, such as... Figure 1 As shown, the speed limit curve calculation method based on multi-parameter fusion according to an embodiment of the present invention specifically includes:

[0034] Step S101 involves acquiring multi-source train data via GNSS / INS module, lidar, and / or signal controller; preprocessing and spatiotemporally aligning the multi-source data; and verifying data validity to obtain the final multi-source train data. Specifically, this includes:

[0035] The system receives BeiDou / GPS dual-frequency signals via a GNSS module, outputs latitude, longitude, altitude, and speed, and switches to inertial navigation (INS) when the signal is lost. It also collects equipment status signals when the train passes the track tags.

[0036] The signal lights are read from the trackside LoRa nodes to monitor the status of the turnouts;

[0037] Obstacle information is obtained by filtering and clustering the original point cloud using lidar.

[0038] The IEEE 1588PTP protocol is used to perform microsecond-level time synchronization of train multi-source data, compensate for the fixed delay of each sensor, perform spatiotemporal alignment through calibration matrix, align the coordinate system to the center of the car body, and perform data validity verification according to preset rules to obtain the final train multi-source data.

[0039] In this embodiment of the invention, if the lidar malfunctions, it switches to pure vision mode and the speed limit is reduced to 60%. If communication is interrupted, RFID and odometer are used, and the error alarm threshold is set to ±50m.

[0040] Step S102 involves inputting the multi-source train data into the risk prediction model LSTM, and using the LSTM to perform risk calculation on the multi-source train data using a dynamic fusion algorithm to obtain the risk level of the train; specifically including:

[0041] The obstacle distance, relative speed, signal status, track slippage, train load, curve radius, and gradient are input into the risk prediction model LSTM. The LSTM is used to perform risk calculation on the multi-source train data using a dynamic fusion algorithm. The weights of the multi-source train data are assigned based on sensor reliability to obtain the risk level of the train between 0 and 1.

[0042] Step S103: Based on the risk level and a multi-level braking strategy, a segmented speed limit curve is generated. Based on the segmented speed limit curve, the speed limit command is executed through the Automatic Train Protection (ATP) system, and the control result is fed back. Specifically, this includes:

[0043] According to Formula 1, a speed limit point is calculated every 1 meter:

[0044] V limit =2· amax ·(Dobs- Dsafe ) Formula 1;

[0045] Where, amax is the braking deceleration calculated in real time based on the load, Dobs represents the distance to the obstacle, Dsafe represents the safe distance, and Vlimit represents the speed limit at the speed limit point;

[0046] Based on the priority rules of obstacle > red light > manual speed limit > green light > planned curve and the conflict arbitration rules, a segmented speed limit curve is generated according to the calculated speed limit point, the risk level and the multi-level braking strategy.

[0047] Based on the segmented speed limit curve, a speed limit command frame is generated by encapsulating MVB data packets and sent to the Automatic Train Protection (ATP) system. The ATP system executes the speed limit command and provides feedback on the control results.

[0048] In summary, the technical solutions of the embodiments of the present invention solve the core problems of existing train speed limit protection technologies, such as poor environmental adaptability, difficulty in balancing efficiency and safety, and insufficient multi-parameter coordination. By integrating multi-source parameters such as real-time positioning, signal equipment status, obstacle data, and manual speed limit commands, the following objectives are achieved:

[0049] 1. By integrating obstacle data from lidar, video surveillance, and other sources in real time, as well as dynamically monitoring changes in the status of signals and switches, speed limit curve updates are triggered, improving dynamic risk response capabilities: achieving millisecond-level response to dynamic risks such as sudden obstacles and signal equipment failures.

[0050] 2. This invention generates segmented speed limit curves based on risk levels (high / medium / low), reducing speed only in high-risk sections. By combining parameters such as vehicle load and braking performance, the safe braking distance is dynamically calculated, optimizing the accuracy and efficiency of speed limit control, avoiding globally conservative speed limits, and maximizing transportation efficiency while ensuring safety.

