A vehicle scheduling optimization method based on the Internet of Things

Abstract

The present application provides a vehicle scheduling optimization method based on the Internet of Things, which relates to the field of Internet of Things technology, including: obtaining raw vehicle status data, smoothing the raw data using a Kalman filter algorithm to obtain first processed data; filling missing values using an interpolation algorithm based on the first processed data to obtain second processed data; extracting key status parameters from the second processed data, and correcting the parameters in real time using a time series prediction model to obtain third processed data; based on the fifth processed data, performing a secondary screening of warning signals using a support vector machine classification algorithm to obtain sixth processed data; and extracting vehicle scheduling requirements from the sixth processed data. The present invention utilizes the Apache Spark distributed computing framework and a genetic algorithm to quickly generate an optimal scheduling solution. This method effectively improves the accuracy of vehicle status monitoring and scheduling efficiency, providing comprehensive technical support for fleet management.

Description

Technical Field

[0001] The present invention relates to the technical field of Internet of Things, and in particular to a vehicle scheduling optimization method based on the Internet of Things. Background Art

[0002] During vehicle dispatch, real-time collection and analysis of various vehicle status parameters are required to provide timely insights into vehicle operating conditions and adjust dispatch strategies accordingly. However, data collected by on-board sensors can be inaccurate, incomplete, and delayed, resulting in discrepancies between the acquired vehicle status information and the actual situation. Furthermore, external factors such as traffic congestion, accidents, and weather can affect vehicle driving conditions, making accurate estimated arrival times difficult to estimate. When triggering warning mechanisms, establishing appropriate warning rules based on vehicle status parameter thresholds to avoid excessive false alarms or missed alerts is also a pressing issue. Furthermore, dynamic adjustments to dispatch strategies require comprehensive consideration of multiple factors, such as vehicle location, cargo load, and remaining fuel levels. Rapidly identifying the optimal dispatch solution within this massive amount of data places high demands on the dispatch system's computing power. Addressing these technical challenges requires optimization and innovation across multiple processes, including data collection, transmission, storage, analysis, and decision-making, to improve the accuracy and efficiency of vehicle dispatch. Summary of the Invention

[0003] The present invention provides a vehicle scheduling optimization method based on the Internet of Things, which mainly includes:

[0004] Obtain the original data of vehicle status, and use the Kalman filter algorithm to smooth the original data to obtain first processed data; based on the first processed data, use the interpolation algorithm to fill the missing values to obtain second processed data; extract the key state parameters in the second processed data, and use the time series prediction model to perform real-time correction on the parameters to obtain third processed data; based on the third processed data, combine road condition information, weather forecast data and historical driving records, and use a multi-factor regression model to dynamically evaluate the vehicle driving status to obtain fourth processed data; extract the vehicle location distribution, cargo loading, and remaining fuel information from the fourth processed data, and use the adaptive threshold algorithm to dynamically adjust the warning rules to obtain fifth processed data; based on the fifth processed data, use the support vector machine classification algorithm to perform secondary screening on the warning signal to obtain sixth processed data; extract the vehicle scheduling requirements from the sixth processed data, and use the distributed computing framework to parallel optimize the scheduling strategy to obtain seventh processed data; based on the seventh processed data, use the genetic algorithm to quickly generate the optimal scheduling plan to obtain the final scheduling result.

[0005] Furthermore, the raw data is smoothed using a Kalman filter algorithm to obtain first processed data, including: obtaining vehicle speed, acceleration and angular data collected by on-board sensors, establishing a Kalman filter model, and setting a process noise covariance matrix and a measurement noise covariance matrix; constructing a state transfer matrix, an observation matrix and a noise matrix of the Kalman filter; inputting the raw data into the Kalman filter model, and recursively estimating the optimal value of the vehicle state at each moment through prediction and updating; in the prediction step, predicting the state prior estimate and the error covariance matrix at the current moment based on the state estimate value at the previous moment and the state transfer matrix; in the update step, calculating the Kalman gain, and correcting the predicted value using the observation value at the current moment to obtain the state posterior estimate and the updated error covariance matrix; iteratively executing the prediction and update steps until the raw vehicle state data at all time points are processed; evaluating the vehicle state estimate after Kalman filter processing, and if the error exceeds a preset threshold, adjusting the process noise covariance matrix and the measurement noise covariance matrix parameters of the Kalman filter.

[0006] Furthermore, an interpolation algorithm is used to fill missing values to obtain second processed data, including: constructing a data model based on the distribution characteristics and change trends of existing data to predict the value range of missing data; using a linear interpolation method to estimate and fill missing continuous values; using a nearest neighbor algorithm to interpolate missing categorical data; performing outlier detection on the filled data, analyzing the distribution characteristics of the outliers, setting an anomaly threshold, and eliminating abnormal data points introduced by the interpolation process; expanding the data scale of the data after removing outliers through data enhancement technology; for continuous values, generating new data samples by adding random noise; for categorical data, using a synthetic minority class oversampling technology algorithm to synthesize new minority class samples to balance the category distribution of the data set; dividing the expanded data set into a training set and a test set; applying different interpolation algorithms to fill missing data, comparing and evaluating the prediction performance of each algorithm on the training set and the test set, selecting the interpolation model with the best filling effect, and using it to estimate all missing values in the original data set.

[0007] Furthermore, the parameters are corrected in real time using a time series prediction model to obtain third processed data, including: obtaining key parameter values reflecting the operating status of the equipment; using the obtained historical parameter value sequence to establish an autoregressive integral sliding average model; using the established autoregressive integral sliding average model, according to the historical state parameter value sequence of the equipment, the state parameters obtained at the current moment are predicted and corrected in real time to obtain the actual operating parameters of the equipment at the current moment; substituting the corrected current state parameters into the autoregressive integral sliding average model, and performing continuous rolling prediction of the equipment state parameters in a subsequent period of time in the form of a sliding time window to obtain a predicted parameter sequence for a period of time in the future; setting a normal value range threshold based on the predicted equipment operating parameter sequence to determine whether the predicted parameter exceeds the threshold; if the predicted parameter at a certain moment exceeds the normal range, issuing an early warning message; obtaining the real-time operating parameters of the equipment, and comparing them in real time with the parameter sequence predicted by the autoregressive integral sliding average model to calculate the deviation value between the actual parameter and the predicted parameter.

[0008] Furthermore, in combination with road condition information, weather forecast data and historical driving records, a multi-factor regression model is used to dynamically evaluate the vehicle driving state to obtain fourth processed data, including: obtaining real-time road congestion and accident information; obtaining temperature, humidity and visibility data for a period of time in the future; reading the vehicle's historical driving data, the historical driving data including driving speed, acceleration and braking frequency data; inputting the obtained real-time road condition information, the weather forecast data and the historical driving record data into a vehicle driving state evaluation model pre-trained based on a neural network algorithm; the vehicle driving state evaluation model calculates the current vehicle's driving state evaluation score by weighting and comprehensively analyzing various factors; the vehicle driving state score calculated according to the vehicle driving state evaluation model is compared with a preset safety threshold; if the score exceeds the safety threshold, it is determined that the current vehicle is in a dangerous driving state, and a safety warning is automatically issued to the driver; the vehicle driving state evaluation score is compared with the data in the vehicle's historical driving record database, and a clustering algorithm is used to identify whether there is an abnormal change trend in the vehicle's driving state.

