Commercial vehicle intelligent connection system based on Beidou positioning

Abstract

The invention belongs to the technical field of commercial vehicles, and discloses a commercial vehicle intelligent connection system based on Beidou positioning. The sub-meter high-precision positioning of the electric vehicle is realized; the system carries out multi-dimensional state monitoring on a vehicle power system, a battery system and vehicle-mounted equipment, constructs a vehicle health index model, forms a space-time trajectory knowledge graph based on a big data mining technology, optimizes the energy efficiency and charging planning of the electric vehicle based on a dynamic energy management strategy network, and improves the energy efficiency of the electric vehicle. And an inter-vehicle collaborative perception and distributed decision-making mechanism is constructed. According to the invention, a multi-level vehicle association cooperation framework and an urban intelligent traffic system are deeply integrated, three-layer cooperation of vehicle individuals, vehicle groups and the urban traffic system is realized, an integrated intelligent travel platform is formed, the energy consumption is effectively reduced, and the traffic efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the field of commercial vehicle technology, and more specifically, to a commercial vehicle intelligent connection system based on Beidou positioning. Background Art

[0002] With the rapid development of intelligent transportation and Internet of Vehicles technology, the intelligence and networking of commercial vehicles have become an important trend in the development of the industry. However, the existing intelligent systems of commercial vehicles still have many limitations: insufficient positioning accuracy, making it difficult to cope with complex environments such as high-rise buildings and numerous tunnels in cities; incomplete vehicle status monitoring, unable to timely detect potential fault risks; low energy management efficiency, making it difficult to optimize the range and charging strategy of electric commercial vehicles; lack of group collaboration capabilities, failing to give full play to the overall advantages of the Internet of Vehicles; low integration with urban transportation systems, making it difficult to achieve seamless connection between commercial vehicles and urban infrastructure; limited data mining and knowledge extraction capabilities, failing to fully utilize massive vehicle data to support high-level decision-making; insufficient system adaptability and robustness, and poor performance in complex and changeable actual operating environments. These problems seriously restrict the operational efficiency and safety of commercial fleets, and also hinder the coordinated development of smart logistics and smart cities.

[0003] In view of this, the present invention proposes a commercial vehicle intelligent connection system based on Beidou positioning to solve the above problems. Summary of the invention

[0004] In order to overcome the above-mentioned defects of the prior art and to achieve the above-mentioned purpose, the present invention provides the following technical solution: a commercial vehicle intelligent connection system based on Beidou positioning, comprising: The data acquisition module is used to perform fusion processing and precision enhancement on the Beidou multi-modal positioning signals acquired by commercial vehicles to obtain high-precision spatiotemporal positioning data; Intelligent analysis module, used to collect and intelligently analyze the state parameters of the power system, battery system and on-board equipment of commercial vehicles in multiple dimensions to obtain the vehicle health index model; An intelligent mining module, based on the high-precision spatiotemporal positioning data and the vehicle health index model, performs commercial vehicle group behavior pattern mining to obtain a spatiotemporal trajectory knowledge graph; An energy optimization module, which is used to optimize the energy efficiency and charge planning of commercial vehicles by using the spatiotemporal trajectory knowledge graph to obtain a dynamic energy management strategy network; The multi-level vehicle connection module builds a collaborative perception and distributed decision-making mechanism between vehicles based on high-precision spatiotemporal positioning data and a dynamic energy management strategy network, and obtains a multi-level vehicle connection collaboration framework; the multi-level vehicle connection collaboration framework is integrated with the urban intelligent transportation system to obtain a comprehensive solution for commercial vehicle intelligent travel.

[0005] Furthermore, the fusion processing and precision enhancement of the Beidou multi-modal positioning signals obtained by the commercial vehicle to obtain high-precision spatiotemporal positioning data includes: Collecting Beidou navigation system multi-frequency signals and ground augmentation station differential signals to obtain an original positioning data set, and performing signal quality evaluation on the original positioning data set to obtain a signal reliability vector; Based on the signal reliability vector, weighted fusion is performed on multi-source positioning information to obtain a preliminary fused positioning result, and the positioning noise is suppressed by using a Kalman filter algorithm to obtain filtered positioning data; Combining the inertial navigation unit data, adaptively compensating the filtered positioning data to obtain a continuous positioning trajectory, and applying a trajectory prediction algorithm in the signal blocking area to obtain a complete positioning sequence; Performing map matching and constraint optimization on the complete positioning sequence to obtain road-level precise positioning data, and adaptively correcting the positioning accuracy in complex scenarios based on a deep learning model to obtain sub-meter positioning results; The sub-meter positioning results are subjected to spatiotemporal index construction and time stamp synchronization to obtain high-precision spatiotemporal positioning data.

