Fire-fighting vehicle rollover risk prevention and control method and system

Abstract

The invention provides a fire-fighting vehicle rollover risk prevention and control method and system, and the method comprises the steps: S2, integrating an inclination angle and tire pressure data according to a standardized sensor data set, calculating the real-time center-of-gravity offset of a vehicle, and generating a center-of-gravity offset feature value; s4, according to the rollover risk level, optimizing the pressure value and the displacement amount of the hydraulic lateral movement device, and generating a hydraulic regulation and control instruction set; s7, according to the dynamic environment adaptation model, fusing the gravity center shift characteristic value and the stationary movement control sequence, generating a transverse displacement path plan, and obtaining a safe displacement track containing path point data; and S8, extracting path point data from the safe displacement track, transmitting the path point data to a hydraulic lateral displacement device in real time, executing lateral displacement operation, and generating a vehicle dynamic state data set. The rollover risk of the vehicle can be effectively prevented and controlled, and the driving safety is improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent driving technology for fire trucks, and in particular to a method and system for preventing and controlling the risk of fire truck rollover. Background Technology

[0002] As core equipment for urban emergency rescue, fire trucks are crucial for fire fighting and disaster response due to their efficiency and safety. With increasing urban building density and the emergence of complex terrain, the need for fire trucks to be mobile in confined spaces or special scenarios is becoming increasingly prominent, especially in lateral movement operations, where it is necessary to ensure vehicle stability and operational precision.

[0003] Currently, lateral displacement of fire trucks mainly relies on manual judgment and operation, which generally suffers from response lag due to environmental complexity. Especially during high-intensity rescue missions, operators find it difficult to monitor the vehicle's dynamic status in real time, easily leading to safety hazards due to misjudgments of the environment or vehicle posture. In addition, existing systems mostly rely on data from a single sensor, lacking the ability to comprehensively analyze and dynamically control multi-source data, making it difficult to achieve precise coordination in complex scenarios.

[0004] During lateral movement, fire trucks must cope with factors such as terrain changes, obstacle distribution, and shifts in their own center of gravity. The core challenge lies in achieving dynamic safety judgment and real-time control capabilities. Real-time monitoring of vehicle tilt is the primary technical challenge, as it directly affects the safety of the movement operation. For example, in narrow alleys, if uneven ground causes the fire truck to tilt beyond a safe range during lateral movement, existing systems cannot automatically halt the operation, easily leading to a rollover risk. Furthermore, the dynamic changes in tilt require the system to adjust the working intensity of the hydraulic lateral movement device in real time based on tire pressure distribution to ensure a smooth movement rhythm.

[0005] However, existing technologies lack sufficient synergy between multi-sensor data fusion and dynamic control, making it difficult to achieve accurate responses in rapidly changing rescue scenarios.

[0006] Therefore, how to dynamically determine the safety of the lateral displacement of fire trucks and coordinate the control rhythm of hydraulic devices by analyzing multi-source sensor data in real time has become a key issue in improving the mobility and safety of fire trucks. Summary of the Invention

[0007] In a first aspect, the present invention provides a method for preventing the rollover risk of fire trucks, the method comprising: S1 acquires tilt angle, tire pressure, terrain height, and obstacle distance data from multi-source sensors in real time, removes noise, and generates a standardized sensor dataset. S2, based on a standardized sensor dataset, integrates tilt angle and tire pressure data to calculate the real-time center of gravity offset of the vehicle and generate center of gravity offset feature values; S3. Based on the comparison between the center of gravity offset characteristic value and the preset first safety threshold, when the center of gravity offset characteristic value exceeds the preset first safety threshold, the correlation between the tilt angle and the terrain height is analyzed through the rollover risk assessment model to generate a rollover risk level. S4, based on the rollover risk level, optimize the pressure value and displacement of the hydraulic lateral displacement device, and generate a hydraulic control command set; S5 extracts the target pressure value and displacement from the hydraulic control instruction set, combines the tire pressure data in the standardized sensor dataset, calculates the moving speed and frequency parameters, and generates a smooth moving control sequence. S6. Based on the smooth movement control sequence and obstacle distance data in the standardized sensor dataset, when the obstacle distance is less than the preset distance threshold, analyze the obstacle distribution, terrain height change trend and center of gravity offset characteristic value to generate a dynamic environment adaptation model. S7. Based on the dynamic environment adaptation model, the center of gravity offset characteristic value and the steady movement control sequence are integrated to generate the lateral displacement path planning and obtain the safe displacement trajectory containing path point data. S8 extracts path point data from the safe displacement trajectory and transmits it to the hydraulic lateral displacement device in real time to perform lateral displacement operation and generate a vehicle dynamic status dataset. S9 updates the standardized sensor dataset based on the vehicle dynamic state dataset, and iteratively executes the process from generating center of gravity offset feature values ​​to generating safe displacement trajectory, continuously optimizing displacement control parameters.

[0008] Optionally, step S2, based on a standardized sensor dataset, integrates tilt angle and tire pressure data to calculate the real-time center of gravity offset of the vehicle and generate center of gravity offset feature values, including: Step S21: Obtain standardized tilt angle and tire pressure data from the sensor dataset, remove noise through median filtering, and obtain the cleaned dataset; Step S22: Based on the cleaned dataset, calculate the weighted average of the tilt angle and tire pressure data to generate a preliminary estimate of the center of gravity offset. Step S23: If the preliminary estimate of the center of gravity offset exceeds the preset second safety threshold, the estimate of the center of gravity offset is smoothed to obtain the smoothed center of gravity offset value. Step S24: For the smoothed centroid offset value, calculate the covariance matrix and extract the first two largest eigenvalues ​​to generate the centroid offset feature vector. Step S25: Obtain the offset direction and magnitude information from the center of gravity offset feature vector, and calculate the real-time center of gravity offset feature value; Step S26: Based on the real-time center of gravity offset feature value, determine the vehicle's center of gravity offset state and obtain the vehicle stability classification result; Step S27: Generate output data based on the vehicle stability classification results.

[0009] Optionally, step S25, obtaining offset direction and magnitude information from the center of gravity offset feature vector and calculating real-time center of gravity offset feature value, includes: The following function is used to calculate the real-time centroid offset eigenvalues: f(x) = gx + b Where f(x) is the real-time centroid offset feature function, x is the amplitude information, g is the preset weight, and b is the offset constant.

[0010] Optionally, step S3 involves comparing the center of gravity offset characteristic value with a preset first safety threshold. If the center of gravity offset characteristic value exceeds the preset first safety threshold, the correlation between the tilt angle and terrain height is analyzed using a rollover risk assessment model to generate a rollover risk level, including: Step S31: Obtain real-time vehicle center of gravity offset feature value and terrain height data through sensors; Step S32: If the real-time centroid offset feature value exceeds the preset first safety threshold, input it into the support vector machine classifier to obtain the initial risk assessment result. Step S33: Extract the tilt angle from the initial risk assessment results; Step S34: Fit the tilt angle and terrain height data to determine the correlation coefficient; Step S35: Calculate the rollover risk index value based on the correlation coefficient and vehicle stability parameters; Step S36: If the rollover risk index value is higher than the preset risk threshold, the rollover risk level is adjusted in combination with environmental impact factors to determine the final risk level. Step S37: Update the dynamic change of the center of gravity based on real-time sensor data and terrain complexity to obtain the updated center of gravity offset feature value; Step S38: If the updated centroid offset feature value exceeds the preset first safety threshold, then re-input it into the support vector machine classifier to obtain a new initial risk assessment result. Step S39: Extract a new tilt angle from the new initial risk assessment results; Step S310: Fit the new tilt angle and terrain height data to determine the new correlation coefficient; Step S311: Calculate the new rollover risk index value based on the new correlation coefficient and vehicle stability parameters. Step S312: If the new rollover risk index value is higher than the preset risk threshold, then the new rollover risk level is adjusted in combination with environmental impact factors to determine the continuous risk level.

[0011] Optionally, step S35, calculating the rollover risk index value based on the correlation coefficient and vehicle stability parameters, includes: The rollover risk index is calculated using the following formula: Rollover risk index = Pearson correlation coefficient × (1 - vehicle stability parameter), The Pearson correlation coefficient represents the linear correlation between tilt and terrain, and the vehicle stability parameter is a preset range of 0 to 1 based on vehicle load.

[0012] Optionally, in step S36, if the rollover risk index value is higher than a preset risk threshold, the rollover risk level is adjusted in conjunction with environmental impact factors to determine the final risk level, including: The final risk level is calculated using the following formula: Final risk level = Initial risk level + Environmental impact factor × 0.5 The initial risk level is extracted from the initial risk assessment, and the environmental impact factor is a preset value based on wind speed and road surface moisture.