[0009] Furthermore, an adaptive threshold algorithm is used to dynamically adjust the warning rules to obtain fifth processed data, including: according to the vehicle location distribution information, a density-based clustering algorithm is used to divide the vehicles into regions to obtain the vehicle distribution density in different regions; for each region, the cargo loading and remaining fuel data of the vehicles in the region are obtained, and outliers and redundant data are removed to obtain a normalized feature data set; according to the vehicle's historical operation data, the judgment criteria for the vehicle status are defined, and the feature data set is annotated to obtain label data of the vehicle status; using the label data and the feature data, a correlation model between the cargo loading and remaining fuel and the vehicle status is trained through a logistic regression algorithm to obtain parameters and thresholds of the correlation model; during the operation of the vehicle, the cargo loading and remaining fuel data of the vehicle are collected in real time to construct time series data; the Kalman filter algorithm is used to filter the time series data to estimate the optimal values of the cargo loading and remaining fuel; the filtered cargo loading and remaining fuel data are input into the correlation model to judge the vehicle status.

[0010] Furthermore, the warning signal is screened twice using a support vector machine classification algorithm to obtain sixth processed data, including: extracting the features of the warning signal to form a warning signal feature vector; using the labeled warning signal sample data, using the support vector machine algorithm, and training a support vector machine classification model by adjusting the kernel function and penalty coefficient parameters; for new warning signals, its feature vector is input into the trained support vector machine classification model to perform a binary classification prediction; the support vector machine classification model will divide the warning signal into two categories: normal signals or abnormal signals; for warning signals predicted as abnormal signals by the support vector machine classification model, they are marked as potential false alarms or missed alarms; a confidence threshold is set, and when the confidence of the abnormal signal is lower than the confidence threshold, the warning signal is determined to be a false alarm or missed alarm and is filtered out; the warning signal with a confidence higher than the confidence threshold is retained as the sixth processed data; according to the classification results of the support vector machine classification model, the warning signal with a higher confidence is screened out from the sixth processed data as the final effective warning signal.

[0011] Furthermore, the genetic algorithm is used to quickly generate the optimal scheduling scheme and obtain the final scheduling result, including: initializing a group of scheduling schemes as the initial population, each scheduling scheme contains the execution order of tasks and the allocated resource information; for each scheduling scheme in the population, the fitness value of the scheme is calculated by weighted summation according to the priority, execution time and resource utilization factors of the task, which is used as an indicator to evaluate the quality of the scheduling scheme; the population is selected, crossed and mutated according to the fitness value to generate a new scheduling scheme; the selection operation adopts the roulette wheel selection method to select from the current population according to the size of the fitness value The larger the fitness value of an individual, the higher the probability of being selected; the crossover operation adopts the double-point crossover method, randomly selects two crossover points, and exchanges the gene fragments of the two selected individuals between the crossover points; the mutation operation adopts the basic bit mutation method, inverts certain genes of the individual with a certain probability, and introduces new gene combinations; it is judged whether the maximum fitness value of the current population has not changed significantly for multiple consecutive iterations, or has reached the preset maximum number of iterations; if it meets the requirements, the scheduling plan with the highest fitness value is output as the optimal plan, the execution order of tasks and resource allocation are determined, and the final scheduling result is generated.

[0012] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0013] The present invention discloses a vehicle scheduling optimization method based on the Internet of Things. In view of the problems of inaccuracy, incompleteness and delay in vehicle-mounted sensor data, the present invention adopts Kalman filtering, interpolation algorithm and time series prediction model to process and correct the data. Combined with real-time road conditions, weather and historical records, the vehicle driving status is dynamically evaluated through a multi-factor regression model. In order to solve the problems of state threshold setting and false alarm and missed alarm, the present invention adopts an adaptive threshold algorithm and a support vector machine classification algorithm to optimize the early warning rules. Finally, in view of the massive data processing and real-time scheduling requirements of multiple vehicles, multiple routes and multiple target points, the present invention uses the Apache Spark distributed computing framework and genetic algorithm to quickly generate the optimal scheduling plan. This method effectively improves the accuracy of vehicle status monitoring and scheduling efficiency, and provides comprehensive technical support for fleet management. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 This is a flow chart of a vehicle scheduling optimization method based on the Internet of Things of the present invention. DETAILED DESCRIPTION

[0015] To help those skilled in the art better understand the technical solutions in this specification, the following will provide a clear and complete description of the technical solutions in the embodiments of this specification, in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of this specification, not all of them. All other embodiments derived by those skilled in the art based on the embodiments in this specification without creative effort shall fall within the scope of protection of this specification.

[0016] like Figure 1 In this embodiment, a vehicle dispatch optimization method based on the Internet of Things may specifically include:

[0017] Step S101: acquiring original vehicle status data from an onboard sensor, and using a Kalman filter algorithm to smooth the original data to obtain first processed data to address the problem of data inaccuracy.

[0018] Real-time vehicle state data, such as speed, acceleration, and heading angle, is acquired from onboard sensors. To address data inaccuracies, a Kalman filter algorithm is used for data preprocessing. Based on the characteristics of the vehicle state data, a Kalman filter model is established, and appropriate parameters for the process noise covariance matrix Q and the measurement noise covariance matrix R are set. The covariance matrices Q and R represent the uncertainty of the vehicle state estimate and the uncertainty of the measurement noise, respectively, and need to be appropriately set based on the vehicle's motion characteristics and sensor performance. Based on the set Q and R parameters, the Kalman filter's state transition matrix, observation matrix, and noise matrix are constructed to form a complete Kalman filter model. The acquired raw vehicle state data is input into the Kalman filter model at each time point. Through two steps, prediction and update, the optimal vehicle state at each moment is recursively estimated. In the prediction step, the current state prior estimate and error covariance matrix are predicted based on the previous state estimate and state transition matrix. In the update step, the Kalman gain is calculated, and the predicted value is corrected using the current observation to obtain the state posterior estimate and the updated error covariance matrix. The prediction and update steps are iteratively performed until the raw vehicle state data for all time points has been processed. A fixed time window, such as 5 seconds, can be set to determine the termination condition for the iteration. The vehicle state estimate after Kalman filtering is evaluated for smoothness and accuracy relative to the raw data using metrics such as mean square error. If the error exceeds a preset threshold, the Kalman filter's Q and R parameters are appropriately adjusted based on the evaluation results to improve data smoothing. Optimization algorithms such as grid search can be used to search for the optimal Q and R parameter combination within a range of values. Through continuous iterative Kalman filtering and parameter optimization, smooth and accurate vehicle state estimates are ultimately obtained, serving as input for subsequent vehicle control and decision-making.