[0006] Furthermore, the multi-dimensional collection and intelligent analysis of state parameters of the power system, battery system and on-board equipment of the commercial vehicle to obtain a vehicle health index model includes: Collecting power system parameters through a multi-source sensor network to obtain a power system state matrix, and performing abnormality detection and fault diagnosis on the power system state matrix to obtain a power system health vector; Monitor the key parameters of the battery pack to obtain the battery status data set, and evaluate the battery health status based on electrochemical impedance spectroscopy analysis technology to obtain the battery capacity decay model; Collecting the operating status data of the vehicle control unit, communication module and auxiliary equipment to obtain a set of equipment performance indicators, and constructing an equipment reliability evaluation network based on the equipment performance indicator set to obtain an equipment health score; Performing multi-level fusion on the power system health vector, the battery capacity attenuation model and the equipment health score to obtain a vehicle comprehensive state feature space, and applying clustering and classification algorithms in the vehicle comprehensive state feature space to obtain a vehicle state assessment result; Based on the vehicle status assessment results, a deep neural network prediction model is constructed to dynamically predict the remaining service life of key vehicle components, and combined with historical operating data to generate a vehicle health index model.

[0007] Furthermore, based on the high-precision spatiotemporal positioning data and the vehicle health index model, commercial vehicle group behavior pattern mining is performed to obtain a spatiotemporal trajectory knowledge graph, including: Performing trajectory segmentation and feature extraction on the high-precision spatiotemporal positioning data to obtain a vehicle movement pattern feature set, and performing trajectory clustering based on the vehicle movement pattern feature set to obtain a typical driving path library; Combining the vehicle health index model with the driving behavior data, establishing a correlation analysis model between driving habits and vehicle status, and obtaining a driving-vehicle status mapping matrix; Frequent pattern mining is performed on the spatiotemporal trajectory data of a large number of vehicles to identify high-frequency travel paths and time windows, obtain a spatiotemporal activity pattern set, and build an urban traffic flow prediction model based on the spatiotemporal activity pattern set to obtain a dynamic traffic flow distribution map; Use graph neural networks to model the interaction behaviors between vehicles, obtain the vehicle social network structure, identify key nodes and community structures in the vehicle social network structure, and obtain the vehicle group influence distribution; The typical driving path library, the driving-vehicle state mapping matrix, the dynamic traffic distribution map and the vehicle group influence distribution are integrated to construct a multi-level knowledge representation model and obtain a spatiotemporal trajectory knowledge graph.

[0008] Furthermore, the spatiotemporal trajectory knowledge graph is used to optimize the energy efficiency and charge planning of commercial vehicles, and a dynamic energy management strategy network is obtained, including: Based on the spatiotemporal trajectory knowledge graph, the energy consumption of commercial vehicle driving routes is analyzed to obtain a road section energy consumption feature library, and a multi-factor energy consumption prediction model is established in combination with terrain, weather and traffic condition data to obtain accurate energy consumption estimation results; Calculate the vehicle's cruising range based on the precise energy consumption estimation result and the battery status information to obtain a mileage prediction curve, and provide personalized energy-saving driving advice to the driver based on the mileage prediction curve to obtain a real-time driving optimization strategy; Using the distribution data of urban charging facilities and the real-time status information of charging stations, combined with the knowledge graph of spatiotemporal trajectories, an intelligent charging recommendation model is constructed to obtain the optimal charging station selection plan. Based on the grid load and electricity price fluctuation data, the optimal charging time period is recommended to obtain an economic charging strategy. Coordinated scheduling of multiple vehicle charging demands to obtain a balanced charging resource allocation plan, and optimized the charging resource sharing mechanism between vehicles based on the game theory model to obtain a collaborative charging network; The real-time driving optimization strategy, the economic charging strategy and the collaborative charging network are integrated to construct a hierarchical decision support system and obtain a dynamic energy management strategy network.

[0009] Furthermore, based on the high-precision spatiotemporal positioning data and the dynamic energy management strategy network, a vehicle-to-vehicle collaborative perception and distributed decision-making mechanism is constructed to obtain a multi-level vehicle-to-vehicle collaborative framework, including: Based on the vehicle wireless communication technology, a direct communication link between vehicles is established to obtain a vehicle communication topology network, and a data exchange protocol is defined according to the vehicle communication topology network to obtain an efficient information sharing mechanism; Integrate the vehicle's own sensor data and the perception information shared by surrounding vehicles to obtain enhanced perception results, and build a dynamic map of the local environment based on the enhanced perception results to obtain a shared situation model; According to the high-precision spatiotemporal positioning data, the relative position relationship between vehicles is calculated to obtain a set of conditions for forming a convoy, and based on the set of conditions for forming a convoy, an adaptive convoy organization algorithm is defined to obtain an intelligent convoy formation strategy; Combining the shared situation model and the dynamic energy management strategy network, a distributed decision-making mechanism is defined to enable each vehicle to make the global optimal decision based on local information, thus obtaining a collaborative decision-making framework; A fault-tolerant mechanism and a backup decision-making strategy are defined to obtain a robustness assurance solution, and the efficient information sharing mechanism, the intelligent vehicle fleet formation strategy, the collaborative decision-making framework and the robustness assurance solution are integrated to obtain a multi-level vehicle-connected collaborative framework.