[0051] 3. This invention employs a priority rule: obstacle > red traffic light > manual speed limit > green traffic light, and fuzzy logic arbitration: for ambiguous scenarios (such as a green traffic light but radar detecting an obstacle), a tiered response (speed reduction + manual confirmation) is triggered. This resolves multi-parameter conflict issues and intelligently arbitrates contradictions between the signal system, obstacle detection, and manual instructions.

[0052] 4. This invention employs a modular design, reserving standardized interfaces (such as those for lidar and turnout health monitoring data). A dynamic weight allocation algorithm automatically adapts to newly added parameter types. This enhances system scalability and compatibility, supporting rapid integration of new sensors and data sources.

[0053] 5. This invention automatically identifies scenarios such as temporary construction zones and weather disasters, generates recommended speed limit curves, and activates an autonomous protection mode (based on local sensor data) in abnormal situations (such as communication interruptions). This reduces reliance on manual intervention and minimizes safety hazards caused by overload or delays in manual speed limit commands. Table 1 compares this invention with existing technologies.

[0054] Table 1

[0055]

[0056] It should be noted that the application scenarios of the technical solutions in the embodiments of the present invention are as follows:

[0057] 1. Sudden Obstacle: Laser detects an object encroaching on the track → immediately generates a steep speed reduction curve to avoid collision. 2. Signal Failure: Switch disconnection but no obstruction ahead → dynamically relaxes the speed limit according to manual speed limit instructions to reduce delays. 3. Temporary Construction: Automatically identifies the electronic fence of the construction area → locally limits speed, while full speed is restored in non-construction sections.

[0058] The technical solutions of the embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0059] like Figure 2 As shown, the technical solution of this embodiment of the invention specifically includes the following processes:

[0060] Step 1: Multi-source data acquisition and time synchronization

[0061] 1. GNSS / INS Positioning Data Acquisition: Receives BeiDou / GPS dual-frequency signals via a GNSS module, outputting latitude, longitude, altitude, and speed (frequency 10Hz). In case of signal loss, switches to inertial navigation (INS), with an error accumulation of <0.1% / min. RFID Calibration: Collects equipment status signals when the train passes a track tag (encoding format: RFID_KM123.5_LAT31.2_LON121.4).

[0062] 2. Signal Data: Signal light status is read via trackside LoRa nodes (anomaly detection: a fault alarm is triggered if three consecutive red lights are read). Turnout status is also monitored: strain gauges measure turnout switch rail displacement (accuracy ±0.1mm).

[0063] 3. Obstacle Detection Process: LiDAR Processing: Filtering the original point cloud (removing fixed structure points on the track). DBSCAN Clustering (minimum number of cluster points = 5, neighborhood radius = 0.3m). Bounding Box Generation (outputting obstacle center coordinates and dimensions).

[0064] 4. Manual instruction parsing: GSM-R message decoding example: 0xA1 0x02 [starting KM][ending KM][rate limit] 0xAA; Request retransmission if checksum error (maximum 3 times).

[0065] Step 2: Data Preprocessing and Spatiotemporal Alignment

[0066] 1. Obstacle detection process: PTP master clock (GNSS module) synchronizes with each slave device: ptpd -i eth0-Gu 0 -f / etc / ptpd.conf, to compensate for fixed sensor delays: LiDAR: +12ms; Camera: +50ms.

[0067] 2. Coordinate System 1: Transform all data to the vehicle coordinate system (X forward, Y left, Z upward): The transformation matrix from lidar to vehicle coordinates is as follows:

[0068] T_lidar2body=[0.99 -0.05 0 1.2; 0.05 0.99 0 0; 0 0 1 2.5];

[0069] 3. Data validity verification: Outlier removal rules: Data is deemed invalid and marked when either of the following two conditions is met: the train speed collected by GNSS is greater than 400km / h; or the difference between the mileage marker value read by the RFID device and the previous mileage marker value is greater than 0.5km.