[0019] Specifically, the Kalman filter algorithm is an excellent data preprocessing method, particularly suitable for processing on-board sensor data. This algorithm recursively estimates the vehicle state by establishing a state-space model and combining measurements with prior estimates. When processing vehicle state data, the state vector must first be determined, typically consisting of position, velocity, and acceleration. Taking the longitudinal motion of a vehicle as an example, the state vector can be defined as [position, velocity, acceleration]. The state transition matrix describes the evolution of the vehicle state in discrete time. Assuming a sampling interval of Δt, it can be expressed as: [[1, Δt, 0.5Δt^2], [0, 1, Δt], [0, 0, 1]]. The observation matrix is determined based on the actual measurable state variables. If only position and velocity can be measured directly, the observation matrix is: [[1, 0, 0], [0, 1, 0]]. The setting of the process noise covariance matrix Q and the measurement noise covariance matrix R is crucial. Q reflects the uncertainty of the state-space model, while R reflects the measurement uncertainty. For example, for a high-precision GPS receiver, the position measurement error can be in the meter level, and the velocity measurement error can be in the centimeter / second level. Based on this, the diagonal elements of the R matrix can be preliminarily set. In practical applications, a sliding time window method can be used to process continuous data streams. Assuming a window size of 5 seconds and a sampling frequency of 10 Hz, 50 data points are processed at a time. For the data within each time window, the prediction and update steps of the Kalman filter are performed. In the prediction step, the optimal estimate and the state transition matrix at the previous moment are used to calculate the prior estimate for the current moment. In the update step, the prior estimate is corrected using the Kalman gain based on the observed values to obtain the posterior estimate. To evaluate the filtering effect, the root mean square error (RMSE) between the filtered state estimate and the original measurement can be calculated. If the RMSE exceeds a preset threshold, such as a position error greater than 0.5 meters or a velocity error greater than 0.2 meters per second, the Q and R matrices need to be adjusted. A grid search method can be used during this adjustment process. For example, the diagonal elements of the Q matrix are searched within the range [10^-6, 10^-5, ..., 10^-1], and the diagonal elements of the R matrix are searched within the range [10^-2, 10^-1, ..., 10^2]. A Kalman filter is applied to each parameter combination, the RMSE is calculated, and the parameter combination that minimizes the RMSE is selected. Through repeated iterative optimization, the resulting vehicle state estimate data becomes smoother and more accurate. This processed data can be used for advanced driver assistance features such as vehicle trajectory prediction and adaptive cruise control, improving driving safety and comfort. For example, in adaptive cruise control, accurate speed and acceleration estimates can more accurately adjust the following distance to the vehicle ahead, achieving smoother acceleration and deceleration.

[0020] Step S102: Based on the first processed data, an interpolation algorithm is used to fill in missing values to solve the problem of incomplete data, thereby obtaining second processed data.

[0021] Based on the distribution characteristics and changing trends of the existing data, the data mining tool SPSS Modeler was used to construct a data model to predict the possible value range of the missing data. Linear interpolation was used to estimate and fill missing continuous values; the nearest neighbor algorithm was used for interpolation of missing categorical data. The imputed data was exported as a second dataset. Outlier detection was performed on the second dataset, and the distribution characteristics of outliers were analyzed using statistical charts such as box plots. Anomalous data points introduced by the interpolation process were removed by setting an appropriate threshold. The data after outlier removal was saved as a third dataset. Based on the third dataset, the training data size was expanded using data augmentation techniques. For continuous values, new data samples were generated by adding random noise; for categorical data, the SMOTE algorithm was used to synthesize new minority class samples to balance the class distribution of the dataset. The expanded dataset was randomly divided into training and test sets. Using the Python Scikit-learn library, different interpolation algorithms, such as linear interpolation and spline interpolation, were applied to fill missing data. The prediction performance of each algorithm on the training and test sets was compared and evaluated. By calculating evaluation metrics such as root mean square error and mean absolute error, the interpolation model with the best imputation effect was selected and used to estimate all missing values in the original dataset to obtain a complete dataset. Finally, the complete dataset after imputation was randomly divided into a training set and a validation set in an 8:2 ratio. The model was trained using the XGBoost algorithm on the training set, and its generalization performance was evaluated on the validation set. A grid search was used to optimize the XGBoost hyperparameters to obtain the optimal machine learning model for subsequent prediction tasks.

[0022] Specifically, when building data models, the data mining tool SPSS Modeler can use algorithms such as decision trees and neural networks to predict the likely range of values for missing data. For example, in a car sales dataset, one might need to predict the missing "sales price." Analysis of existing data reveals significant correlations between "model," "mileage," and "age" and "sales price." Leveraging these features, SPSS Modeler can build a prediction model that provides a reasonable estimate range for the missing "sales price." Linear interpolation is a simple and effective method for filling missing values for continuous data. Consider an environmental monitoring dataset where temperature data for a particular time point is missing. Linear interpolation can be performed using the temperature values at adjacent time points. If the temperatures before and after the missing point are 20°C and 22°C, respectively, with equal time intervals, the temperature at the missing point can be estimated to be 21°C. This method assumes that temperature changes follow a linear trend over short periods of time and is suitable for data with gradual fluctuations. The nearest neighbor algorithm is very effective for handling missing categorical data. Taking medical diagnosis data as an example, if a patient's "symptom type" information is missing, we can find records of patients with similar characteristics (such as age, gender, and other symptoms) and use the most frequent symptom type among these similar patients to fill in the missing value. This method, based on the assumption that similar patients are likely to have similar symptoms, preserves the inherent correlation of the data. Using boxplots for outlier detection is an intuitive and effective method. In a dataset of student grades, we can plot boxplots of scores for each subject. Suppose a boxplot of math scores shows that the upper quartile is 90 and the lower quartile is 60. Any score below 15 (lower quartile minus 1.5 times the interquartile range) or above 135 (upper quartile plus 1.5 times the interquartile range) could be considered an outlier. These outliers may be caused by data entry errors or improper interpolation and require further verification or removal. Data augmentation techniques play an important role in expanding the size of training data. For continuous numerical data, such as stock price data, small random fluctuations can be added to the original prices to generate new samples. For example, for a stock price of 100 yuan, random noise can be added within a range of ±5% to generate new price data between 95 yuan and 105 yuan. This method can simulate small market fluctuations and increase the model's sensitivity to price changes. The SMOTE algorithm is very effective when working with unbalanced datasets. Suppose in a credit card fraud detection dataset, fraudulent transactions only account for 1% of the total transactions. The SMOTE algorithm can be used to synthesize new fraudulent transaction samples, increasing the proportion of fraudulent transactions to 10%. This balance can help the model better learn the characteristics of fraudulent transactions and improve detection accuracy. Root mean square error (RMSE) and mean absolute error (MAE) are commonly used metrics when evaluating the performance of interpolation algorithms. For example, in a weather forecast dataset, missing rainfall data may need to be filled in.Assuming the RMSE of linear interpolation is 2.5mm and the RMSE of spline interpolation is 1.8mm, spline interpolation would tend to be chosen as the final filling method. Spline interpolation can better capture the nonlinear variation characteristics of rainfall, and therefore performs better in this case. The XGBoost algorithm performs well when processing structured data. In a customer churn prediction model, there may be features such as "customer age", "monthly spending amount", and "number of customer service contacts". XGBoost can automatically learn the complex interactions between these features, such as discovering the rule that "customers aged 30-40 with monthly spending less than 100 yuan are more likely to churn." By optimizing hyperparameters such as the learning rate and tree depth through grid search, a model that can accurately predict churn risk without overfitting can be obtained.

[0023] Step S103 , extracting key state parameters from the second processed data, and using a time series prediction model to perform real-time correction on the parameters to address the data delay problem, thereby obtaining third processed data.