[0010] A commercial vehicle intelligent connection device based on Beidou positioning, the commercial vehicle intelligent connection device based on Beidou positioning comprising: a memory and at least one processor, wherein the memory stores instructions; The at least one processor calls the instructions in the memory so that the commercial vehicle intelligent connection device based on Beidou positioning executes the commercial vehicle intelligent connection system based on Beidou positioning.

[0011] A computer-readable storage medium stores instructions, and when the instructions are executed by a processor, the commercial vehicle intelligent connection system based on Beidou positioning is implemented.

[0012] The technical effects and advantages of the commercial vehicle intelligent connection system based on Beidou positioning of the present invention are as follows: The present invention constructs a complete set of intelligent networking solutions for commercial vehicles by fusing and enhancing the Beidou multimodal positioning signals obtained by commercial vehicles, combined with comprehensive monitoring and analysis of vehicle status. This solution realizes the leap from single-vehicle intelligence to group collaboration, organically combining vehicle positioning, status monitoring, behavior analysis, energy management and collaborative control to form a closed-loop intelligent travel ecosystem. The system greatly improves the positioning accuracy, energy efficiency and travel experience of commercial vehicles through advanced technologies such as multi-source data fusion, machine learning, and distributed decision-making. At the same time, through deep integration with urban intelligent transportation systems, it promotes the coordinated development of commercial vehicles and urban transportation infrastructure. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1It is a schematic diagram of the commercial vehicle intelligent connection system based on Beidou positioning of the present invention. DETAILED DESCRIPTION

[0014] The embodiment of the present application provides a commercial vehicle intelligent connection system based on Beidou positioning. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described here can be implemented in an order other than that illustrated or described here. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0015] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 In the embodiments of the present application, an embodiment of a commercial vehicle intelligent connection system based on Beidou positioning includes: The data acquisition module is used to perform fusion processing and precision enhancement on the Beidou multi-modal positioning signals acquired by commercial vehicles to obtain high-precision spatiotemporal positioning data; It is understandable that the execution subject of this application can be a commercial vehicle intelligent connection system based on Beidou positioning, or an intelligent control unit or cloud service platform on a commercial vehicle, which is not limited here. The embodiment of this application is explained by taking the intelligent control unit on a commercial vehicle as the execution subject.

[0016] Specifically, the multi-frequency satellite signals provided by the Beidou navigation system, including navigation information of different frequencies such as B1, B2, and B3, are collected, and the differential signals provided by the ground enhancement station are received at the same time. These signals together constitute the original positioning data set. The quality of each signal in the original positioning data set is evaluated, including factors such as signal-to-noise ratio, multipath effect, and satellite geometric distribution, to obtain the signal reliability vector. Based on the signal reliability vector, the multi-source positioning information is weighted and fused, and the weight distribution is proportional to the signal reliability, so as to improve the positioning accuracy and obtain the preliminary fusion positioning result. The Kalman filter algorithm is used to suppress the noise in the preliminary fusion positioning result. The Kalman filter effectively removes the influence of random noise through two steps of prediction and update to obtain the filtered positioning data. Combined with the acceleration, angular velocity and other data of the vehicle-mounted inertial navigation unit (IMU), the filtered positioning data is adaptively compensated to compensate for the temporary loss of satellite signals or the decrease in accuracy, and a continuous positioning trajectory is obtained. In signal-blocked areas such as tunnels and high-rise buildings, a prediction algorithm based on historical trajectories is applied, combined with road network constraints, to infer the possible position of the vehicle and obtain a complete positioning sequence. The complete positioning sequence is matched with the high-precision electronic map, and the vehicle position is accurately matched to the actual road through the road constraint optimization algorithm to obtain road-level precise positioning data. For complex scenes such as urban canyons and multi-story elevated roads, a deep learning model is used to adaptively correct the positioning results. The model learns the historical positioning error pattern and performs real-time correction to obtain sub-meter positioning results. Finally, a spatiotemporal index structure is established for the sub-meter positioning results and synchronized with the Universal Time (UTC) to ensure that the spatiotemporal reference system of all vehicles is consistent and obtain high-precision spatiotemporal positioning data.