[0013] Optionally, step S4, based on the rollover risk level, optimizes the pressure value and displacement of the hydraulic lateral displacement device to generate a hydraulic control command set, including: Step S41: Acquire real-time data from the sensor, process the real-time data using a Kalman filter, and obtain vehicle operating condition parameters; Step S42: Calculate the rollover risk level based on the vehicle operating condition parameters; Step S43: Construct an objective function based on the rollover risk level, initial pressure value, and initial displacement. Step S44: Iteratively optimize the initial pressure value and initial displacement based on the objective function to obtain the optimal pressure value and optimal displacement. Step S45: Generate a hydraulic control command set using the optimal pressure value and the optimal displacement. Step S46: If the control accuracy of the hydraulic control instruction set is lower than the preset control accuracy threshold, then adjust the objective function and recalculate the new initial pressure value and the new initial displacement. Step S47: Obtain real-time data from the sensor to verify the execution efficiency of the hydraulic control instruction set; Step S48: Generate a new hydraulic control instruction set based on instruction execution efficiency.

[0014] Optionally, step S42, calculating the rollover risk level based on vehicle operating condition parameters, includes: The rollover risk level is calculated using the following formula: R = k1v + k2a + k3θ Where R is the rollover risk level, v is the speed, a is the acceleration, θ is the tilt angle, and k1, k2, and k3 are preset coefficients.

[0015] Optionally, step S43, based on the rollover risk level, the initial pressure value, and the initial displacement, constructs an objective function, including: The objective function is: ,in, Let P be the objective function, D be the initial pressure value, and P0 and D0 be the initial displacement values. , , As weight.

[0016] A second aspect of the present invention provides a fire truck rollover risk prevention and control system, which uses the method described above to prevent and control the risk of fire truck rollover, the system comprising: The sensor data preprocessing module is used to acquire tilt angle, tire pressure, terrain height and obstacle distance data from multi-source sensors in real time, and use filtering and standardization algorithms to remove noise and generate a standardized sensor dataset. The center of gravity offset calculation module is used to integrate tilt angle and tire pressure data based on a standardized sensor dataset using a weighted fusion algorithm to calculate the real-time center of gravity offset of the vehicle and generate center of gravity offset feature values. The rollover risk assessment module is used to compare the center of gravity offset characteristic value with a preset first safety threshold. When the center of gravity offset characteristic value exceeds the preset first safety threshold, the rollover risk assessment model analyzes the correlation between the tilt angle and the terrain height to generate a rollover risk level. The hydraulic control optimization module is used to optimize the pressure value and displacement of the hydraulic lateral displacement device based on the rollover risk level using a dynamic programming algorithm, and to generate a hydraulic control instruction set. The smooth movement control module is used to extract target pressure and displacement values ​​from the hydraulic control command set, combine tire pressure data from the standardized sensor dataset, calculate movement speed and frequency parameters, and generate a smooth movement control sequence. The environmental perception modeling module is used to generate a dynamic environmental adaptation model by analyzing obstacle distribution, terrain height change trends and center of gravity offset characteristics through environmental perception algorithms when the obstacle distance is less than a preset distance threshold, based on the obstacle distance data in the stable motion control sequence and the obstacle distance data in the standardized sensor dataset. The path planning generation module is used to generate a lateral displacement path plan based on the dynamic environment adaptation model, by integrating the center of gravity offset feature value and the steady movement control sequence, and obtain a safe displacement trajectory containing path point data. The displacement execution module is used to extract path point data from the safe displacement trajectory, transmit it to the hydraulic lateral displacement device in real time, perform lateral displacement operations, and generate a vehicle dynamic status dataset. The data loop update module is used to update the standardized sensor dataset based on the vehicle dynamic status dataset, and loop through the process from generating center of gravity offset feature values ​​to generating safe displacement trajectory, continuously optimizing displacement control parameters.

[0017] The technical solution provided by this invention has the following beneficial effects: This invention discloses a method and system for preventing vehicle rollover risks. It utilizes multi-source sensors to collect real-time data on vehicle tilt angle, tire pressure, terrain elevation, and obstacle distance, generating a sensor dataset after filtering and standardization. Based on this dataset, the invention employs a weighted fusion algorithm to calculate the vehicle's center of gravity offset and combines this with a preset safety threshold to assess the rollover risk level. According to the risk level, the invention uses a dynamic programming algorithm to optimize the control parameters of the hydraulic lateral displacement device, generating a smooth movement control sequence. Simultaneously, the invention also incorporates an environmental perception algorithm to dynamically adjust the displacement path when an obstacle is detected.

[0018] Ultimately, this invention, through continuous iterative optimization, controls the hydraulic lateral displacement device to perform lateral displacement operations in real time, effectively preventing vehicle rollover risks and improving driving safety. Attached Figure Description

[0019] Figure 1 This is a flowchart of a method for preventing the rollover risk of fire trucks according to the present invention.

[0020] Figure 2 This is a schematic diagram of a fire truck rollover risk prevention method according to the present invention.

[0021] Figure 3 This is another schematic diagram of a fire truck rollover risk prevention method according to the present invention.

[0022] Figure 4 This is a structural schematic diagram of a fire truck rollover risk prevention and control system according to the present invention. Detailed Implementation

[0023] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] like Figure 1-3 As shown, in a first aspect, the present invention provides a method for preventing the rollover risk of fire trucks, which may specifically include: S1 acquires tilt angle, tire pressure, terrain height, and obstacle distance data in real time from multiple source sensors, and uses filtering and standardization algorithms to remove noise and generate a standardized sensor dataset.

[0025] Optionally, this step also includes: Step S11: Collect tilt angle, tire pressure, terrain height, and obstacle distance data from multi-source sensors in real time to generate the original sensor dataset.

[0026] Step S12: The original sensor dataset is processed using a Kalman filter to remove noise and obtain a clean sensor dataset.

[0027] Step S13: Use the min-max normalization tool to process the clean sensor dataset to generate a normalized sensor dataset.

[0028] Step S14: Extract tilt angle and tire pressure data from the standardized sensor dataset. If at least one of them exceeds a preset parameter threshold, then fuse terrain height and obstacle distance data using a weighted average tool. Specifically, first determine the weights as follows: terrain height weight 0.6 and obstacle distance weight 0.4, then calculate the weighted sum to obtain comprehensive environmental feature data.

[0029] Step S15: Obtain the time series of each data point from the comprehensive environmental feature data, calculate the difference between adjacent points as the trend of change, and generate the environmental dynamic feature vector.

[0030] Step S16: Use principal component analysis to reduce the dimensionality of the environmental dynamic feature vectors to obtain a low-dimensional feature dataset.

[0031] Step S17: Calculate the variance from the low-dimensional feature dataset. If it is lower than the preset variance threshold, use the K-means tool to cluster the low-dimensional feature dataset to generate classification environment data.

[0032] For example, in the sensor data processing scenarios of autonomous vehicles, multi-source sensor data needs to be collected in real time to support navigation decisions. Tire angle sensors monitor vehicle attitude, tire pressure sensors reflect tire condition, terrain height sensors provide information on road surface undulations, and obstacle distance sensors detect the distance to objects ahead. Assuming an autonomous vehicle is driving on mountainous terrain with a tilt angle of 15 degrees, tire pressure of 2.5 bar, terrain height change of 0.8 meters, and obstacle distance of 5 meters, this constitutes the initial sensor dataset.

[0033] In one possible implementation, Kalman filtering is used to remove noise. Kalman filtering smooths out data fluctuations by fusing sensor measurements with the system model through prediction and update steps.

[0034] For example, the tilt angle might be noisy due to vibration; after filtering, a stable value of 14.8 degrees is obtained. Tire pressure, terrain height, and obstacle distance are processed in the same way to generate a clean dataset. This method effectively reduces noise interference and improves data reliability.

[0035] Specifically, min-max normalization maps the clean dataset to the [0,1] interval. Assuming tire pressure ranges from 2.0 to 3.0 bar, 2.5 bar is normalized to 0.5; and tilt angle ranges from 0 to 30 degrees, 14.8 degrees is normalized to 0.493. The normalized dataset facilitates subsequent analysis, eliminates dimensional differences, and improves model compatibility.

[0036] For example, if the preset parameter thresholds include a tilt angle threshold of 0.5 or a tire pressure threshold of 0.6, the current data triggers the fusion condition. The weighted average tool calculates the comprehensive environmental features with a terrain height weight of 0.6 and an obstacle distance weight of 0.4. Assuming a standardized terrain height of 0.8 and an obstacle distance of 0.2, the weighted sum is 0.8 × 0.6 + 0.2 × 0.4 = 0.56. This fusion emphasizes the impact of terrain on driving while also considering obstacle information, reflecting environmental complexity.