[0024] Based on the processed data, key parameters reflecting the equipment's operating status, such as temperature, pressure, and speed, are obtained. To address potential data transmission delays, an ARIMA time series prediction model is established using the acquired historical parameter value sequence. Using this established ARIMA model, the delayed state parameters acquired at the current moment are predicted and corrected in real time based on the equipment's historical state parameter value sequence, resulting in the equipment's actual operating parameters at the current moment. The corrected current state parameters are substituted into the ARIMA model, and a continuous rolling prediction of the equipment's state parameters over a subsequent period is performed using a sliding time window, resulting in a series of predicted parameters for the future. Based on the predicted equipment operating parameter sequence, a threshold within the normal value range is set to determine whether the predicted parameter exceeds the threshold. If the predicted parameter at a given moment exceeds the normal range, an alert is issued via email or text message, alerting relevant personnel to the impending equipment operating anomaly and informing them to prepare for the event. The equipment's real-time operating parameters are obtained and compared in real time with the parameter sequence predicted by the ARIMA model. The deviation between the actual and predicted parameters is calculated. A deviation threshold is set. If the actual deviation exceeds the threshold, an abnormality is detected, triggering an anomaly alarm and displaying the anomaly information on the human-machine interface. At the same time, abnormal parameters are fed back to the ARIMA prediction model, and the deviation values are used to correct the correlation coefficient terms in the model to enable the model to adapt to changes in equipment status. Based on the adjusted and optimized ARIMA model parameters, equipment operating parameters are re-predicted and continuously compared and monitored with real-time equipment parameters until the equipment returns to normal operation. Finally, the optimized and corrected ARIMA model serves as the basis for equipment operating status prediction and abnormality diagnosis. The ARIMA model parameters and the corrected equipment operating parameter sequence predicted by the model are output, providing data support for equipment health management.

[0025] Specifically, the ARIMA model is a commonly used time series forecasting method suitable for data with trends and seasonality. In equipment condition monitoring, ARIMA can effectively capture the patterns of parameter changes. For example, in an industrial cooling system, temperature is a key parameter. Suppose historical data shows that the temperature fluctuates between 20 and 25°C, exhibiting a similar daily variation pattern. The ARIMA model can learn this pattern and predict future temperature changes. The ARIMA model's rolling forecasting capability is very useful for addressing data transmission delays. For example, if sensor data is transmitted every 5 minutes but there is actually a 10-minute delay, the ARIMA model can predict the current temperature based on historical data. This prediction fills the data gap caused by the delay, enabling the monitoring module to respond promptly. Setting a threshold within the normal value range is key to early warning mechanisms. Continuing with the cooling system example, if the normal operating temperature range is 22-26°C, a warning threshold of 27°C can be set. If the ARIMA model predicts that the temperature will reach 27°C within 30 minutes, an alert will be issued. This gives operators ample time to take preventive measures, such as increasing cooling water flow or inspecting radiators. Real-time comparison of deviations between actual and predicted parameters is a key mechanism for model adaptation. Suppose an ARIMA model predicts a temperature of 24°C at a certain moment, but the actual measured value is 25.5°C. If the deviation threshold is set to 1°C, this 1.5°C difference will trigger an anomaly alert. This deviation will also be used to adjust the ARIMA model parameters for more accurate future predictions. Continuous model optimization is crucial for long-term monitoring. For example, as cooling system components age, their efficiency may gradually decline, resulting in a slight increase in overall temperature. An optimized ARIMA model can adapt to this gradual change and avoid frequent false alarms. The parameter series output by the ARIMA model provides comprehensive data support for equipment health management. For example, by analyzing the temperature change rate, it is possible to predict a decrease in radiator efficiency. If the temperature rise rate is found to be 20% faster than historical data, this may indicate the need for maintenance. This proactive maintenance strategy can significantly reduce the risk of equipment failure and improve reliability. In practical applications, ARIMA models can be combined with other technologies. For example, ARIMA predictions can be input into a machine learning model to further improve the accuracy of anomaly detection. This multi-model fusion approach leverages the strengths of both time series analysis and complex pattern recognition to provide a more comprehensive solution for equipment condition monitoring.

[0026] Step S104 , based on the third processed data, combined with the road condition information, weather forecast data and historical driving records obtained in real time from the traffic management department, a multi-factor regression model is used to dynamically evaluate the vehicle driving state to obtain fourth processed data.

[0027] Real-time road congestion and accident information is obtained through the traffic management department's API as one input factor for vehicle driving condition assessment. Weather forecast data, such as temperature, humidity, and visibility, is accessed from the Meteorological Bureau's public database for a period of time and used as another input factor for vehicle driving condition assessment. Historical driving data, including speed, acceleration, and braking frequency, is retrieved from the vehicle's onboard optical boresight (OBD) device and uploaded to the data center for storage and management, serving as the basis for assessing vehicle driving condition. Multiple factors, including real-time road conditions, weather forecast data, and historical driving records, are input into a vehicle driving condition assessment model pre-trained using a neural network algorithm. The vehicle driving condition assessment model calculates a driving condition score by weighting various factors and conducting a comprehensive analysis. The vehicle driving condition score calculated by the vehicle driving condition assessment model is compared with a preset safety threshold. If the score exceeds the threshold, the vehicle is deemed to be in a dangerous driving state and an automatic voice and visual safety warning is issued to the driver, prompting the driver to take measures such as slowing down or stopping to avoid the risk. The vehicle's driving status assessment score is compared and analyzed with the data in the vehicle's historical driving record database, and a clustering algorithm is used to identify whether there are any abnormal changes in the vehicle's driving status. If the scores of multiple consecutive assessments show an upward trend, it is predicted that the vehicle's driving risk will continue to increase in the future. Based on the vehicle's driving status risk prediction results, driving parameters such as the vehicle's cruising speed and following distance are automatically adjusted dynamically. When the predicted risk level is high, the vehicle is controlled to slow down by 10% and the following distance is increased to more than 100 meters to ensure driving safety. When the predicted risk level returns to normal, the cruising speed and following distance are restored to the original set values to improve driving efficiency.