[0017] Intelligent analysis module, used to collect and intelligently analyze the state parameters of the power system, battery system and on-board equipment of commercial vehicles in multiple dimensions to obtain the vehicle health index model; Specifically, through the installation of multiple sensors on key components such as motors and electronic control units, power system parameters such as motor temperature, speed, torque, output power, efficiency, etc. are collected in real time to form a power system state matrix. The power system state matrix is ​​subjected to anomaly detection algorithms based on machine learning to identify potential fault modes and signs of performance degradation, and fault diagnosis is performed in combination with the expert knowledge base to obtain a power system health vector, which quantifies the health status of each component of the power system. The battery management system (BMS) monitors key parameters such as the cell voltage, current, temperature distribution, internal resistance change, and number of charge and discharge cycles of the battery pack to form a battery state data set. Based on the electrochemical impedance spectroscopy (EIS) analysis technology, the internal electrochemical state of the battery is evaluated by applying a small signal AC current and measuring the corresponding voltage response. The battery capacity attenuation model is established in combination with historical data. The model can accurately predict the remaining capacity and aging rate of the battery. Through the on-board diagnostic system (OBD) and the communication network, the operating status data of auxiliary equipment such as the on-board control unit, communication module, air conditioning system, and lighting system are collected, including indicators such as processor load, memory usage, communication quality, and power consumption, to form a set of equipment performance indicators. According to the equipment performance indicator set, the Bayesian network is used to build an equipment reliability assessment network. The network takes into account the interdependence and failure mode between each device to obtain the equipment health score. Through the deep fusion algorithm, the power system health vector, battery capacity decay model and equipment health score are integrated at multiple levels to construct the vehicle comprehensive state feature space. In this feature space, K-means clustering and random forest classification algorithms are applied to finely divide and evaluate the vehicle health status to obtain the vehicle status assessment results. Based on the evaluation results, a deep neural network prediction model containing LSTM (long short-term memory network) and attention mechanism is constructed. The model can learn the health status change trend of each component of the vehicle and dynamically predict the remaining service life of key components. Combined with the vehicle's historical operation data, maintenance records and data of the same model vehicle group, a comprehensive vehicle health index model is generated. The model intuitively represents the overall health status of the vehicle with a score of 0-100, and provides the health score of the subsystem and the expected maintenance time point.

[0018] An intelligent mining module, based on the high-precision spatiotemporal positioning data and the vehicle health index model, performs commercial vehicle group behavior pattern mining to obtain a spatiotemporal trajectory knowledge graph; Specifically, the high-precision spatiotemporal positioning data is segmented in time and space dimensions to extract features such as vehicle acceleration, deceleration, turning angle, and dwell time to form a vehicle movement pattern feature set. Based on this feature set, the DBSCAN (density-based spatial clustering applications and noise) algorithm is used to cluster trajectories and identify typical driving modes in different driving scenarios, such as urban commuting, high-speed cruising, and congested travel, to form a typical driving path library. Combining the vehicle health index model with driving behavior data from on-board sensors, including accelerator pedal position, braking force, steering angular velocity, etc., a driving habit and vehicle state association analysis model is established. This model quantifies the degree of influence of different driving habits on vehicle health status through multivariate regression analysis and generates a driving-vehicle state mapping matrix. The Apriori algorithm and sequential pattern mining technology are applied to the historical trajectory data of large-scale vehicle groups to identify high-frequency travel paths and time windows in the city, such as commuting routes during peak hours in the morning and evening, and visit patterns in weekend leisure areas, to obtain a spatiotemporal activity pattern set. Based on this pattern set, combined with the time series analysis method, a hybrid urban traffic flow prediction model based on ARIMA (autoregressive integrated moving average) and neural network is constructed to generate dynamic traffic flow distribution maps reflecting traffic flow in different time periods and different regions. Using graph neural network (GNN) technology, the interaction behaviors between vehicles (such as appearing at the same location at the same time, similar driving paths, etc.) are modeled as a graph structure, where nodes represent vehicles and edges represent the interaction relationship between vehicles, and the vehicle social network structure is obtained. In this network structure, centrality analysis and community detection algorithms are applied to identify key vehicle nodes with high influence and closely connected vehicle communities, and the influence distribution of vehicle groups is obtained. The typical driving path library, driving-vehicle state mapping matrix, dynamic flow distribution map and vehicle group influence distribution are integrated, and a multi-level knowledge representation model is constructed using knowledge graph technology. The model is stored in the form of a graph database, containing entities (vehicles, drivers, road sections, time windows, etc.), attributes and relationships, supporting complex queries and reasoning, and forming a spatiotemporal trajectory knowledge graph.