[0070] Step 3: Dynamic Risk Quantification and Weight Allocation

[0071] 1. LSTM Risk Prediction Model: Input Layer (7-dimensional): [Obstacle Distance, Relative Velocity, Signal Status (0 / 1), Track Slipperyness, Load, Curve Radius, Gradient]

[0072] The network structure is as follows:

[0073] model.add(LSTM(64,input_shape=(10,7))) #10 historical time steps

[0074] model.add(Dense(1,activation='sigmoid'))

[0075] input_shape = (10, 7) represents the time series structure of the model input:

[0076] The first dimension, 10, represents data from 10 historical time steps (i.e., observations from 10 consecutive moments).

[0077] The second dimension, number 7: 7-dimensional feature data corresponding to each time step.

[0078] LSTM(64) indicates that this layer contains 64 LSTM neurons (memory units).

[0079] Dense(1) indicates that the output layer contains 1 neuron, corresponding to the final risk level prediction result.

[0080] activation='sigmoid' uses the sigmoid activation function to compress the output value between 0 and 1.

[0081] 2. Dynamic Weight Calculation: Confidence-based weight allocation: LiDAR weight (w_lidar): When the LiDAR confidence (lidar.confidence) is greater than 0.8, the weight is assigned 0.9; otherwise (confidence ≤ 0.8), the weight is assigned 0.4. Signal data weight (w_signal): When the signal data lifetime (signal.age, i.e., the time since data generation) is less than 100 milliseconds, the weight is assigned 0.7; otherwise (data lifetime ≥ 100 milliseconds), the weight is assigned 0.3. The LiDAR weight, signal weight, and manual input data weight (w_manual) are normalized.

[0082] Step 4: Speed ​​Limit Curve Generation and Conflict Arbitration

[0083] 1. Segmented speed limit calculation: A speed limit value is calculated for every 1 meter. If the distance from the current calculation point to the train plus the required braking distance is still less than the total distance from the obstacle to the train, the base speed is maintained; otherwise, the speed limit value at that point is calculated by formula 2.

[0084] 2. Multi-stage braking strategy, as shown in Table 2:

[0085] Table 2

[0086]

[0087] Different braking schemes with varying intensities are adopted based on the risk level of train operation:

[0088] Low risk: using 0.8m / s 2 The deceleration is such that it takes 1200 meters to come to a complete stop from 160 km / h, offering the highest level of comfort (4 stars).

[0089] Medium risk: using 1.5m / s 2 The deceleration is moderate, requiring 650 meters to come to a complete stop from 160 km / h. Comfort level is moderate (2 stars).

[0090] High risk: using 3.0m / s 2 The deceleration is such that it takes 320 meters to come to a complete stop from 160 km / h, resulting in the lowest level of comfort (1 star).

[0091] The higher the risk level, the greater the braking intensity (greater deceleration), the shorter the required braking distance, but the worse the comfort.

[0092] When multiple control signals (obstacle risk, signal status, manual commands, etc.) are present, they shall be processed according to the following priority:

[0093] If the obstacle risk level is >0.8 (high risk):

[0094] Even if the traffic signal shows green (allowed to proceed), the "Traffic Signal Override" mechanism will be triggered, forcibly performing a speed reduction operation (ignoring the green signal) and recording a "obstacle takes precedence over traffic signal" conflict log. Finally, the emergency braking speed (corresponding to the high-risk braking parameters in the table) will be returned.

[0095] If the obstacle risk level is ≤0.8 (not high risk):

[0096] When an active manual speed limit instruction exists, the smaller value between the manual speed limit and the line design speed is taken as the current speed limit (i.e., following the logic of directional safety).

[0097] The pseudocode for the arbitration logic is as follows:

[0098]

[0099] Step 5: Issuance and Verification of Control Commands

[0100] 1. MVB data packet encapsulation: The speed limit instruction includes the speed limit, the distance corresponding to the speed limit (i.e., the distance to the current position), and the braking gear value.

[0101] 2. Braking system response test: Verification items: from 160km / h to 80km / h, the actual deceleration error is <5%, and the command transmission delay is <8ms.

[0102] 3. Human-Computer Interface Interaction: HUD Display Elements: [Current] 128km / h [Restriction] 80km / h (Distance to Conflict Point: 215m), Obstacle Warning Ahead!