[0028] Specifically, obtaining real-time road congestion and accident information is crucial for assessing vehicle driving conditions. For example, in large cities like Beijing, traffic management departments may provide an API that returns congestion indexes and the latest accident reports for each major road. This data can be updated every five minutes, quantifying the congestion index for the Third Ring Road from 0 to 10, while also recording the specific location and time of the accident. This information directly impacts driving condition assessments, as congestion or accidents can lead to frequent acceleration and deceleration, increasing the risk of collisions. Weather conditions significantly impact driving safety. Accessing the meteorological bureau's database provides hourly weather forecasts for the next 24 hours. For example, a forecast indicating visibility will drop below 200 meters and temperatures near 0°C in two hours suggests the possibility of icing. This information can be incorporated into the vehicle driving condition assessment model to proactively adjust driving parameters, such as reducing recommended speeds and increasing safe distances. Historical driving data recorded by the vehicle's on-board optical borehole detection (OBD) device provides a personalized basis for assessments. Suppose a vehicle has an average speed of 60 km / h, a maximum acceleration of 2 m / s², and an average of three brakes per kilometer over the past month. Once this data is uploaded to the data center, it can be used to establish a baseline for the vehicle's normal driving patterns. Any significant deviation from this baseline may be considered a potential risk. The vehicle driving state assessment model, trained using a neural network algorithm, comprehensively analyzes multiple factors. The model may weight real-time road conditions 30%, weather forecasts 20%, and historical driving data 50%. For example, if a traffic accident is detected 2 kilometers ahead, visibility is low, and the vehicle has recently braked at an above-average rate, the model may output a score from 0 to 100, with scores above 80 considered high risk. When the assessment score exceeds the safety threshold, the vehicle's automatic warning function is crucial. If the safety threshold is set at 75, when the score reaches 78, the vehicle system will immediately issue an audible warning: "Account has occurred 2 kilometers ahead. Please slow down." The recommended deceleration and safe distance between vehicles will also be displayed on the central control screen. Analyzing historical data using a clustering algorithm can identify abnormal driving state trends. For example, if a vehicle's assessment score shows an upward trend over the past week, gradually rising from an average of 65 points to 75 points, it is predicted that if no action is taken, the risk score of the next trip may exceed the high-risk limit of 80 points. Based on the risk prediction results, the vehicle's driving parameters can be proactively adjusted. When a high risk is predicted, if the vehicle's original cruising speed is set to 100km / h, it will be automatically reduced to 90km / h. At the same time, if the original following distance is set to 50 meters, it will be increased to more than 100 meters. These adjustments can effectively reduce the risk of collision and improve driving safety.

[0029] Step S105, extracting vehicle location distribution, cargo loading, and remaining fuel information from the fourth processed data, and dynamically adjusting the warning rules using an adaptive threshold algorithm based on historical data and current operating conditions to address the state threshold setting problem, thereby obtaining the fifth processed data.

[0030] Based on vehicle location distribution information, the DBSCAN density clustering algorithm is used to divide vehicles into regions and determine the vehicle distribution density in each region. For each region, data on the cargo load and remaining fuel level of vehicles in that region is collected. Through data cleaning and feature engineering, outliers and redundant data are removed to obtain a normalized feature dataset. Based on historical vehicle operation data, vehicle status judgment criteria are defined and the feature dataset is annotated to obtain vehicle status label data. Using this label and feature data, a logistic regression algorithm is used to train a correlation model between cargo load, remaining fuel level, and vehicle status, determining the model parameters and thresholds. During vehicle operation, cargo load and remaining fuel level data are collected in real time to construct time series data. A Kalman filter is used to filter the time series data to estimate optimal values for cargo load and remaining fuel level, minimizing the impact of measurement noise. The filtered cargo load and remaining fuel level data are then fed into the correlation model to determine vehicle status. If the judgment result exceeds the threshold, a warning of the corresponding level is triggered. Based on the warning level, the system dynamically adjusts the transport assignments of vehicles with warnings by matching pre-set vehicle dispatch strategies, reducing the risks of vehicle overload and fuel shortages. The adjusted vehicle dispatch strategies and transport assignments are output to form an optimized vehicle operation plan, guiding safe and efficient vehicle operation. Simultaneously, the vehicle operation plan is transmitted to relevant departments and personnel, ensuring timely monitoring of vehicle status and task progress, facilitating coordination and management.

[0031] Specifically, vehicle location distribution information is a key factor in logistics scheduling. The DBSCAN density clustering algorithm can effectively identify vehicle density distribution in different areas, providing a basis for subsequent scheduling decisions. For example, in Beijing, the algorithm might classify the area within the Fifth Ring Road as a high-density area and the suburbs as a low-density area. This classification helps optimize vehicle allocation and improve transportation efficiency. Obtaining data on the cargo load and remaining fuel level of vehicles within the region is crucial for assessing vehicle status. During data cleaning and feature engineering, it may be discovered that some vehicles report abnormally high loads, such as exceeding 120% of the vehicle's maximum load capacity. Such data should be considered outliers and removed to ensure the accuracy of subsequent analysis. The definition of vehicle status criteria requires consideration of multiple factors. For example, a vehicle with a load exceeding 90% and a remaining fuel level below 20% can be labeled as "high-risk." This criterion helps promptly identify vehicles at risk of being overloaded or running low on fuel. When training a model to associate cargo load and remaining fuel level with vehicle status, a logistic regression algorithm might find that the weight of the load on vehicle status is 0.7, while the weight of the remaining fuel level is 0.3. This means that changes in load have a more significant impact on vehicle status assessment. The Kalman filter algorithm effectively reduces the impact of measurement noise when processing real-time data. Suppose a vehicle's cargo load sensor is subject to vibration interference, resulting in large fluctuations in data over a short period of time. The Kalman filter can smooth these fluctuations, providing a more stable estimate and thus improving the accuracy of status assessment. An early warning mechanism, triggered when vehicle status assessment results exceed a threshold, is crucial for ensuring transportation safety. For example, if a vehicle's status score reaches 85 (assuming the threshold is 80), a yellow warning is immediately issued. This timely warning prevents potential safety hazards before they occur. Once the warning is triggered, transport tasks are dynamically adjusted according to pre-set scheduling strategies. For example, for vehicles approaching overload, some cargo may be transferred to nearby, lighter-loaded vehicles. This dynamic adjustment not only reduces risk but also improves overall transportation efficiency. The optimized vehicle operation plan is the result of a comprehensive consideration of multiple factors. It may include specific instructions such as recommending that certain vehicles refuel early or adjusting routes to avoid congested areas. These optimization measures can significantly improve the overall operational efficiency and safety of the fleet. Transmitting operational plans to relevant departments and personnel facilitates collaborative management. For example, the dispatch center can monitor the status and progress of each vehicle in real time, enabling rapid response to any anomalies. This information-sharing mechanism significantly enhances the flexibility of the logistics system and its ability to respond to emergencies.

[0032] Step S106: Based on the fifth processed data, a support vector machine classification algorithm is used to perform a secondary screening on the warning signal to solve the problem of false alarms and missed alarms, thereby obtaining sixth processed data.

[0033] Based on the fifth processed data, warning signal features, including signal strength, duration, and frequency, are extracted to form a warning signal feature vector. Using the labeled warning signal sample data, a support vector machine (SVM) algorithm is used to train a SVM classification model by adjusting parameters such as the kernel function and penalty coefficient. For each new warning signal, its feature vector is input into the trained SVM classification model for binary classification prediction. The SVM classification model classifies the warning signal as either a normal signal or an abnormal signal. Warning signals predicted as abnormal by the SVM classification model are marked as potential false positives or false negatives. A confidence threshold is set. When the confidence level of an abnormal signal falls below the threshold, the warning signal is considered a false positive or false negative and is filtered out. Warning signals with confidence levels above the threshold are retained as the sixth processed data. Based on the classification results of the SVM classification model, warning signals with higher confidence levels are screened from the sixth processed data as valid warning signals. These validated warning signals are then passed to subsequent business processes to guide decision-making and resolution. By using a support vector machine classification model to perform secondary classification and filtering on warning signals, we can effectively reduce the false alarm and missed alarm rates, and improve the accuracy and reliability of warning signals. Furthermore, the support vector machine classification model can be retrained based on continuously updated sample data, continuously optimizing classification results and adapting to changes in actual business needs.