[0019] An energy optimization module, which is used to optimize the energy efficiency and charge planning of commercial vehicles by using the spatiotemporal trajectory knowledge graph to obtain a dynamic energy management strategy network; Specifically, based on the spatiotemporal trajectory knowledge graph, the energy consumption data of commercial vehicles in different sections of the city at different times are statistically analyzed, and the road section energy consumption feature library is established by considering factors such as road conditions, slope, speed limit, etc. Combined with meteorological data (temperature, humidity, wind speed), real-time traffic conditions and vehicle load information, machine learning methods such as gradient boosting decision tree (GBDT) are applied to construct a multi-factor energy consumption prediction model. By inputting the current road conditions and environmental conditions, the model can accurately estimate the energy consumption of future trips and obtain accurate energy consumption estimation results. According to the accurate energy consumption estimation results and the battery status information (SOC, SOH, etc.) provided by the battery management system, the physical model is combined with the data-driven method to calculate the vehicle's endurance under different driving modes and environmental conditions, and generate a mileage prediction curve. Based on the prediction curve, combined with the driver's historical driving habits data, a personalized energy-saving driving recommendation system is developed. The system can provide optimal speed recommendations, acceleration timing prompts and energy recovery strategies during driving to form a real-time driving optimization strategy. By using the geographical distribution data of urban charging infrastructure and the real-time operation status information of charging stations (number of idle charging piles, queuing status, charging rate, etc.), combined with the vehicle location and travel plan in the spatiotemporal trajectory knowledge graph, an intelligent charging recommendation system is constructed. The system adopts a multi-objective optimization algorithm, comprehensively considers factors such as distance, charging time, waiting time, etc., recommends the most suitable charging station, and obtains the optimal charging station selection plan. Combined with the grid load data and time-of-use electricity price policy, the charging cost differences in different time periods are analyzed, and the charging time recommendation with the best economic efficiency is generated to form an economic charging strategy. By constructing a regional charging demand prediction model, the spatiotemporal distribution analysis of the charging demand of multiple vehicles is carried out to achieve the coordinated scheduling of charging resources. The scheduling system can avoid charging resource congestion in specific time and area through reservation mechanism and real-time guidance, and obtain a balanced charging resource allocation plan. Based on the game theory model, the charging resource sharing mechanism between vehicles is defined, including the private charging pile sharing pricing model, charging rights and interests trading rules and mutual credit system, to form a collaborative charging network. Integrate real-time driving optimization strategy, economic charging strategy and collaborative charging network to build a hierarchical decision support system including perception layer, decision layer and execution layer. The system can adaptively adjust energy management strategies based on dynamic factors such as vehicle status, driving habits, energy prices, and charging infrastructure distribution to form a dynamic energy management strategy network.

[0020] The multi-level vehicle connection module builds a collaborative perception and distributed decision-making mechanism between vehicles based on high-precision spatiotemporal positioning data and a dynamic energy management strategy network, and obtains a multi-level vehicle connection collaboration framework. The multi-level vehicle connection collaboration framework is integrated with the urban intelligent transportation system to obtain a comprehensive solution for commercial vehicle intelligent travel. Specifically, based on wireless communication technologies such as DSRC (dedicated short-range communication) and C-V2X (cellular vehicle-to-everything), direct information exchange between vehicles is realized, and a dynamic vehicle communication topology network is established. The network can adapt to changes in vehicle density and communication environment and maintain stable connectivity. According to the characteristics of the vehicle communication topology network, a lightweight, low-latency data exchange protocol is defined, including message format definition, transmission priority strategy and congestion control mechanism, to form an efficient information sharing mechanism. By integrating the vehicle's own radar, camera, lidar and other sensor data, as well as the perception information shared by surrounding vehicles through V2V (vehicle-to-vehicle) communication, the perception range of a single vehicle is expanded, the perception blind spot is eliminated, and environmental collaborative perception is achieved. Sensor data fusion algorithms, including Kalman filtering, particle filtering and deep learning methods, are applied to suppress noise and process consistency of multi-source perception data to obtain enhanced perception results. Based on the enhanced perception results, a local environment dynamic map containing elements such as roads, vehicles, pedestrians and obstacles is constructed. The map represents the environmental state in a unified coordinate system and can be updated in real time, supporting context sharing between vehicles to form a shared context model. According to the high-precision spatiotemporal positioning data, the relative position, relative speed and driving direction between adjacent vehicles are calculated. Combined with the vehicle intention inference results, the feasibility of forming a convoy is evaluated and the convoy formation condition set is obtained. Based on the convoy formation condition set, an adaptive convoy organization algorithm that takes into account road conditions, vehicle performance and traffic rules is defined. The algorithm can dynamically determine the convoy size, vehicle spacing and formation form, support the splitting and merging operations of the convoy, and form an intelligent convoy formation strategy. Combined with the shared situation model and the dynamic energy management strategy network, a distributed decision-making mechanism based on the consistency algorithm is defined. This mechanism enables each vehicle to reach a consistent decision based on local information and information shared by neighboring vehicles in accordance with common decision-making rules without relying on central control, forming a collaborative decision-making framework. For abnormal situations such as communication interruption, node failure and malicious attack, a fault-tolerant mechanism including redundant communication path, decision degradation strategy and security authentication mechanism is defined. At the same time, a backup decision strategy based on local information is pre-configured to ensure that basic functions can be maintained when part of the collaborative system fails, forming a robustness guarantee solution. Integrate efficient information sharing mechanism, intelligent fleet formation strategy, collaborative decision framework and robustness assurance solution to build a multi-level vehicle-connected collaboration framework including communication layer, perception layer, decision layer and security layer. The framework supports different levels of vehicle-connected collaboration, from simple information sharing to complex collaborative decision-making, and adapts to application scenarios under different penetration rates and technical conditions.