[0037] In one possible implementation, the comprehensive environmental feature data is arranged in a time series, and the changing trend is calculated by the difference between adjacent points. Assuming the comprehensive features at three consecutive time points are 0.56, 0.58, and 0.60, with differences of 0.02 and 0.02, a dynamic environmental feature vector is formed. This vector captures the rate of environmental change and helps predict potential risks.

[0038] Specifically, principal component analysis reduces the dimensionality of dynamic eigenvectors while retaining key information.

[0039] For example, if the original vector contains multi-dimensional features, dimensionality reduction yields a 2D dataset, reducing computation while preserving 90% of the variance. This method reduces processing complexity and is suitable for real-time applications.

[0040] For example, if the variance of a low-dimensional dataset is below a preset variance threshold of 0.1, it indicates that the data variation is small, making K-means clustering suitable. Assume the clusters are divided into two classes: stationary and complex. The stationary class represents flat roads, and the complex class represents rugged terrain. Classifying environmental data guides vehicles to adjust speed or route planning, improving safety.

[0041] In one possible implementation, the above process ensures end-to-end optimization of data from collection to classification. Kalman filtering improves data accuracy, standardization unifies the scale, feature fusion and dimensionality reduction reduce computational burden, and clustering provides a basis for environmental classification. This approach supports efficient decision-making in autonomous driving systems, enhancing driving stability and safety.

[0042] S2, based on a standardized sensor dataset, uses a weighted fusion algorithm to integrate tilt angle and tire pressure data, calculates the real-time center of gravity offset of the vehicle, and generates center of gravity offset feature values.

[0043] Optionally, this step also includes: Step S21: Obtain standardized tilt angle and tire pressure data from the sensor dataset, remove noise through median filtering, and obtain the cleaned dataset.

[0044] Step S22: Based on the cleaned dataset, calculate the weighted average of the tilt angle and tire pressure data to generate a preliminary estimate of the center of gravity offset.

[0045] Step S23: If the initial estimate of the center of gravity offset exceeds the preset second safety threshold, the center of gravity offset estimate is smoothed by Kalman filtering to obtain the smoothed center of gravity offset value.

[0046] Step S24: For the smoothed centroid offset value, calculate the covariance matrix and extract the first two largest eigenvalues ​​to generate the centroid offset eigenvector.

[0047] Step S25: Obtain the offset direction and magnitude information from the center of gravity offset feature vector, and calculate the real-time center of gravity offset feature value through a linear mapping function.

[0048] Preferably, the following linear mapping function is used to calculate the real-time centroid offset feature value: f(x) = gx + b, where f(x) is the real-time centroid offset eigenvalue function, x is the amplitude information, g is the preset weight, and b is the offset constant.

[0049] Step S26: Based on the real-time center of gravity offset feature value, a decision tree is used to determine the vehicle's center of gravity offset state and obtain the vehicle stability classification result.

[0050] Step S27: Generate output data based on the vehicle stability classification results.

[0051] For example, when acquiring standardized tilt angle and tire pressure data from sensor datasets, multi-source sensors can be used to collect vehicle tilt angle and tire pressure data in real time under different terrain conditions. Assuming a vehicle is traveling on a mountain road with varying slopes, tilt angle sensors record the vehicle's angle relative to the horizontal plane, such as 10 degrees or 12 degrees, while tire pressure sensors record the pressure values ​​of the four tires, such as 2.5 bar or 2.4 bar. After standardization, the tilt angle may be mapped to a range of 0 to 1, and the tire pressure is converted into a ratio relative to the standard value, such as 0.8 or 0.75. This data provides a uniform scale for subsequent analysis.

[0052] In one possible implementation, when removing noise through median filtering, the median of five consecutive time points can be taken from the standardized data to replace outliers.

[0053] For example, given a tilt angle sequence of 0.8, 0.9, 2.0, 0.85, and 0.82, where 2.0 is clearly abnormal, replacing it with the median value of 0.85 yields a smoothed sequence. This method effectively preserves the data trend while eliminating abrupt changes caused by sensor jitter or external interference.

[0054] Specifically, when calculating the weighted average to generate a preliminary estimate of the center of gravity offset, the tilt angle weight can be set to 0.7 and the tire pressure weight to 0.3.

[0055] For example, with a standardized tilt angle of 0.8 and tire pressure of 0.75, the weighted average is 0.8 × 0.7 + 0.75 × 0.3 = 0.785. If the preset second safety threshold is 0.6, 0.785 exceeds the second safety threshold, triggering subsequent smoothing processing. This weighted design takes into account the greater impact of tilt angle on center of gravity shift.

[0056] For example, when smoothing the estimated center of gravity shift using Kalman filtering, an initial estimate of 0.785 can be used as input. Combining this with the vehicle motion model and noise covariance, a smoothed value such as 0.78 can be output. The smoothed value better reflects the trend of center of gravity changes in vehicle stability. Next, when calculating the covariance matrix and extracting the first two largest eigenvalues, a matrix can be generated from the time series of the smoothed values. The eigenvalues ​​reflect the main direction and magnitude of the shift.

[0057] For example, eigenvalues ​​1.2 and 0.8 might represent a primary offset along the vertical axis and a secondary offset along the horizontal axis.

[0058] In one possible implementation, after generating the center-of-gravity offset feature vector, real-time feature values ​​are calculated using a linear mapping function. Assuming the function is f(x) = 0.5x + 0.1, and the magnitude x is 1.2, then the real-time feature value is 0.5 × 1.2 + 0.1 = 0.7. This mapping converts the offset magnitude into a quantifiable stability index. When the decision tree judges vehicle stability, it can use the feature value of 0.7, combined with preset rules, such as a feature value greater than 0.65 indicating an unstable state, and output a classification result of "needs adjustment". Ultimately, the generated output data can include the classification result and feature values, used by the vehicle control system to optimize driving strategies.

[0059] Specifically, this method employs multi-level processing to form a complete analysis chain from raw data to classification results. Each step fully utilizes sensor data to progressively extract key features, ensuring the accuracy and real-time nature of stability assessments. This method is particularly effective in supporting dynamic vehicle adjustments in complex terrain.

[0060] S3. Based on the comparison between the center of gravity offset characteristic value and the preset first safety threshold, when the center of gravity offset characteristic value exceeds the preset first safety threshold, the correlation between the tilt angle and the terrain height is analyzed through the rollover risk assessment model to generate a rollover risk level.

[0061] Optionally, this step also includes: Step S31: Obtain real-time vehicle center of gravity offset feature value and terrain height data through sensors.

[0062] Step S32: If the real-time centroid offset feature value exceeds the preset first safety threshold, then input it into the support vector machine classifier to obtain the initial risk assessment result.

[0063] Step S33: Extract the tilt angle from the initial risk assessment results.

[0064] Step S34: Use a linear regression model to fit the tilt angle and terrain height data to determine the Pearson correlation coefficient.

[0065] Step S35: Calculate the rollover risk index value based on the Pearson correlation coefficient and vehicle stability parameters.

[0066] Preferably, the rollover risk index value is calculated according to the following formula: Rollover risk index = Pearson correlation coefficient × (1 - vehicle stability parameter), The Pearson correlation coefficient represents the linear correlation between tilt and terrain, and the vehicle stability parameter is a preset range of 0 to 1 based on vehicle load.

[0067] Step S36: If the rollover risk index value is higher than the preset risk threshold, the rollover risk level is adjusted in combination with environmental impact factors to determine the final risk level.

[0068] Preferably, the final risk level is calculated using the following formula: Final risk level = Initial risk level + Environmental impact factor × 0.5 The initial risk level is extracted from the initial risk assessment, and the environmental impact factor is a preset value based on wind speed and road surface moisture.

[0069] Step S37: Update the dynamic change of the center of gravity using real-time sensor data and terrain complexity to obtain the updated center of gravity offset feature value.

[0070] Step S38: If the updated centroid offset feature value exceeds the preset first safety threshold, then re-input it into the support vector machine classifier to obtain a new initial risk assessment result.

[0071] Step S39: Extract the new tilt angle from the new initial risk assessment results.

[0072] Step S310: Use a linear regression model to fit the new tilt angle and terrain height data to determine the new Pearson correlation coefficient.

[0073] Step S311: Calculate the new rollover risk index value based on the new Pearson correlation coefficient and vehicle stability parameters.

[0074] Step S312: If the new rollover risk index value is higher than the preset risk threshold, then the new rollover risk level is adjusted in combination with environmental impact factors to determine the continuous risk level.