[0034] Specifically, extracting warning signal features is key to accurate classification. Signal strength reflects the severity of an anomaly, duration indicates the persistence of the problem, and frequency reveals the pattern of anomaly occurrence. For example, in a logistics system, the strength of a vehicle overload warning signal may be proportional to the amount of overload, duration reflects how long the overload persists, and frequency indicates whether the vehicle is repeatedly overloaded. Support vector machine algorithms offer significant advantages in warning signal classification. By adjusting the kernel function, such as the radial basis function (RBF) kernel, nonlinearly separable warning signals can be effectively processed. Adjusting the penalty coefficient balances the complexity of the support vector machine classification model with classification accuracy. For example, in vehicle status warnings, a large penalty coefficient may cause the support vector machine classification model to be overly sensitive, misclassifying normal minor fluctuations as anomalies; whereas a small penalty coefficient may overlook potential risk signals. After the support vector machine classification model is trained, a new warning signal is input for classification. Suppose a new warning signal has a feature vector of [0.8, 30, 0.05], representing signal strength, duration (minutes), and frequency (times / hour), respectively. Based on these features, the support vector machine classification model may classify it as an anomaly. For signals identified as anomalies, further confidence assessment is crucial. Setting a reasonable threshold, such as 0.85, can effectively filter out low-confidence anomalies. For example, an anomaly signal with a confidence of 0.92 will be retained, while a signal with a confidence of 0.78 will be considered a potential false alarm and filtered out. This secondary classification and filtering mechanism significantly improves the reliability of the early warning mechanism. In practical applications, it can effectively reduce false alarms caused by sensor failures or data transmission errors. For example, a large logistics center may generate hundreds of early warning signals daily. After screening by the support vector machine classification model, only dozens of high-confidence signals may be retained, significantly reducing the burden on management personnel. Continuous optimization of the support vector machine classification model is also key to ensuring the long-term effectiveness of the early warning mechanism. As new sample data accumulates, the support vector machine classification model can be regularly retrained. This dynamic update mechanism enables the support vector machine classification model to adapt to changes in the business environment. For example, as fleet size expands and road conditions change, existing warning criteria may no longer apply. Retraining the support vector machine classification model can capture these changes and maintain the effectiveness of the warning mechanism. Ultimately, effective warning signals with high confidence are used to guide actual operations. In a logistics system, this may mean timely adjusting vehicle routes, arranging maintenance, or reallocating cargo. In this way, the warning mechanism not only improves operational safety but also optimizes resource utilization, ultimately enhancing the efficiency and reliability of the entire logistics network.

[0035] Step S107 , extracting vehicle scheduling requirements from the sixth processed data, and using the Apache Spark distributed computing framework to parallel optimize the scheduling strategy for the massive data processing problem generated by multiple vehicles, multiple routes, and multiple destinations, to obtain seventh processed data.

[0036] To meet vehicle scheduling requirements, the Scrapy distributed crawler framework was used to capture relevant data such as vehicles, routes, and destinations from massive data sources. Data was cleaned and standardized using the Pandas and NumPy libraries to remove noise and convert the data into a unified format. Key vehicle and route features, such as vehicle location, speed, and fuel consumption, as well as route length, road conditions, and speed limits, were extracted. Feature engineering was used to construct the input vectors for the vehicle scheduling optimization model. The XGBoost algorithm was used to train a vehicle scheduling strategy classification model. Based on vehicle and route characteristics, the optimal scheduling strategy was predicted and a preliminary scheduling plan was generated. The preliminary scheduling plan was loaded into the Spark distributed computing framework and partitioned using the RDD data structure to form multiple subtasks that can be executed in parallel. Using Spark's Map and Reduce operations, the scheduling subtasks were executed in parallel on multiple nodes in the cluster to optimize the vehicle scheduling strategy. Within Spark's Map operation, the tabu search algorithm was used for local optimization of each scheduling subtask. Constraints such as vehicle load, mileage, and time window are set, and a feasible scheduling solution that satisfies these constraints is searched. Taboo step sizes and taboo table lengths are set to prevent the search from becoming trapped in local optima. The optimization results of each subtask are aggregated through a Reduce operation to form a set of optimized candidate scheduling solutions. This set of candidate scheduling solutions is then globally optimized using a simulated annealing algorithm. Each vehicle's scheduling strategy is treated as a particle. By setting parameters such as the initial temperature and cooling rate, the particle's motion in state space is simulated. Good solutions are accepted, while poor solutions are probabilistically accepted, escaping local optima to find the globally optimal vehicle scheduling strategy. The optimized vehicle scheduling strategy is converted into scheduling instructions in JSON format and distributed via a Kafka message queue. Each vehicle terminal subscribes to the Kafka scheduling instruction topic, receives scheduling instructions in real time, and executes the corresponding scheduling tasks, achieving intelligent vehicle scheduling. During the execution of the scheduling task, vehicles collect information such as trajectory, speed, and fuel consumption through GPS, OBD, and other devices. This collected data is then transmitted to the stream processing framework via Kafka. Spark Streaming is used to analyze vehicle operation data in real time. If anomalies such as vehicle deviation from the planned route or speeding are detected, warnings are generated immediately, triggering dynamic optimization of the dispatch strategy. Vehicle operation data and optimized dispatch strategies are persistently stored in HDFS and HBase, forming a historical dataset for vehicle dispatch. Spark SQL is used to analyze and mine this historical data, extracting key factors influencing vehicle dispatch, optimizing the dispatch strategy model, and continuously improving the efficiency and accuracy of vehicle dispatch.

[0037] Specifically, vehicle scheduling optimization is a critical component of the logistics industry, and data-driven and intelligent algorithms can significantly improve efficiency. First, the Scrapy distributed crawler framework is used to obtain vehicle and route information from multiple data sources. For example, this data can be captured from public traffic data, weather forecasts from meteorological departments, and real-time location information uploaded by onboard devices. This data is processed using Pandas and NumPy to form standardized feature vectors. During the feature engineering phase, composite features can be constructed. For example, a "estimated time of arrival" feature can be generated by comprehensively considering information such as the vehicle's current location, destination, road conditions, and weather. This feature more comprehensively reflects the factors influencing scheduling decisions. The XGBoost algorithm performs well in training scheduling strategy classification models. It effectively handles interactions between features, such as the impact of the combination of vehicle load and road conditions on fuel consumption. By learning from historical data, the model can predict the optimal scheduling strategy under given conditions. Importing the preliminary scheduling plan into the Spark framework fully leverages the advantages of distributed computing. For example, a large-scale scheduling problem involving 100 vehicles and 20 delivery points can be broken down into multiple subtasks, each responsible for optimizing the scheduling solution for a specific area or time period. In the Map phase, the application of the Tabu Search algorithm effectively avoids being trapped in local optima. For example, when optimizing a truck's route, the algorithm might temporarily accept a seemingly inferior solution (such as a detour to avoid traffic congestion). This decision may increase travel distance in the short term, but may be a more optimal option in the long run. The global optimization phase utilizes the simulated annealing algorithm, which uniquely tolerates a certain degree of "bad" solutions during the search process. This property allows the algorithm to escape local optima and find a better global solution. For example, when optimizing the scheduling strategy for an entire fleet, it might temporarily accept suboptimal routes for some vehicles in exchange for improved overall scheduling efficiency. The optimized scheduling strategy is distributed to each vehicle terminal via a Kafka message queue. This real-time communication mechanism enables the scheduling system to rapidly respond to emergencies. For example, if a traffic accident occurs on a road, the scheduling system can immediately send new routing instructions to affected vehicles. Spark Streaming's real-time analysis capabilities provide the foundation for dynamic adjustments. For example, the dispatch system can monitor the speed and location of each vehicle in real time. If a vehicle consistently falls below expected speed, this could indicate deteriorating traffic conditions in the area, prompting the dispatch system to promptly adjust the routes of subsequent vehicles. Finally, by storing dispatch data in HDFS and HBase and performing in-depth analysis using SparkSQL, potential optimization opportunities can be identified. For example, by analyzing historical data, delivery efficiency may be particularly high during certain time periods or along certain routes. These findings can guide future dispatch decisions and continuously improve overall operational efficiency.