[0021] In a specific embodiment, the execution process of the data acquisition module specifically includes the following steps: (1) Collecting Beidou navigation system multi-frequency signals and ground augmentation station differential signals to obtain an original positioning data set, and performing signal quality assessment on the original positioning data set to obtain a signal reliability vector; (2) Based on the signal reliability vector, weighted fusion is performed on multi-source positioning information to obtain a preliminary fused positioning result, and the positioning noise is suppressed by using the Kalman filter algorithm to obtain filtered positioning data; (3) Adaptively compensating the filtered positioning data in combination with the inertial navigation unit data to obtain a continuous positioning trajectory, and applying a trajectory prediction algorithm in the signal-blocked area to obtain a complete positioning sequence; (4) performing map matching and constraint optimization on the complete positioning sequence to obtain road-level precise positioning data, and adaptively correcting the positioning accuracy in complex scenarios based on a deep learning model to obtain sub-meter positioning results; (5) Constructing a spatiotemporal index and synchronizing the timestamp of the sub-meter positioning results to obtain high-precision spatiotemporal positioning data.

[0022] Specifically, the Beidou receiver on the commercial vehicle simultaneously collects signals from multiple frequency points such as B1 (1561.098MHz), B2 (1207.14MHz) and B3 (1268.52MHz) of the Beidou satellite navigation system, and receives differential correction signals from the ground augmentation station network. The differential signal contains correction parameters such as satellite orbit error, clock error and atmospheric delay, which can significantly improve positioning accuracy. These signals together constitute the original positioning data set D containing information such as satellite pseudorange, carrier phase, Doppler frequency shift, etc. The quality of each signal in the original positioning data set D is evaluated, and quality indicators such as signal-to-noise ratio (SNR), multipath effect index, and geometric precision factor (GDOP) are calculated to form a signal quality evaluation matrix Q. Based on the matrix Q, the fuzzy comprehensive evaluation method is used to calculate the reliability weight of each signal source to obtain the signal reliability vector ; Where W is the evaluation index weight vector determined by expert experience. Based on the signal reliability vector R, the multi-source positioning information is weighted and fused. The weight is proportional to the signal reliability. The fusion formula is as follows: ; Where P_i represents the position information provided by the i-th signal source, is the signal reliability vector of the i-th signal source, and the preliminary fusion positioning result P_fusion is obtained. The random noise in the preliminary fusion positioning result is suppressed by the Kalman filter algorithm. The Kalman filter includes two steps: prediction and update: Prediction steps: ; Update steps: ; Where X_k represents the state vector at time k (including position, speed and other information), A is the state transfer matrix, B is the control matrix, u_k is the control vector, Z_k is the observation vector, and H is the observation matrix. is the transfer value of the state vector at time k; K_k is the Kalman gain. Through repeated iterative prediction and update steps, the filtered positioning data is obtained. Combined with the acceleration and angular velocity data provided by the vehicle-mounted inertial navigation unit (IMU), the filtered positioning data is adaptively compensated. According to the IMU data, attitude solution and motion integration are performed to obtain the displacement estimate from the last valid positioning point to the current moment, compensating for the possible temporary loss of satellite signals or decreased accuracy. The calculation formula is: ; Where P_last represents the last valid positioning point, a(t) represents the acceleration function, To compensate for the effective positioning point, the displacement estimate is obtained by quadratic integration of acceleration to form a continuous positioning trajectory T_con. In signal-blocked areas such as tunnels and high-rise buildings, a prediction algorithm based on historical trajectory characteristics and road network constraints is applied. The kinematic characteristics of the vehicle's historical trajectory and the road topology are used to predict the possible position of the vehicle in the signal-missing area. The prediction algorithm includes: The trajectory prediction model trained based on historical data and the application of road constraints; through the above prediction algorithm, the position of the signal-blocked area is calculated to obtain the complete positioning sequence T_comp. The complete positioning sequence is matched with a high-precision electronic map, and the hidden Markov model (HMM) and other map matching algorithms are applied to accurately match the vehicle position to the actual road. The map matching process takes into account the position measurement error, road network topology and vehicle motion constraints, and obtains the road-level precise positioning data P_road by maximizing the likelihood probability: For complex scenes such as urban canyons and multi-story elevated roads, a deep learning model including convolutional neural networks (CNN) and recurrent neural networks (RNN) is used to adaptively correct positioning results. The model learns the relationship between error patterns and environmental features in a large amount of historical positioning data, establishes an error compensation model, and achieves real-time correction: P_corrected=P_road-CNN-RNN(Env); Among them, P_corrected is the corrected data of the road-level precise positioning data; Env includes environmental features such as the number of visible satellites, building height, and road type. Through correction, sub-meter positioning results are obtained. Finally, a quadtree or grid-type spatiotemporal index structure is established for the sub-meter positioning results, and it is synchronized with the Universal Time (UTC) with high precision to ensure that the spatiotemporal reference systems of all vehicles are consistent, and high-precision spatiotemporal positioning data is obtained. The data contains accurate location coordinates, timestamps, accuracy assessments, and credibility indicators.