[0103] Step 6: Fault Handling and Degradation Mode:

[0104] 1. Handling sensor failures: LiDAR failure → switch to pure vision mode (speed limit reduced to 60%), GNSS loss → rely on RFID + odometer (error alarm threshold ±50m).

[0105] 2. Communication interruption handling:

[0106] If the signal interruption between the train and the ground control system lasts for more than 10 seconds, the train will perform the following actions:

[0107] ● Continue operation using the last valid control command received.

[0108] Activate the limited autonomous mode (in which the train autonomously controls itself based on locally stored track data and sensor information to achieve safe deceleration and stopping, rather than relying on real-time ground commands).

[0109] The system used in the above method is a "four-layer three-bus" architecture, which realizes closed-loop control of the entire process from data acquisition to rate limiting execution.

[0110] 2. The four-layer architecture is shown in Table 3:

[0111] Table 3

[0112]

[0113] 3. The three-bus design is shown in Table 4:

[0114] Table 4

[0115]

[0116]

[0117] For example, a train is traveling at 160 km / h. A lidar detects a tool left on the track 200 meters away (risk factor 0.8), the signal light is green, and the dispatcher has not issued a speed limit. Using existing technology, maintaining the 160 km / h speed relies on the driver's visual braking. The technical solution of this invention identifies high-risk obstacles and covers signal status. It generates a segmented speed limit curve: 0–50m: 160 → 100 km / h (deceleration 1.0 m / s²). 2 ), 50~150m: 100→30km / h (deceleration 2.5m / s²) 2 It can stop completely in front of obstacles, and the whole process is automated and requires no human intervention.

[0118] The technical solution of this invention achieves high precision, high real-time performance, and strong robustness in train speed limit protection through dynamic fusion of multi-source data and intelligent decision-making mechanisms. The following are six key points and their technical implementations:

[0119] 1. Strict spatiotemporal alignment of multiple sensors

[0120] Technical Implementation: Microsecond-level time synchronization is achieved using the IEEE 1588PTP protocol to compensate for the fixed delays of each sensor (e.g., LiDAR +12ms, camera +50ms). The coordinate system is aligned to the vehicle body center, and data is transformed through a calibration matrix: Tlidar→body=[0.99-0.0501.20.050.99000012.5]Tlidar→body=0.990.050-0.050.9900011.202.5. This eliminates fusion errors caused by time asynchrony or coordinate system differences.

[0121] 2. Dynamic Risk Quantification Model

[0122] Technical implementation: The LSTM neural network takes seven parameters as input, including obstacle distance, relative speed, signal status, track slipperiness, load, curve radius and gradient. After processing by the network, it outputs a risk coefficient between 0 and 1 to quantify the risk level of train operation.

[0123] Meanwhile, the system has real-time adaptive capabilities. For example, in rainy weather, it will automatically increase the weight of track slippage in risk assessment (increasing the weight coefficient λ from 1.0 to 1.5) to more accurately reflect the impact of rain on train safety.

[0124] When the risk score calculated by the model based on the input parameters is greater than 0.7, it means that the current risk is high, and the system will trigger emergency braking to ensure train operation safety. This overcomes the inability of traditional threshold methods to quantify risks in complex scenarios.

[0125] 3. Multi-parameter dynamic weight allocation

[0126] Technical Implementation: Weight Calculation Based on Sensor Reliability

[0127]

[0128] in:

[0129] wi represents the weight of the i-th sensor.

[0130] Ri represents the reliability of the i-th sensor (the larger the value, the higher the reliability).

[0131] λ is an adjustment coefficient (used to control the degree of influence of unreliability on the weights, default 1.0, 1.5 for rainy or snowy days).

[0132] e is the natural constant (approximately 2.718).

[0133] n is the total number of sensors.

[0134] 4. Generation of segmented speed limit curves

[0135] Technical implementation: Calculate a speed limit point every 1 meter, formula:

[0136] Vlimit =2· amax ·(Dobs- Dsafe )

[0137] Where amax is the braking deceleration calculated in real time based on the load (1.2 m / s² when unloaded). 2 Heavy load 0.8m / s 2 This solves the problem of low efficiency caused by globally unified rate limiting.