[0038] Step S108: Based on the seventh processed data, a genetic algorithm is used to quickly generate an optimal scheduling solution for the computing power requirement of real-time scheduling to obtain a final scheduling result.

[0039] A set of randomly initialized scheduling plans serves as the initial population. Each scheduling plan includes information about the execution order of tasks and allocated resources. For each scheduling plan in the population, a weighted summation method is used to calculate the fitness value of the plan, based on factors such as task priority, execution time, and resource utilization. This fitness value serves as a metric for evaluating the performance of the scheduling plan. The fitness value is calculated as: F = w1 × P + w2 × (1 / T) + w3 × R, where F is the fitness value, P is the task priority, T is the task execution time, R is the resource utilization, and w1, w2, and w3 are the weight coefficients of each factor. Based on the fitness value, the population undergoes selection, crossover, and mutation operations to generate new scheduling plans. The selection operation uses a roulette wheel selection method, selecting individuals from the current population based on their fitness values. The higher the fitness value, the higher the probability of selection. The crossover operation uses a two-point crossover method, randomly selecting two crossover points and swapping the gene segments between the two selected individuals. The mutation operation uses a base bit mutation method, inverting certain genes in an individual with a certain probability to introduce new gene combinations. Determine whether the maximum fitness value of the current population has remained unchanged for multiple consecutive iterations, or whether it has reached the preset maximum number of iterations. If so, proceed to step 6. Otherwise, add the newly generated scheduling solution to the population and repeat steps 2-4 for iterative optimization. The scheduling solution with the highest fitness value is output as the optimal solution. The task execution order and resource allocation are determined, generating the final scheduling result. The scheduling result is sent to the task execution module to guide the scheduling of real-time tasks. Simultaneously, task execution is monitored to obtain feedback information such as task completion time and resource usage. Using this feedback information, the genetic algorithm parameters, including population size, crossover probability, and mutation probability, are dynamically adjusted to enable the genetic algorithm to adapt to dynamic changes in task execution and improve the efficiency and quality of subsequent scheduling. Based on the adjusted genetic algorithm parameters, return to step 1 and begin a new round of real-time scheduling tasks. Repeat the above steps to continuously optimize the scheduling solution.

[0040] Specifically, the application of genetic algorithms in vehicle scheduling optimization demonstrates their powerful search capabilities. During the initialization phase, a diverse set of scheduling solutions can be generated based on historical data. For example, in an urban logistics distribution system, the initial population may contain 10 different delivery route combinations, each involving 20 vehicles and 100 delivery points. Fitness calculation is key to evaluating the quality of solutions. In practical applications, task priority may reflect the urgency of an order, execution time corresponds to the total delivery time, and resource utilization reflects vehicle loading efficiency. Assume that a delivery solution has a task priority of 0.8, an estimated execution time of 2 hours, and a resource utilization of 0.7, with weights of 0.4, 0.3, and 0.3, respectively. Its fitness value is: 0.4 × 0.8 + 0.3 × (1 / 2) + 0.3 × 0.7 = 0.62. The selection operation simulates the natural law of "survival of the fittest." When using a roulette wheel selection method, solutions with higher fitness values are more likely to be retained. This ensures that excellent features are inherited and developed within the population. The crossover operation integrates the advantages of different solutions. For example, two delivery routes may excel in certain areas. By crossing two points, these excellent fragments can be combined to create a more optimal solution. Mutation introduces randomness to prevent the algorithm from falling into local optima. In vehicle scheduling, mutations might involve randomly adjusting the order of delivery point visits or replacing the vehicle assigned to a particular delivery task. These small random changes can sometimes yield unexpected results, such as discovering a more efficient route combination. The iterative optimization process continuously improves solution quality. When the change in the maximum fitness value over multiple consecutive iterations (e.g., 50) is less than a preset threshold (e.g., 0.001) or the maximum number of iterations (e.g., 1000) is reached, the algorithm is considered to have converged to a relatively optimal solution. The output and execution of the optimal solution are the implementation phase of the algorithm. For example, the finalized dispatch plan might specify that vehicle 1 delivers to point A first and then to point B, while vehicle 2 delivers to points C, D, and E. These instructions are communicated to the driver via the onboard terminal to guide actual delivery work. Feedback mechanisms and dynamic adjustments enable the algorithm to adapt to changing circumstances. For example, if certain road sections are frequently congested, causing mission delays, the mutation probability can be appropriately increased to increase the opportunity to explore new routes. Alternatively, if population diversity is observed to be decreasing, the population size can be increased to introduce more possible solutions. This iterative optimization process enables the scheduling system to continuously learn and improve. Through accumulated experience, the scheduling system may discover that delivery efficiency is particularly high on specific routes during certain times, or that certain vehicle models are more suitable for specific delivery tasks. These findings can guide future scheduling decisions and continuously improve overall operational efficiency.

[0041] Obviously, those skilled in the art may make various changes and modifications to the embodiments of the present application without departing from the spirit and scope of the embodiments of the present application. Thus, if these modifications and variations of the embodiments of the present application fall within the scope of the claims of the present application and their equivalents, the present application also intends to include such modifications and variations.

Claims

1. A vehicle scheduling optimization method based on the Internet of Things, characterized in that: include: Acquire original vehicle state data, and smooth the original data using a Kalman filter algorithm to obtain first processed data; Filling missing values using an interpolation algorithm based on the first processed data to obtain second processed data; extracting key state parameters from the second processed data, and modifying the parameters in real time using a time series prediction model to obtain third processed data; Based on the third processed data, combined with road condition information, weather forecast data and historical driving records, a multi-factor regression model is used to dynamically evaluate the vehicle driving state to obtain fourth processed data; Extracting vehicle location distribution, cargo loading, and remaining fuel information from the fourth processed data, and dynamically adjusting warning rules using an adaptive threshold algorithm to obtain fifth processed data; Based on the fifth processed data, the warning signal is secondary screened using a support vector machine classification algorithm to obtain sixth processed data; Extracting vehicle dispatch requirements from the sixth processed data, and optimizing the dispatch strategy in parallel using a distributed computing framework to obtain seventh processed data; According to the seventh processed data, the optimal scheduling plan is quickly generated using a genetic algorithm to obtain a final scheduling result; Get real-time road congestion and accident information; Obtain temperature, humidity, and visibility data for a period of time in the future; Read the vehicle's historical driving data, including driving speed, acceleration, and braking frequency data; The acquired real-time traffic information, weather forecast data and historical driving record data are input into a vehicle driving state evaluation model pre-trained based on a neural network algorithm.