[0023] In a specific embodiment, the execution process of the intelligent analysis module specifically includes the following steps: (1) Collecting power system parameters such as motor temperature, speed, and output power through a multi-source sensor network to obtain a power system state matrix, and performing abnormality detection and fault diagnosis on the power system state matrix to obtain a power system health vector; (2) Monitor key parameters of the battery pack, such as voltage, current, temperature, and number of cycles, to obtain a battery status data set. The battery health status is evaluated based on electrochemical impedance spectroscopy analysis technology to obtain a battery capacity decay model. (3) Collecting the operating status data of the vehicle control unit, communication module and auxiliary equipment to obtain a set of equipment performance indicators, and constructing an equipment reliability evaluation network based on the equipment performance indicator set to obtain an equipment health score; (4) performing multi-level fusion of the power system health vector, the battery capacity attenuation model, and the equipment health score to obtain a vehicle comprehensive state feature space, and applying clustering and classification algorithms in the feature space to obtain a vehicle state assessment result; (5) Based on the vehicle status assessment results, a deep neural network prediction model is constructed to dynamically predict the remaining service life of key vehicle components, and combined with historical operating data to generate a vehicle health index model.

[0024] Specifically, by installing temperature sensors, Hall effect sensors, current sensors, vibration sensors and other sensors on key power system components such as motors, motor controllers, and reducers, power system parameters such as motor temperature, stator current, speed, torque, output power, efficiency, noise and vibration spectrum are collected in real time. These parameters are organized into a power system state matrix M according to time series. Each row of the matrix represents a time point and each column represents a parameter. Principal component analysis (PCA) is applied to the power system state matrix M to reduce the dimension and extract key features. Then, an anomaly detection method based on the isolation forest algorithm is used to identify potential anomalies. The mathematical expression for identifying potential anomalies is: ;in is the abnormal score of sample x in the total number of samples, h(x) is the path length of sample x, c(n) is the normalization factor, is the total number of samples, is the abbreviation of expected path length. It represents the average path length of sample x from the root node to the leaf node in the process of constructing the isolation tree multiple times. For the detected anomalies (anomaly score greater than the threshold), fault diagnosis is performed in combination with the knowledge-based reasoning system and case-based reasoning (CBR) to identify the specific fault type and severity. According to the diagnosis results, a power system health vector H_power is constructed, which contains the health score, expected service life and maintenance recommendations of each key component. The battery management system (BMS) is used to monitor the key parameters of the battery pack in real time, such as single cell voltage, charge and discharge current, temperature distribution, internal resistance change, number of charge and discharge cycles, and number of deep charge and discharge cycles. These parameters together constitute the battery state data set B. Using electrochemical impedance spectroscopy (EIS) analysis technology, a small signal AC current (usually in the frequency range of mHz to kHz) is applied to the battery at different states of charge (SOC), the corresponding voltage response is measured, and the impedance spectrum Z is calculated. ; Where |Z| is the impedance amplitude, φ is the phase angle, is the frequency, The impedance spectrum is fitted through an equivalent circuit model (such as the Randles model) to extract the internal parameters of the battery, including ohmic impedance, charge transfer impedance, diffusion impedance, etc. These parameters are closely related to the aging state and health state of the battery. Combined with historical charge and discharge data and impedance parameter change trends, a battery capacity attenuation model is established. : ; Where C_0 is the initial capacity, k is the attenuation coefficient, f is the attenuation function (exponential function, etc.), N is the number of cycles, DOD is the depth of discharge, T is the temperature, and I is the charge and discharge current. This model can accurately predict the remaining capacity and aging rate of the battery under different usage conditions. Through the on-board diagnostic system (OBD) and the on-board Ethernet, the operating status data of auxiliary equipment such as the on-board control unit (ECU), communication module (4G / 5G, DSRC), air conditioning system, lighting system, and infotainment system are collected, including indicators such as processor load, memory usage, communication quality, response time, power consumption, and error log to form a device performance indicator set P. According to the device performance indicator set P, the Bayesian network is used to construct a device reliability assessment