[0075] For example, in a real-time vehicle center of gravity shift monitoring scenario, the process of sensors acquiring real-time center of gravity shift characteristic values ​​and terrain height data can be achieved based on high-precision accelerometers and laser rangefinders. The accelerometer measures the vehicle's acceleration in three axes in real time, and combined with the vehicle's mass distribution model, calculates the real-time center of gravity shift characteristic value. The laser rangefinder, on the other hand, acquires terrain height data by scanning the ground.

[0076] For example, when a heavy truck is driving in a mountainous area, the sensor records a center of gravity offset feature value of 0.45 and a terrain height change of 2 meters.

[0077] It should be noted that the real-time center of gravity offset characteristic value, after being standardized by accelerometer data, reflects the degree of vehicle tilt relative to the horizontal plane, while the terrain height data provides a basis for subsequent correlation analysis.

[0078] In one possible implementation, if the real-time center of gravity offset feature value of 0.45 exceeds a preset first safety threshold of 0.3, the data is input into a support vector machine classifier. The classifier, trained on historical data, distinguishes between normal and abnormal states and outputs an initial risk assessment result, such as a tilt angle of 15 degrees. A linear regression model then fits the 15-degree tilt angle with 2-meter terrain height data, calculating the Pearson correlation coefficient. Assuming a coefficient of 0.8, this indicates a strong correlation between the tilt angle and terrain height. The vehicle stability parameter is set to 0.6 based on the load weight, and the rollover risk index is 0.8 × (1 - 0.6) = 0.32. If the preset risk threshold is 0.25, then 0.32 is higher than the risk threshold, indicating a rollover risk.

[0079] For example, the environmental impact factor can be adjusted based on wind speed and road surface moisture. Assuming a wind speed of 10 m / s and road surface moisture of 80%, the preset environmental impact factor is 0.5. The initial risk level is "medium," and the final risk level after adjustment is "medium + 0.5 = medium-high." This adjustment considers the comprehensive impact of the external environment on vehicle stability.

[0080] It should be noted that wind speed and humidity are acquired in real time through meteorological sensors to ensure the dynamic nature of the risk assessment.

[0081] In one possible implementation, when updating the dynamic changes in the center of gravity, the sensors detect an increase in terrain complexity, such as a slope changing from 10 degrees to 20 degrees. The updated center of gravity offset eigenvalue is 0.48, exceeding the first safety threshold of 0.3. It is then re-input into a support vector machine to obtain a new tilt angle of 18 degrees. A linear regression model is fitted to 18 degrees and the new terrain height of 3 meters, calculating a new Pearson correlation coefficient of 0.85. The new rollover risk index is 0.85 × (1 - 0.6) = 0.34, higher than the risk threshold of 0.25. Combining this with an environmental impact factor of 0.5, the adjusted risk level is "high".

[0082] Understandably, continuous monitoring and updates ensure the real-time nature of risk assessments and adapt to complex terrain changes.

[0083] For example, the assessment of terrain complexity can be based on continuous scanning data from a laser rangefinder, combined with a map database, to determine the frequency and magnitude of slope changes. This approach improves the accuracy of center of gravity shift monitoring through multi-dimensional data fusion.

[0084] It should be noted that the real-time update mechanism allows the system to respond quickly to changes in terrain, dynamically adjust the risk level, and provide timely warnings to drivers.

[0085] S4. Based on the rollover risk level, a dynamic programming algorithm is used to optimize the pressure and displacement of the hydraulic lateral displacement device and generate a hydraulic control command set.

[0086] Optionally, this step also includes: Step S41: Acquire real-time data from the sensor, process the real-time data using a Kalman filter, and obtain the vehicle operating condition parameters.

[0087] Step S42: Calculate the rollover risk level R based on the vehicle operating parameters. Preferably, the rollover risk level is calculated according to the following formula: R = k1v + k2a + k3θ Where R is the rollover risk level, v is the speed, a is the acceleration, θ is the tilt angle, and k1, k2, and k3 are preset coefficients.

[0088] Step S43: Construct the objective function J based on the rollover risk level R, the initial pressure value P, and the initial displacement D.

[0089] Preferably, the objective function for: ,in, Let P be the objective function, D be the initial pressure value, and P0 and D0 be the initial displacement values. , , For weights.

[0090] Step S44, according to the objective function The initial pressure value P and the initial displacement D are optimized using the Bellman equation iteratively to obtain the optimal pressure value P_opt and the optimal displacement D_opt.

[0091] Step S45: Generate a hydraulic control command set using the optimal pressure value P_opt and the optimal displacement D_opt.

[0092] Step S46: If the control accuracy of the hydraulic control command set is lower than the preset control accuracy threshold, then adjust the weights of the objective function. , , The new initial pressure value P and the new initial displacement D are then recalculated.

[0093] Step S47: Obtain real-time data from the sensor to verify the execution efficiency of the hydraulic control instruction set.

[0094] Step S48: Generate a new hydraulic control instruction set based on instruction execution efficiency.

[0095] For example, the process of acquiring real-time data from sensors and processing it using a Kalman filter to obtain vehicle operating parameters can be understood as acquiring vehicle state information, such as speed, acceleration, and tilt angle, in real time through multiple source sensors. The Kalman filter, through prediction and update steps, fuses sensor data, reduces noise interference, and generates more accurate operating parameters.

[0096] For example, assuming a vehicle is traveling on complex terrain, and the sensors collect data showing a speed of 20 m / s, an acceleration of 2 m / s², and a tilt angle of 15°, a Kalman filter will smooth these data, eliminating abnormal fluctuations caused by road bumps, and ultimately output stable operating parameters. This method effectively improves data reliability and provides accurate input for subsequent risk assessment.

[0097] In one possible implementation, the rollover risk level R can be calculated based on operating condition parameters using the formula R=k1v+k2a+k3θ.

[0098] For example, by setting k1=0.5, k2=0.3, and k3=0.2, and combining this with the aforementioned speed of 20 m / s, acceleration of 2 m / s², and tilt angle of 15°, the rollover risk level R value can be obtained. This method comprehensively reflects the rollover risk of the vehicle under the current condition by weighting different parameters.

[0099] Preferably, the coefficients k1, k2, and k3 can be dynamically adjusted according to vehicle type and terrain features. For example, when a heavy truck is driving on a slope, the weight of k3 can be increased to highlight the impact of the tilt angle on risk. This flexibility ensures that the assessment results are more in line with actual scenarios.

[0100] Specifically, when constructing the objective function J, the deviations between the rollover risk R, the initial pressure value P, and the initial displacement D and the reference value need to be comprehensively considered.

[0101] For example, assuming the initial pressure value P is 500 kPa, the reference value P0 is 480 kPa, the initial displacement D is 0.1 m, the reference value D0 is 0.08 m, and the weights w1, w2, and w3 are 0.4, 0.3, and 0.3 respectively, the objective function J can reflect the comprehensive performance of the system control.

[0102] Preferably, the weights can be adjusted according to operating conditions, such as increasing w2 on wet or slippery surfaces to emphasize pressure stability. This design optimizes the control precision of the hydraulic system.

[0103] For example, when using the Bellman equation to iteratively optimize the initial pressure value P and the initial displacement D, the optimal solution can be gradually approximated through dynamic programming.

[0104] For example, the initial iteration starts with P=500kPa and D=0.1m, and through multiple iterations, the optimal pressure value P_opt and displacement D_opt are obtained. This method can effectively balance the system's response speed and stability, ensuring the high efficiency of the hydraulic control command set.

[0105] In one possible implementation, after generating the hydraulic control instruction set, if the control accuracy is lower than the control accuracy threshold, the weights w1, w2, and w3 can be readjusted.

[0106] For example, if the vehicle tilt angle does not significantly improve after the command is executed, the weight of w3 can be increased, and P and D can be recalculated. This closed-loop adjustment mechanism can quickly adapt to dynamically changing operating conditions.

[0107] Specifically, when verifying the execution efficiency of the hydraulic control command set, the vehicle response can be monitored in real time using sensors.

[0108] For example, if the tilt angle drops from 15° to 10° after the command is executed, it indicates that the command is effective.

[0109] Preferably, a more accurate instruction set can be generated through multiple iterations of verification. This continuous optimization approach can significantly improve system robustness.

[0110] S5 extracts the target pressure value and displacement from the hydraulic control command set, combines the tire pressure data in the standardized sensor dataset, calculates the moving speed and frequency parameters, and generates a smooth moving control sequence.

[0111] Optionally, this step also includes: Step S51: Obtain the target pressure value and displacement from the hydraulic control command set, and extract the numerical parameters from the command set using regular expressions to obtain the target pressure value and displacement.