[0138] 5. Multi-level conflict arbitration mechanism

[0139] Technical implementation: Priority rule: Obstacle > Red traffic light > Manual speed limit > Green traffic light > Planned curve; For example: If the obstacle risk level is >0.8 (high risk) and the traffic light is green (allowed to pass), a forced speed reduction is still required.

[0140] This solves the problem of system rigidity when parameters are contradictory.

[0141] 6. Fault Degradation and Autonomous Mode

[0142] Technical Implementation: Sensor Failure: When the LiDAR fails, the system switches to pure vision mode (speed limit automatically reduced to 60%). Communication Interruption: After GSM-R is lost, local RFID + odometer combined positioning is activated (alarm triggered if error ±50m). Self-Test Rule: If the GSM-R communication interruption lasts for more than 10 seconds and there is no valid RFID signal, the system will enter safety mode, at which time the speed limit is set to 40 (unit usually km / h). This solves the problem of single-point failure leading to functional loss in traditional systems.

[0143] System Implementation Examples

[0144] According to embodiments of the present invention, a speed limit curve calculation system based on the fusion of multiple parameters is provided. Figure 3 This is a schematic diagram of a speed limit curve calculation system based on the fusion of multiple parameters according to an embodiment of the present invention, such as... Figure 3 As shown, the speed limit curve calculation system based on multi-parameter fusion according to an embodiment of the present invention specifically includes:

[0145] The perception layer 30 is used to acquire multi-source train data through GNSS / INS modules, lidar, and / or signal controllers. Specifically, it is used to: receive BeiDou / GPS dual-frequency signals through the GNSS module, output latitude, longitude, altitude, and speed, and switch to inertial navigation (INS) when the signal is lost; collect equipment status signals when the train passes track tags; use the signal controller to read the signal light status through the trackside LoRa node and monitor the switch status; filter and cluster the raw point cloud through lidar to obtain obstacle information; if the lidar fails, switch to pure vision mode and reduce the speed limit to 60%; if communication is interrupted, use RFID and odometer, and set the error alarm threshold to ±50m.

[0146] Edge computing layer 32 is used to preprocess and spatiotemporally align the multi-source data, perform data validity verification, and obtain the final train multi-source data. Specifically, it is used to: perform microsecond-level time synchronization of the train multi-source data using the IEEE 1588PTP protocol, compensate for the fixed delay of each sensor, perform spatiotemporal alignment through a calibration matrix, align the coordinate system to the center of the vehicle body, and perform data validity verification according to preset rules to obtain the final train multi-source data.

[0147] The decision layer 34 is used to input the multi-source train data into the risk prediction model LSTM, and to perform risk calculation on the multi-source train data using a dynamic fusion algorithm through LSTM to obtain the risk level of the train; based on the risk level and a multi-level braking strategy, a segmented speed limit curve is generated; specifically, it is used to: input obstacle distance, relative speed, signal status, track slippage, train load, curve radius and gradient into the risk prediction model LSTM, and to perform risk calculation on the multi-source train data using a dynamic fusion algorithm through LSTM, and to perform weight allocation on the multi-source train data based on sensor reliability to obtain the risk level of the train between 0 and 1;

[0148] According to Formula 1, a speed limit point is calculated every 1 meter:

[0149] Vlimit =2· amax ·(Dobs- Dsafe ) Formula 1;

[0150] Where, amax is the braking deceleration calculated in real time based on the load, Dobs represents the distance to the obstacle, Dsafe represents the safe distance, and Vlimit represents the speed limit at the speed limit point;

[0151] Based on the priority rules of obstacle > red light > manual speed limit > green light > planned curve and the conflict arbitration rules, a segmented speed limit curve is generated according to the calculated speed limit point, the risk level and the multi-level braking strategy.