2. The method according to claim 1, characterized in that The acquiring of raw vehicle status data and smoothing of the raw data using a Kalman filter algorithm to obtain first processed data includes: Obtain vehicle speed, acceleration, and angular data collected by on-board sensors, establish a Kalman filter model, and set the process noise covariance matrix and the measurement noise covariance matrix; Construct the state transfer matrix, observation matrix and noise matrix of the Kalman filter; Inputting the raw data into the Kalman filter model, and recursively estimating the optimal value of the vehicle state at each moment through prediction and updating; In the prediction step, the state prior estimate and the error covariance matrix of the current moment are predicted based on the state estimate value and the state transfer matrix at the previous moment; In the update step, the Kalman gain is calculated and the predicted value is corrected using the current observation value to obtain the state posterior estimate and the updated error covariance matrix; Iterate the prediction and update steps until all the raw vehicle status data at all time points are processed; The vehicle state estimate after Kalman filtering is evaluated, and if the error exceeds a preset threshold, the process noise covariance matrix and the measurement noise covariance matrix parameters of the Kalman filter are adjusted.

3. The method according to claim 1, characterized in that Filling missing values using an interpolation algorithm based on the first processed data to obtain second processed data includes: Build a data model based on the distribution characteristics and change trends of existing data to predict the value range of missing data; For missing continuous values, linear interpolation method is used to estimate and fill; For missing categorical data, the nearest neighbor algorithm was used for interpolation; Perform outlier detection on the filled data, analyze the distribution characteristics of outliers, set thresholds, and remove abnormal data points introduced by the interpolation process; After removing outliers, the data size is expanded through data enhancement technology; For continuous values, new data samples are generated by adding random noise; For categorical data, a synthetic minority oversampling algorithm is used to synthesize new minority class samples to balance the category distribution of the dataset. Divide the expanded dataset into training and test sets; Different interpolation algorithms are applied to fill missing data, and the prediction performance of each algorithm on the training set and test set is compared and evaluated. The interpolation model with the best filling effect is selected and used to estimate all missing values in the original data set.

4. The method according to claim 1, wherein The step of extracting key state parameters from the second processed data and modifying the parameters in real time using a time series prediction model to obtain third processed data includes: Obtain key parameter values that reflect the operating status of the equipment; Using the obtained historical parameter value sequence, an autoregressive integrated moving average model is established; The established autoregressive integral moving average model is used to predict and correct the delayed state parameters at the current moment in real time according to the historical state parameter value sequence of the device, so as to obtain the real operating parameters of the device at the current moment; Substituting the corrected current state parameters into the autoregressive integral moving average model, continuously rolling forecasting the device state parameters in the subsequent period of time is performed in a sliding time window manner to obtain a forecast parameter sequence for the future period of time; According to the predicted equipment operating parameter sequence, a normal value range threshold is set, and whether the predicted parameter exceeds the threshold is determined; If the forecast parameters at a certain moment exceed the normal range, an early warning message will be issued; The real-time operating parameters of the equipment are obtained, and compared with the parameter sequence predicted by the autoregressive integral moving average model in real time to calculate the deviation value between the actual parameters and the predicted parameters.

5. The method according to claim 1, wherein The fourth processed data is obtained by dynamically evaluating the vehicle driving state using a multi-factor regression model based on the third processed data in combination with road condition information, weather forecast data, and historical driving records, including: The model calculates the current vehicle driving status evaluation score by weighting and comprehensively analyzing various factors; Comparing the vehicle driving status score calculated by the evaluation model with a preset safety threshold; If the score exceeds the threshold, the vehicle is judged to be in a dangerous driving state and the system automatically issues a safety warning to the driver; The vehicle driving status assessment score is compared and analyzed with the data in the vehicle's historical driving record database, and a clustering algorithm is used to identify whether there is an abnormal change trend in the vehicle's driving status.

6. The method according to claim 1, characterized in that The extracting of the vehicle position distribution, cargo loading, and remaining fuel information from the fourth processed data and dynamically adjusting the warning rules using an adaptive threshold algorithm to obtain fifth processed data includes: According to the vehicle location distribution information, a density-based clustering algorithm is used to divide the vehicles into regions to obtain the vehicle distribution density in different regions; For each area, obtain the cargo load and remaining fuel data of the vehicles in the area, remove outliers and redundant data, and obtain a normalized feature dataset; Based on the historical operation data of the vehicle, a judgment standard of the vehicle status is defined, and the feature data set is annotated to obtain the label data of the vehicle status; Using the label data and the feature data, a model of the association between the cargo load, the remaining fuel level, and the vehicle status is trained using a logistic regression algorithm to obtain parameters and thresholds of the model; During vehicle operation, the vehicle's cargo load and remaining fuel data are collected in real time to construct time series data; Using a Kalman filter algorithm to filter the time series data, and estimate the optimal values of the cargo load and the remaining fuel amount; The filtered cargo load and remaining fuel data are input into the correlation model to determine the vehicle status.

7. The method according to claim 1, characterized in that The method of performing secondary screening on the warning signal using a support vector machine classification algorithm based on the fifth processed data to obtain sixth processed data includes: Extract the characteristics of the warning signal, including signal strength, duration, and frequency, to form a warning signal feature vector; Using the labeled warning signal sample data and the support vector machine algorithm, the support vector machine classification model is trained by adjusting the kernel function and penalty coefficient parameters; For a new warning signal, its feature vector is input into the trained support vector machine model to perform a binary classification prediction; The support vector machine model will classify the warning signal into two categories: normal signal or abnormal signal; For warning signals predicted as abnormal signals by the support vector machine model, marking them as potential false positives or false negatives; Set a threshold. When the confidence level of an abnormal signal is lower than the threshold, the warning signal is determined to be a false alarm or missed alarm and is filtered out. retaining the warning signal with a confidence level higher than the threshold as sixth processed data; According to the classification result of the support vector machine model, a warning signal with a higher confidence level is screened out from the sixth processed data as a final effective warning signal.

8. The method according to claim 1, characterized in that The method of rapidly generating an optimal scheduling solution using a genetic algorithm based on the seventh processed data to obtain a final scheduling result includes: Initialize a set of scheduling schemes as the initial population. Each scheduling scheme contains the execution order of tasks and the allocated resources information. For each scheduling scheme in the population, the fitness value of the scheme is calculated by weighted summation based on the task priority, execution time and resource utilization factors, which is used as an indicator to evaluate the quality of the scheduling scheme; The population is selected, crossed, and mutated according to the fitness value to generate a new scheduling plan. The selection operation uses the roulette wheel selection method to select individuals from the current population according to the size of the fitness value. The larger the fitness value, the higher the probability of being selected. The crossover operation uses a two-point crossover method, randomly selecting two crossover points and exchanging the gene segments between the two selected individuals; The mutation operation uses the basic bit mutation method to invert certain genes of an individual with a certain probability and introduce new gene combinations; Determine whether the maximum fitness value of the current population has not changed significantly for multiple consecutive iterations, or has reached the preset maximum number of iterations; If the conditions are met, the scheduling scheme with the highest fitness value is output as the optimal scheme, the execution order of tasks and resource allocation are determined, and the final scheduling result is generated.

Images