[0112] Step S52: Based on the target pressure value and displacement, obtain tire pressure data from the standardized sensor dataset, integrate the target pressure value, displacement and tire pressure data using Kalman filtering, first predict the fusion state and then update the measurement value to obtain the fusion data features.

[0113] Step S53: Based on the features of the fused data, a linear regression algorithm is used to calculate the moving speed. First, the relationship between the features and the speed is fitted, and then the value is predicted to obtain the moving speed parameter.

[0114] Step S54: Based on the moving speed parameters, analyze the periodicity of the fused data features using Fast Fourier Transform, apply Discrete Fourier Transform to the features to calculate the amplitude spectrum, extract the dominant frequency, and obtain the frequency parameters.

[0115] Step S55: If the frequency parameter exceeds the preset frequency threshold, the frequency parameter is adjusted by moving average filtering. First, the window average is calculated and then the value is replaced to obtain the smoothed frequency parameter.

[0116] Step S56: Generate a control sequence based on the smoothing frequency parameter and the moving speed parameter. Construct a smooth moving control sequence using a proportional-integral-derivative controller. Calculate the error first and then adjust the output to obtain the control sequence.

[0117] For example, in a vehicle hydraulic control system, when extracting target pressure and displacement values ​​from a hydraulic control command set, regular expressions can be used to match specific numerical formats. Assuming the command set is in string form, such as "PRESSURE_150.5kPa_DISPLACEMENT_20.3mm", the regular expression can be designed to match the patterns "PRESSURE_" followed by a floating-point number and "DISPLACEMENT_" followed by a floating-point number, extracting 150.5 and 20.3. This method ensures efficient parsing of complex command sets, avoiding errors caused by manual splitting. Especially in scenarios with high real-time requirements, rapid numerical extraction is crucial for subsequent control.

[0118] In one possible implementation, when acquiring tire pressure data from a standardized sensor dataset, the sensor may output pressure values ​​for multiple tires, such as 200.2 kPa for the left front tire and 198.7 kPa for the right front tire. When Kalman filtering integrates the target pressure value of 150.5 kPa, the displacement of 20.3 mm, and the tire pressure data, it first estimates the vehicle's current motion state through a prediction step, then updates the state by combining the sensor measurements, resulting in fused data features such as the overall pressure distribution and displacement trend. This fusion effectively reduces the impact of sensor noise, ensuring the data more closely reflects actual operating conditions.

[0119] Specifically, when using a linear regression algorithm to calculate movement speed based on the features of fused data, pressure, displacement, and time series in the fused data can be used as input features to fit a linear relationship between speed and these features.

[0120] For example, assuming data shows that increased pressure and displacement typically correspond to increased speed, a model can be trained using historical data to predict the current speed as 60.5 km / h. This method is simple, efficient, and suitable for scenarios with high real-time requirements.

[0121] For example, when analyzing the periodicity of fused data features using Fast Fourier Transform (FFT), Discrete Fourier Transform (DFT) can be applied to time-series data of pressure and displacement to calculate the amplitude spectrum and extract the dominant frequency. Suppose the analysis yields a dominant frequency of 2.5 Hz, which may reflect the vibration period of the vehicle suspension system. If this frequency exceeds a preset threshold of 2.0 Hz, it indicates that the vibration may affect stability and requires further processing. This type of analysis helps identify potential instability factors.

[0122] In one possible implementation, when adjusting the frequency parameter using a moving average filter, assuming a window size of 5, the average frequency over the most recent five time points is calculated, such as 2.5Hz, 2.6Hz, 2.4Hz, 2.7Hz, and 2.3Hz, resulting in a smoothed frequency of 2.5Hz. This method can smooth short-term fluctuations, preserve the main trend, and improve the robustness of the control system.

[0123] Specifically, when generating a control sequence based on a smoothing frequency parameter of 2.5Hz and a moving speed of 60.5km / h, a proportional-integral-derivative (PID) controller can be used. First, the error between the current speed and the target speed is calculated; for example, if the target speed is 60.0km / h, the error is 0.5km / h. Then, a smooth control sequence is generated through a fast proportional term, steady-state error elimination via an integral term, and overshoot suppression via a derivative term. This control method ensures smooth vehicle movement under complex operating conditions, improving safety.

[0124] S6. Based on the smooth movement control sequence and obstacle distance data in the standardized sensor dataset, when the obstacle distance is less than the preset distance threshold, the environmental perception algorithm analyzes the obstacle distribution, terrain height change trend and center of gravity offset characteristic value to generate a dynamic environmental adaptation model.

[0125] Optionally, this step also includes: Step S61: Obtain obstacle distance data and smooth movement control sequence from standardized sensors, use a preset distance threshold to determine whether the obstacle distance is less than the distance threshold, and obtain a trigger signal.

[0126] Step S62: Based on the trigger signal, use OpenCV to process the sensor dataset, extract edge features, analyze obstacle distribution, and determine the spatial location of obstacles.

[0127] Step S63: Using the spatial location of the obstacle and combining it with the terrain height data obtained from the sensor, calculate the height difference trend to obtain the terrain change characteristics.

[0128] Step S64: Calculate the center of gravity offset characteristic value and determine the center of gravity offset state by combining the terrain change characteristics with the steady-state movement control sequence.

[0129] Step S65: Based on the center of gravity shift state, use a linear regression model to generate environmental adaptation parameters.

[0130] Step S66: Adjust the smooth movement control sequence using environmental adaptation parameters to obtain the optimized control sequence.

[0131] Step S67: Optimize the control sequence, update the linear regression model, and generate a real-time environmental adaptation scheme.

[0132] For example, when acquiring obstacle distance data from standardized sensors, environmental information can be collected using lidar or ultrasonic sensors. LiDAR emits laser pulses at a high frequency, receives reflected signals, calculates the distance, and generates an obstacle distance dataset. Assuming an automated guided vehicle (AGV) is operating in a warehouse environment, the sensors collect distance data 100 times per second, obtaining obstacle distance values ​​such as 2.5 meters, 3.0 meters, etc. The smooth movement control sequence is provided by a hydraulic system or motor control module, including speed and direction commands, such as "advance at 0.5 meters per second." A preset distance threshold is used to determine if the obstacle distance is less than a safe distance (e.g., 1.0 meter). When the distance is 0.8 meters, a signal is triggered, prompting the system to enter obstacle avoidance mode.

[0133] Specifically, when using OpenCV to process sensor datasets, LiDAR point cloud data can be image-processed to extract obstacle edge features. OpenCV uses the Canny edge detection algorithm to identify obstacle contours, generate edge point sets, and then analyze obstacle distribution. For example, assuming there's a shelf in a warehouse, edge detection shows its width is 1.2 meters, and it's located 1.0 meter in front of a vehicle, determining the obstacle's spatial location as (1.0, 0, 1.2). This process helps the system accurately locate obstacles and avoid collisions.

[0134] In one embodiment, terrain height data acquired by sensors (such as ground height changes measured by an ultrasonic height sensor) is used to calculate the height difference trend. Assuming the vehicle is moving and the terrain height changes from 0.1 meters to 0.3 meters, the height difference is 0.2 meters, and the trend shows the ground gradually rising. The center of gravity offset characteristic value is then calculated using a smooth movement control sequence.

[0135] For example, when a vehicle is driving on a slope, the increased terrain causes the center of gravity to tilt forward, and the offset characteristic value may be 0.15 meters, indicating that the center of gravity offset state is "slightly tilted forward".

[0136] For example, based on the center of gravity offset state, a linear regression model is used to generate environmental adaptation parameters. The model is trained on historical data, taking the terrain elevation difference and center of gravity offset value as input, and outputting adaptation parameters such as "decelerate by 0.2 m / s". Assuming the current offset state is "slightly leaning forward", the model predicts a 10% speed reduction and generates adaptation parameters. The adjusted smooth movement control sequence becomes "advance at 0.3 m / s", forming an optimized control sequence. This sequence can adapt to terrain changes and maintain vehicle stability.

[0137] Specifically, when updating the linear regression model using optimized control sequences, real-time collected terrain data and centroid offset data can be used as new inputs to continuously optimize the model parameters.

[0138] For example, if a vehicle runs continuously on a warehouse ramp for 10 minutes, the model updates its weights based on the data from each run and generates a real-time environmental adaptation plan, such as "reducing the speed to 0.25 m / s and adjusting the direction by 5 degrees." This dynamic update makes the system more adaptable to complex environments.

[0139] In one embodiment, the application of optimized control sequences can significantly improve vehicle stability on uneven terrain.