[0152] Execution layer 36 is used to execute speed-limiting commands through the Automatic Train Protection (ATP) system based on the segmented speed-limiting curves and to provide feedback on the control results. Specifically, it is used to: generate speed-limiting command frames by encapsulating MVB data packets based on the segmented speed-limiting curves, send them to the ATP system, execute the speed-limiting commands through the ATP system, and provide feedback on the control results.

[0153] The system in this embodiment of the invention adopts a "four-layer three-bus" architecture to realize closed-loop control of the entire process from data acquisition to rate-limited execution.

[0154] 2. The four-layer architecture is shown in Table 3:

[0155] Table 3

[0156]

[0157]

[0158] 3. The three-bus design is shown in Table 4:

[0159] Table 4

[0160]

[0161] The embodiments of the present invention are device embodiments corresponding to the above method embodiments. The specific operation of each module can be understood with reference to the description of the method embodiments, and will not be repeated here.

[0162] Device Example 1

[0163] This invention provides an electronic device, such as... Figure 4 As shown, it includes: a memory 40, a processor 42, and a computer program stored in the memory 40 and executable on the processor 42, wherein the computer program, when executed by the processor 42, performs the steps as described in the method embodiment.

[0164] Device Example 3

[0165] This invention provides a computer-readable storage medium storing an information transmission implementation program, which, when executed by a processor 42, performs the steps described in the method embodiment.

[0166] The computer-readable storage media described in this embodiment include, but are not limited to, ROM, RAM, disk, or optical disk.

[0167] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for calculating speed limit curves based on the fusion of multiple parameters, characterized in that, include: Train multi-source data is acquired through GNSS / INS modules, lidar, and / or signal controllers. The multi-source data is preprocessed and spatiotemporally aligned, and the data validity is verified to obtain the final train multi-source data. The train multi-source data is input into the risk prediction model LSTM. The LSTM uses a dynamic fusion algorithm to calculate the risk of the train multi-source data and obtain the risk level of the train. Based on the risk level and a multi-level braking strategy, a segmented speed limit curve is generated. Based on the segmented speed limit curve, the speed limit command is executed through the Automatic Train Protection (ATP) system, and the control results are fed back.

2. The method according to claim 1, characterized in that, Train multi-source data is acquired through GNSS / INS modules, lidar, and / or signal controllers. This multi-source data undergoes preprocessing and spatiotemporal alignment, and data validity verification to obtain the final train multi-source data, which specifically includes: The system receives BeiDou / GPS dual-frequency signals via a GNSS module, outputs latitude, longitude, altitude, and speed, and switches to inertial navigation (INS) when the signal is lost. It also collects equipment status signals when the train passes the track tags. The signal lights are read from the trackside LoRa nodes to monitor the status of the turnouts; Obstacle information is obtained by filtering and clustering the original point cloud using lidar. The IEEE 1588PTP protocol is used to perform microsecond-level time synchronization of train multi-source data, compensate for the fixed delay of each sensor, perform spatiotemporal alignment through calibration matrix, align the coordinate system to the center of the car body, and perform data validity verification according to preset rules to obtain the final train multi-source data.

3. The method according to claim 1, characterized in that, The train multi-source data is input into the risk prediction model LSTM. The LSTM uses a dynamic fusion algorithm to calculate the risk of the train multi-source data, and the specific risk level of the train is obtained, including: The obstacle distance, relative speed, signal status, track slippage, train load, curve radius, and gradient are input into the risk prediction model LSTM. The LSTM is used to perform risk calculation on the multi-source train data using a dynamic fusion algorithm. The weights of the multi-source train data are assigned based on sensor reliability to obtain the risk level of the train between 0 and 1.