[0140] For example, if the warehouse floor has slight undulations, the optimized sequence can prevent vehicles from overturning due to a shift in the center of gravity by slowing down and making minor adjustments to direction. The real-time environmental adaptation solution can also reduce downtime caused by sensor misjudgments, ensuring continuous operation.

[0141] S7. Based on the dynamic environment adaptation model, the center of gravity offset characteristic value and the steady movement control sequence are integrated to generate the lateral displacement path planning and obtain the safe displacement trajectory containing path point data.

[0142] Optionally, this step also includes: step S71, acquiring environmental perception data from the sensor, fusing real-time environmental information, and using a Kalman filter to determine a dynamic environment model.

[0143] Step S72: Calculate the object pose data based on the dynamic environment model to obtain the center of gravity offset feature value.

[0144] Step S73: If the center of gravity offset characteristic value exceeds the preset third safety threshold, the PID controller is used to adjust the control parameters and generate a smooth movement control sequence.

[0145] Step S74: A steady-state motion control sequence is input into Algorithm A and combined with the dynamic environment model to generate a lateral displacement path.

[0146] Step S75: Using the lateral displacement path, the gradient descent algorithm is used to optimize the path point data to obtain the safe displacement trajectory.

[0147] Step S76: If there is a deviation between the safe displacement trajectory and the environmental perception data, the trajectory points are iteratively adjusted to obtain optimized trajectory point data.

[0148] Step S77: Generate the final safe displacement trajectory based on the optimized trajectory point data.

[0149] For example, when acquiring environmental perception data from sensors, distance and position information of surrounding obstacles can be collected using LiDAR and ultrasonic sensors, and combined with an IMU (Inertial Measurement Unit) to obtain terrain tilt angle and acceleration data. The LiDAR collects 1000 point cloud data points per second, the ultrasonic sensors provide obstacle distances within a 10-meter range, and the IMU provides tilt angles with 0.1-degree accuracy. This data fusion forms real-time environmental information, reflecting the surrounding environment of the robot or autonomous vehicle.

[0150] In one possible implementation, a Kalman filter is used to process this data to determine a dynamic environment model. The Kalman filter reduces the impact of noise by fusing multi-source sensor data through prediction and update steps.

[0151] For example, lidar may experience distance deviations due to light reflection, while IMU data may drift due to vibration. Kalman filtering can optimize data through weighted averaging, generating a smooth dynamic environment model that describes real-time changes in obstacles and terrain.

[0152] For example, when calculating object pose data based on a dynamic environment model, the three-dimensional coordinates and orientation of obstacles can be determined through geometric analysis of point cloud data.

[0153] For example, if the point cloud shows that there is a 0.5-meter-high obstacle 2 meters ahead, and the terrain is determined to be an uphill slope of 5 degrees based on the IMU data, the calculated characteristic value of the machine's center of gravity offset is 0.3 meters, which exceeds the preset third safety threshold of 0.2 meters, triggering the adjustment requirement.

[0154] In one possible implementation, a PID controller is used to adjust control parameters to generate a smooth movement control sequence. The PID controller adjusts the motor speed and direction based on the center of gravity offset.

[0155] For example, when the center of gravity shifts by 0.3 meters, the PID controller increases the speed of the left wheel by 10% and decreases the speed of the right wheel by 5% to restore the machine's balance and generate a smooth movement sequence.

[0156] For example, when Algorithm A generates a lateral displacement path using a dynamic environment model, it can analyze obstacle distribution and plan a curved path that avoids obstacles 2 meters away. Path points are marked every 0.5 meters to ensure the robot passes safely at a speed of 0.5 meters per second.

[0157] In one possible implementation, a gradient descent algorithm is used to optimize the path point data and generate a safe displacement trajectory.

[0158] For example, the initial waypoints may deviate due to sudden changes in terrain. Gradient descent can shorten the total path length by 5% and improve movement efficiency by iteratively adjusting the coordinates of the waypoints.

[0159] For example, if there is a deviation between the safe displacement trajectory and the environmental perception data, such as the path point deviating from the actual terrain by 0.2 meters, the trajectory point can be adjusted iteratively.

[0160] For example, recalculate the path point coordinates, combine them with the latest point cloud data, ensure that the trajectory points maintain a safe distance of 1 meter from obstacles, and generate optimized trajectory point data.

[0161] In one possible implementation, the final safe displacement trajectory is generated based on the optimized trajectory point data.

[0162] For example, the trajectory point data is updated every 0.1 seconds to form a continuous trajectory curve, ensuring that the robot moves smoothly at a speed of 0.4 meters per second, avoiding obstacles and adapting to changes in terrain.

[0163] S8 extracts path point data from the safe displacement trajectory and transmits it to the hydraulic lateral displacement device in real time to perform lateral displacement operations and generate a vehicle dynamic status dataset.

[0164] Optionally, this step also includes: Step S81: Obtain initial safe displacement trajectory data from vehicle sensors, and filter the initial safe displacement trajectory data using a Savitzky-Golay filter to obtain the first smooth path point data.

[0165] Step S82: If the deviation between the first smooth path point data and the preset trajectory threshold exceeds the limit, then the missing points are supplemented by linear interpolation to generate the first complete path point data.

[0166] Step S83: Based on the first complete path point data, convert it into control commands using the CAN bus protocol to obtain a first command sequence that the hydraulic lateral displacement device can recognize.

[0167] Step S84: Transmit the first instruction sequence to the hydraulic lateral displacement device via the MQTT protocol to perform lateral displacement operation and generate the first displacement execution record.

[0168] Step S85: Extract the vehicle lateral displacement parameters from the first displacement execution record, and combine them with sensor data to generate the first vehicle dynamic state dataset.

[0169] Step S86: If the key parameters in the first vehicle dynamic state dataset deviate from the preset normal range, the Kalman filter algorithm is used to optimize the data to obtain the second vehicle dynamic state dataset.

[0170] Step S87: Adjust the preset trajectory threshold according to the second vehicle dynamic state dataset to generate the second complete path point data.

[0171] For example, when acquiring initial safe displacement trajectory data from vehicle sensors, environmental information, such as obstacle distances and road conditions, can be collected using lidar and ultrasonic sensors. Suppose an automated guided vehicle (AGV) is driving in a warehouse; lidar detects a shelf 2 meters ahead, and ultrasonic sensors confirm no obstacles within 1.5 meters to the side. This data forms the initial trajectory points, containing position coordinates and velocity information. Using a Savitzky-Golay filter to smooth this data effectively reduces sensor noise interference.

[0172] For example, the filter fits with a 5-point window and a second-order polynomial, resulting in more continuous first-smooth path point data, which is suitable for subsequent path planning.

[0173] In one possible implementation, if the deviation between the first smooth path point data and a preset trajectory threshold exceeds a certain range—for example, the distance between trajectory points is greater than 0.2 meters or the angle deviation exceeds 5 degrees—missing points can be supplemented through linear interpolation. Assuming a path segment lacks intermediate points, the interpolation method calculates the intermediate point positions based on the coordinates of preceding and following points, generating the first complete path point data. This method ensures path continuity and facilitates the generation of control commands.

[0174] Specifically, the first complete path point data is converted into control commands via the CAN bus protocol.

[0175] For example, pathpoint data includes the target position and speed. The CAN bus transmits commands to the hydraulic lateral displacement device at a rate of 500kbps. The commands include a movement distance of 0.3 meters and a speed of 0.1 meters per second. The hydraulic lateral displacement device then executes the precise displacement. This method ensures the real-time performance and reliability of command transmission.

[0176] For example, transmitting the first instruction sequence to the hydraulic lateral displacement device via the MQTT protocol enables low-latency communication. Assuming a warehouse environment, MQTT transmits at QoS 1 level to ensure that instructions are delivered at least once. After execution, a first displacement execution record is generated, recording the actual distance and time the device moves, such as moving 0.29 meters in 3 seconds. These records provide a basis for subsequent analysis.

[0177] In one possible implementation, lateral displacement parameters, such as an offset angle of 3 degrees and an actual displacement distance of 0.29 meters, are extracted from the first displacement execution record. This data is then combined with sensor data to generate a first vehicle dynamic state dataset. If key parameters deviate from the normal range, such as an offset angle exceeding 2 degrees, a Kalman filter algorithm can be used to optimize the data. The Kalman filter, through prediction and update steps, fuses sensor data and historical states to generate a second vehicle dynamic state dataset, reducing the angle deviation to within 1 degree. This method improves data accuracy.

[0178] Specifically, the preset trajectory threshold is adjusted based on the second vehicle dynamic state dataset.

[0179] For example, by tightening the angle deviation threshold from 2 degrees to 1.5 degrees, a second complete waypoint data is generated. This adjustment makes the path more closely match the actual environment, improving the adaptability of navigation.

[0180] S9 updates the standardized sensor dataset based on the vehicle dynamic state dataset, and iteratively executes the process from generating center of gravity offset feature values ​​to generating safe displacement trajectory, continuously optimizing displacement control parameters.

[0181] Optionally, this step also includes: Step S91: Obtain raw sensor data from the vehicle dynamic state dataset, and normalize the data using preset standardization rules to obtain a standardized sensor dataset.

[0182] Step S92: Based on the standardized sensor dataset, the centroid offset feature is extracted using the principal component analysis algorithm to obtain the centroid offset feature value.

[0183] Step S93: If the center of gravity offset feature value exceeds the preset fourth safety threshold, the feature value is smoothed by the Kalman filter algorithm to obtain a smoothed center of gravity offset feature value.

[0184] Step S94: If the fourth safety threshold is not exceeded, the center of gravity offset feature value is used as the smoothed center of gravity offset feature value.

[0185] Step S95: Based on the smooth centroid offset characteristic value, the first safe displacement trajectory is calculated using the Bezier curve generator to obtain the first safe displacement trajectory.

[0186] Step S96: Obtain displacement control parameters from the first safe displacement trajectory, and iteratively optimize the parameters using the SciPy optimize.minimize function to obtain optimized displacement control parameters.

[0187] Step S97: Update the Bezier curve generator based on the optimized displacement control parameters to generate the second safe displacement trajectory.

[0188] Step S98: The second safe displacement trajectory is compared with the vehicle dynamic state dataset using NumPy's array comparison function to determine whether the preset convergence condition is met. If not, the standardized sensor dataset is updated based on the second safe displacement trajectory, and the processing starting from extracting the center of gravity offset feature is repeated to obtain the final safe displacement trajectory.

[0189] For example, when acquiring raw sensor data from a vehicle dynamics dataset, assume the vehicle is equipped with an accelerometer, gyroscope, and wheel speed sensors, which collect data on the vehicle's acceleration, angular velocity, and wheel speed, respectively. The raw data may contain noise, such as sudden acceleration changes caused by road bumps. By normalizing the data using pre-defined standardization rules, data with different dimensions can be unified into a range of 0 to 1.

[0190] For example, acceleration data might range from -10 to 10 m / s², which, after normalization, can be mapped to 0 to 1 for easier subsequent analysis. This processing can eliminate dimensional differences and improve data consistency.

[0191] Specifically, when using principal component analysis (PCA) to extract center of gravity offset features, the normalized dataset is input into the PCA model. The model extracts the main feature vectors reflecting the vehicle's center of gravity offset by calculating the covariance matrix of the data. Assuming the analysis results show that the center of gravity offset feature value reaches 0.8 when the vehicle is turning, exceeding the preset fourth safety threshold of 0.6, this indicates that the vehicle may experience a roll risk due to high-speed cornering, requiring further processing.

[0192] In one embodiment, if the centroid offset feature value exceeds the fourth safety threshold, a Kalman filter algorithm is used for smoothing.

[0193] For example, Kalman filtering, through prediction and update steps, combines historical data and current observations to generate smooth centroid shift eigenvalues, assuming they decrease from 0.8 to 0.65. This smoothing process effectively reduces noise interference and improves the stability of the eigenvalues.

[0194] For example, when generating the first safe displacement trajectory based on the smooth center of gravity offset feature value, the Bezier curve generator uses the feature value as control points to calculate a smooth displacement trajectory. Assuming the vehicle needs to move laterally 0.5 meters to the right, the Bezier curve generator generates a smooth trajectory curve based on the smoothness feature value of 0.65, ensuring smooth vehicle movement and avoiding sudden swaying.

[0195] Specifically, when optimizing displacement control parameters using SciPy's `optimize.minimize` function, the initial parameters are assumed to include a displacement velocity of 0.2 m / s and an acceleration of 0.1 m / s². The optimization process adjusts the velocity to 0.25 m / s and the acceleration to 0.08 m / s² by minimizing the trajectory error. These optimized parameters make the displacement more consistent with the vehicle's dynamic characteristics.

[0196] In one embodiment, when updating the Bezier curve generator to generate a second safe displacement trajectory, a more accurate trajectory is generated based on the optimized parameters.

[0197] For example, the radius of curvature of the second trajectory is adjusted from 5 meters to 5.2 meters to better suit the current road conditions. Using NumPy's array comparison function, the second trajectory is compared with the vehicle dynamic state dataset to check if the trajectory point deviation is less than the preset trajectory point deviation threshold of 0.01 meters. If the deviation is large, such as 0.02 meters, the standardized sensor dataset is updated, the center of gravity offset features are re-extracted, and iterative optimization is performed until the convergence condition is met.

[0198] For example, once the final safe displacement trajectory is generated, it can be used as control command input for the hydraulic lateral displacement device, ensuring the vehicle's safe movement under complex road conditions. This iterative optimization approach improves the accuracy and adaptability of the trajectory, providing reliable data support for vehicle dynamic control.

[0199] like Figure 4 As shown, a second aspect of the present invention provides a fire truck rollover risk prevention and control system, which uses the method described above to prevent and control the risk of fire truck rollover. The system includes: The sensor data preprocessing module is used to acquire tilt angle, tire pressure, terrain height and obstacle distance data from multi-source sensors in real time, and use filtering and standardization algorithms to remove noise and generate a standardized sensor dataset. The center of gravity offset calculation module is used to integrate tilt angle and tire pressure data based on a standardized sensor dataset using a weighted fusion algorithm to calculate the real-time center of gravity offset of the vehicle and generate center of gravity offset feature values. The rollover risk assessment module is used to compare the center of gravity offset characteristic value with a preset first safety threshold. When the center of gravity offset characteristic value exceeds the preset first safety threshold, the rollover risk assessment model analyzes the correlation between the tilt angle and the terrain height to generate a rollover risk level. The hydraulic control optimization module is used to optimize the pressure value and displacement of the hydraulic lateral displacement device based on the rollover risk level using a dynamic programming algorithm, and to generate a hydraulic control instruction set. The smooth movement control module is used to extract target pressure and displacement values ​​from the hydraulic control command set, combine tire pressure data from the standardized sensor dataset, calculate movement speed and frequency parameters, and generate a smooth movement control sequence. The environmental perception modeling module is used to generate a dynamic environmental adaptation model by analyzing obstacle distribution, terrain height change trends and center of gravity offset characteristics through environmental perception algorithms when the obstacle distance is less than a preset distance threshold, based on the obstacle distance data in the stable motion control sequence and the obstacle distance data in the standardized sensor dataset. The path planning generation module is used to generate a lateral displacement path plan based on the dynamic environment adaptation model, by integrating the center of gravity offset feature value and the steady movement control sequence, and obtain a safe displacement trajectory containing path point data. The displacement execution module is used to extract path point data from the safe displacement trajectory, transmit it to the hydraulic lateral displacement device in real time, perform lateral displacement operations, and generate a vehicle dynamic status dataset. The data loop update module is used to update the standardized sensor dataset based on the vehicle dynamic status dataset, and loop through the process from generating center of gravity offset feature values ​​to generating safe displacement trajectory, continuously optimizing displacement control parameters.

[0200] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to specific implementations. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for preventing the risk of fire truck rollover, characterized in that, The method includes: S1 acquires tilt angle, tire pressure, terrain height, and obstacle distance data from multi-source sensors in real time, removes noise, and generates a standardized sensor dataset. S2, based on a standardized sensor dataset, integrates tilt angle and tire pressure data to calculate the real-time center of gravity offset of the vehicle and generate center of gravity offset feature values; S3. Based on the comparison between the center of gravity offset characteristic value and the preset first safety threshold, when the center of gravity offset characteristic value exceeds the preset first safety threshold, the correlation between the tilt angle and the terrain height is analyzed through the rollover risk assessment model to generate a rollover risk level. S4, based on the rollover risk level, optimize the pressure value and displacement of the hydraulic lateral displacement device, and generate a hydraulic control command set; S5 extracts the target pressure value and displacement from the hydraulic control instruction set, combines the tire pressure data in the standardized sensor dataset, calculates the moving speed and frequency parameters, and generates a smooth moving control sequence. S6. Based on the smooth movement control sequence and obstacle distance data in the standardized sensor dataset, when the obstacle distance is less than the preset distance threshold, analyze the obstacle distribution, terrain height change trend and center of gravity offset characteristic value to generate a dynamic environment adaptation model. S7. Based on the dynamic environment adaptation model, the center of gravity offset characteristic value and the steady movement control sequence are integrated to generate the lateral displacement path planning and obtain the safe displacement trajectory containing path point data. S8 extracts path point data from the safe displacement trajectory and transmits it to the hydraulic lateral displacement device in real time to perform lateral displacement operation and generate a vehicle dynamic status dataset. S9 updates the standardized sensor dataset based on the vehicle dynamic state dataset, and iteratively executes the process from generating center of gravity offset feature values ​​to generating safe displacement trajectory, continuously optimizing displacement control parameters.