4. The method according to claim 1, characterized in that, Based on the risk level and a multi-level braking strategy, segmented speed limit curves are generated. Speed ​​limit commands are executed via the Automatic Train Protection (ATP) system based on these segmented speed limit curves, and the control results are fed back. Specifically, this includes: According to Formula 1, a speed limit point is calculated every 1 meter: Vlimit =2 amax ·(Dobs- Dsafe ) Formula 1; Where, amax is the braking deceleration calculated in real time based on the load, Dobs represents the distance to the obstacle, Dsafe represents the safe distance, and Vlimit represents the speed limit at the speed limit point; Based on the priority rules of obstacle > red light > manual speed limit > green light > planned curve and the conflict arbitration rules, a segmented speed limit curve is generated according to the calculated speed limit point, the risk level and the multi-level braking strategy. Based on the segmented speed limit curve, a speed limit command frame is generated by encapsulating MVB data packets and sent to the Automatic Train Protection (ATP) system. The ATP system executes the speed limit command and provides feedback on the control results.

5. The method according to claim 1, characterized in that, The method further includes: If the lidar malfunctions, switch to pure vision mode and reduce the speed limit to 60%. If communication is interrupted, use RFID and odometer, and set the error alarm threshold to ±50m.

6. A speed limit curve calculation system based on multi-parameter fusion, characterized in that, include: The perception layer is used to acquire multi-source train data through GNSS / INS modules, lidar, and / or signals; The edge computing layer is used to preprocess and spatiotemporally align the multi-source data, perform data validity verification, and obtain the final train multi-source data. The decision layer is used to input the multi-source train data into the risk prediction model LSTM, and to calculate the risk level of the train by using a dynamic fusion algorithm on the multi-source train data through LSTM. Based on the aforementioned risk level and a multi-level braking strategy, a segmented speed limit curve is generated. The execution layer is used to execute speed limit commands through the Automatic Train Protection (ATP) system based on the segmented speed limit curves and to provide feedback on the control results.

7. The system according to claim 6, characterized in that, The perception layer is specifically used for: receiving BeiDou / GPS dual-frequency signals through a GNSS module, outputting latitude, longitude, altitude, and speed, and switching to inertial navigation (INS) when the signal is lost; collecting equipment status signals when the train passes the track tag; using a signal controller to read the signal light status through a trackside LoRa node and monitoring the switch status; and filtering and clustering the original point cloud using a lidar to obtain obstacle information. The edge computing layer is specifically used for: performing microsecond-level time synchronization of train multi-source data using the IEEE 1588PTP protocol, compensating for the fixed delay of each sensor, performing spatiotemporal alignment through a calibration matrix, aligning the coordinate system to the center of the vehicle body, verifying the validity of the data according to preset rules, and obtaining the final train multi-source data. The decision layer is specifically used to: input obstacle distance, relative speed, signal status, track slipperiness, train load, curve radius and gradient into the risk prediction model LSTM; use LSTM to perform risk calculation on the multi-source train data using a dynamic fusion algorithm; and allocate weights to the multi-source train data based on sensor reliability to obtain the risk level of the train between 0 and 1. According to Formula 1, a speed limit point is calculated every 1 meter: Vlimit =2 amax ·(Dobs- Dsafe ) Formula 1; Where, amax is the braking deceleration calculated in real time based on the load, Dobs represents the distance to the obstacle, Dsafe represents the safe distance, and Vlimit represents the speed limit at the speed limit point; Based on the priority rules of obstacle > red light > manual speed limit > green light > planned curve and the conflict arbitration rules, a segmented speed limit curve is generated according to the calculated speed limit point, the risk level and the multi-level braking strategy. The execution layer is specifically used to: generate a speed limit instruction frame by encapsulating MVB data packets based on the segmented speed limit curve, send it to the Automatic Train Protection (ATP) system, execute the speed limit instruction through the ATP system, and provide feedback on the control results.

8. The system according to claim 6, characterized in that, The sensing layer is further used for: If the lidar malfunctions, switch to pure vision mode and reduce the speed limit to 60%. If communication is interrupted, use RFID and odometer, and set the error alarm threshold to ±50m.

9. An electronic device, characterized in that, include: The memory, the processor, and the computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the speed limit curve calculation method based on the fusion of multiple parameters as described in any one of claims 1 to 5.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores an information transmission implementation program, which, when executed by a processor, implements the steps of the speed limit curve calculation method based on the fusion of multiple parameters as described in any one of claims 1 to 5.

Images