2. The method for preventing the rollover risk of fire trucks according to claim 1, characterized in that, Step S2, based on a standardized sensor dataset, integrates tilt angle and tire pressure data to calculate the real-time center of gravity offset of the vehicle and generate center of gravity offset feature values, including: Step S21: Obtain standardized tilt angle and tire pressure data from the sensor dataset, remove noise through median filtering, and obtain the cleaned dataset; Step S22: Based on the cleaned dataset, calculate the weighted average of the tilt angle and tire pressure data to generate a preliminary estimate of the center of gravity offset. Step S23: If the preliminary estimate of the center of gravity offset exceeds the preset second safety threshold, the estimate of the center of gravity offset is smoothed to obtain the smoothed center of gravity offset value. Step S24: For the smoothed centroid offset value, calculate the covariance matrix and extract the first two largest eigenvalues ​​to generate the centroid offset feature vector. Step S25: Obtain the offset direction and magnitude information from the center of gravity offset feature vector, and calculate the real-time center of gravity offset feature value; Step S26: Based on the real-time center of gravity offset feature value, determine the vehicle's center of gravity offset state and obtain the vehicle stability classification result; Step S27: Generate output data based on the vehicle stability classification results.

3. The method for preventing the rollover risk of fire trucks according to claim 2, characterized in that, Step S25 involves obtaining offset direction and magnitude information from the center of gravity offset feature vector and calculating the real-time center of gravity offset feature value, including: The following function is used to calculate the real-time centroid offset eigenvalues: f(x) = gx + b Where f(x) is the real-time centroid offset feature function, x is the amplitude information, g is the preset weight, and b is the offset constant.

4. The method for preventing the rollover risk of fire trucks according to claim 1, characterized in that, Step S3 involves comparing the center of gravity offset characteristic value with a preset first safety threshold. If the center of gravity offset characteristic value exceeds the preset first safety threshold, the correlation between the tilt angle and terrain height is analyzed using a rollover risk assessment model to generate a rollover risk level, including: Step S31: Obtain real-time vehicle center of gravity offset feature value and terrain height data through sensors; Step S32: If the real-time centroid offset feature value exceeds the preset first safety threshold, input it into the support vector machine classifier to obtain the initial risk assessment result. Step S33: Extract the tilt angle from the initial risk assessment results; Step S34: Fit the tilt angle and terrain height data to determine the correlation coefficient; Step S35: Calculate the rollover risk index value based on the correlation coefficient and vehicle stability parameters; Step S36: If the rollover risk index value is higher than the preset risk threshold, the rollover risk level is adjusted in combination with environmental impact factors to determine the final risk level. Step S37: Update the dynamic change of the center of gravity based on real-time sensor data and terrain complexity to obtain the updated center of gravity offset feature value; Step S38: If the updated centroid offset feature value exceeds the preset first safety threshold, then re-input it into the support vector machine classifier to obtain a new initial risk assessment result. Step S39: Extract a new tilt angle from the new initial risk assessment results; Step S310: Fit the new tilt angle and terrain height data to determine the new correlation coefficient; Step S311: Calculate the new rollover risk index value based on the new correlation coefficient and vehicle stability parameters. Step S312: If the new rollover risk index value is higher than the preset risk threshold, then the new rollover risk level is adjusted in combination with environmental impact factors to determine the continuous risk level.

5. A method for preventing the rollover risk of fire trucks according to claim 4, characterized in that, Step S35, based on the correlation coefficient and vehicle stability parameters, calculates the rollover risk index value, including: The rollover risk index is calculated using the following formula: Rollover risk index = Pearson correlation coefficient × (1 - vehicle stability parameter), The Pearson correlation coefficient represents the linear correlation between tilt and terrain, and the vehicle stability parameter is a preset range of 0 to 1 based on vehicle load.

6. The method for preventing the rollover risk of fire trucks according to claim 5, characterized in that, In step S36, if the rollover risk index value is higher than the preset risk threshold, the rollover risk level is adjusted in conjunction with environmental impact factors to determine the final risk level, including: The final risk level is calculated using the following formula: Final risk level = Initial risk level + Environmental impact factor × 0.5 The initial risk level is extracted from the initial risk assessment, and the environmental impact factor is a preset value based on wind speed and road surface moisture.

7. The method for preventing the rollover risk of fire trucks according to claim 1, characterized in that, Step S4 involves optimizing the pressure and displacement of the hydraulic lateral displacement device based on the rollover risk level, and generating a hydraulic control command set, including: Step S41: Acquire real-time data from the sensor, process the real-time data using a Kalman filter, and obtain vehicle operating condition parameters; Step S42: Calculate the rollover risk level based on the vehicle operating condition parameters; Step S43: Construct an objective function based on the rollover risk level, initial pressure value, and initial displacement. Step S44: Iteratively optimize the initial pressure value and initial displacement based on the objective function to obtain the optimal pressure value and optimal displacement. Step S45: Generate a hydraulic control command set using the optimal pressure value and the optimal displacement. Step S46: If the control accuracy of the hydraulic control instruction set is lower than the preset control accuracy threshold, then adjust the objective function and recalculate the new initial pressure value and the new initial displacement. Step S47: Obtain real-time data from the sensor to verify the execution efficiency of the hydraulic control instruction set; Step S48: Generate a new hydraulic control instruction set based on instruction execution efficiency.

8. A method for preventing the rollover risk of fire trucks according to claim 9, characterized in that, Step S42, which calculates the rollover risk level based on vehicle operating condition parameters, includes: The rollover risk level is calculated using the following formula: R = k1v + k2a + k3θ Where R is the rollover risk level, v is the speed, a is the acceleration, θ is the tilt angle, and k1, k2, and k3 are preset coefficients.

9. A method for preventing the rollover risk of fire trucks according to claim 8, characterized in that, Step S43 involves constructing an objective function based on the rollover risk level, initial pressure value, and initial displacement, including: The objective function is: ,in, Let P be the objective function, D be the initial pressure value, and P0 and D0 be the initial displacement values. , , For weights.

10. A fire truck rollover risk prevention and control system, characterized in that, The system for preventing the risk of fire truck rollover by means of any one of claims 1-9, the system comprising: The sensor data preprocessing module is used to acquire tilt angle, tire pressure, terrain height and obstacle distance data from multi-source sensors in real time, and use filtering and standardization algorithms to remove noise and generate a standardized sensor dataset. The center of gravity offset calculation module is used to integrate tilt angle and tire pressure data based on a standardized sensor dataset using a weighted fusion algorithm to calculate the real-time center of gravity offset of the vehicle and generate center of gravity offset feature values. The rollover risk assessment module is used to compare the center of gravity offset characteristic value with a preset first safety threshold. When the center of gravity offset characteristic value exceeds the preset first safety threshold, the rollover risk assessment model analyzes the correlation between the tilt angle and the terrain height to generate a rollover risk level. The hydraulic control optimization module is used to optimize the pressure value and displacement of the hydraulic lateral displacement device based on the rollover risk level using a dynamic programming algorithm, and to generate a hydraulic control instruction set. The smooth movement control module is used to extract target pressure and displacement values ​​from the hydraulic control command set, combine tire pressure data from the standardized sensor dataset, calculate movement speed and frequency parameters, and generate a smooth movement control sequence. The environmental perception modeling module is used to generate a dynamic environmental adaptation model by analyzing obstacle distribution, terrain height change trends and center of gravity offset characteristics through environmental perception algorithms when the obstacle distance is less than a preset distance threshold, based on the obstacle distance data in the stable motion control sequence and the obstacle distance data in the standardized sensor dataset. The path planning generation module is used to generate a lateral displacement path plan based on the dynamic environment adaptation model, by integrating the center of gravity offset feature value and the steady movement control sequence, and obtain a safe displacement trajectory containing path point data. The displacement execution module is used to extract path point data from the safe displacement trajectory, transmit it to the hydraulic lateral displacement device in real time, perform lateral displacement operations, and generate a vehicle dynamic status dataset. The data loop update module is used to update the standardized sensor dataset based on the vehicle dynamic status dataset, and loop through the process from generating center of gravity offset feature values ​​to generating safe displacement trajectory, continuously optimizing displacement control parameters.

Images