network. The Bayesian network is a probabilistic graph model in which the dependencies between variables are represented by a directed acyclic graph (DAG), and each node is associated with a conditional probability table (CPT): P(X_i|Parents(X_i)); Where X_i represents the device state variable, and Parents(X_i) represents the set of parent node variables that affect X_i. Through Bayesian reasoning, the probability distribution of the health state of the device under given observation evidence is calculated to obtain the device health score. Through the deep fusion algorithm, the power system health vector H_power, the battery capacity decay model C(t) and the device health score are integrated at multiple levels. First, the feature vectors are spliced; then the decision-level fusion is performed, and the decision results of each subsystem are integrated through weighted voting or to construct the vehicle comprehensive state feature space. The K-means++ clustering algorithm is applied to the vehicle comprehensive state feature space to divide the vehicle state into multiple categories. Then, the random forest classification algorithm is used to accurately classify and evaluate the current state of the vehicle based on historical data and expert annotations to obtain the vehicle state evaluation result. Based on the vehicle state evaluation result, a deep neural network prediction model containing LSTM (long short-term memory network) and Transformer attention mechanism is constructed. The core structure of LSTM includes input gate, forget gate and output gate. The model dynamically predicts the remaining service life RUL_i of key components (such as motors, batteries, transmissions, controllers, etc.) by learning the historical change trends of the health status of various vehicle components. Combined with the vehicle's historical operation data, maintenance records and data of the same model vehicle group, the transfer learning method is applied to build a comprehensive vehicle health index model ; ; Among them, w_i is the weight coefficient, f_i is the evaluation function (can be a product function), S_i is the component health status (quantitative value), RUL_i is the remaining service life, and Imp_i is the component importance. The health index model intuitively represents the overall health status of the vehicle with a score of 0-100, and provides the health score of the subsystem and the expected maintenance time point, providing a scientific basis for vehicle maintenance and use decisions.

[0025] In a specific embodiment, the execution process of the intelligent mining module 103 specifically includes the following steps: (1) performing trajectory segmentation and feature extraction on the high-precision spatiotemporal positioning data to obtain a vehicle movement pattern feature set, and performing trajectory clustering based on the vehicle movement pattern feature set to obtain a typical driving path library; (2) combining the vehicle health index model and driving behavior data, establishing a driving habit and vehicle state correlation analysis model to obtain a driving-vehicle state mapping matrix; (3) Conduct frequent pattern mining on the spatiotemporal trajectory data of a large number of vehicles to identify high-frequency travel paths and time windows, obtain a spatiotemporal activity pattern set, and build an urban traffic flow prediction model based on the spatiotemporal activity pattern set to obtain a dynamic traffic flow distribution map; (4) Use a graph neural network to model the interaction behavior between vehicles, obtain a vehicle social network structure, and identify key nodes and community structures in the vehicle social network structure to obtain the vehicle group influence distribution; (5) Integrate the typical driving path library, the driving-vehicle state mapping matrix, the dynamic traffic distribution map, and the vehicle group influence distribution to construct a multi-level knowledge representation model and obtain a spatio-temporal trajectory knowledge graph.

[0026] Specifically, the high-precision spatio-temporal positioning data T is segmented according to time windows and spatial features, and a piecewise linear approximation (PLA) or a piecewise method based on important inflection points is applied to divide the continuous trajectory into a series of driving segments. Feature vectors are extracted for each driving segment, including: Kinetic features: average speed, maximum speed, acceleration distribution, deceleration distribution, driving stability; Geometric features: trajectory curvature, turning angle, path complexity; Spatio-temporal features: driving time, driving distance,停留 time, time regularity; These features together constitute the vehicle movement pattern feature set. Trajectory clustering is performed based on the vehicle movement pattern feature set. The parameters of trajectory clustering include radius ε and minimum number of points Pts. The process of trajectory clustering is as follows: For any unvisited point p, mark it as visited; Search the ε-neighborhood N_ε(p) of p; if |N_ε(p)| < Pts, mark p as a noise point; otherwise create a new cluster C and add p to C; for each point q in C, repeat the above process to expand cluster C; Through trajectory clustering, typical driving patterns in different driving scenarios, such as urban commuting, highway cruising, congested travel, etc., are identified to form a typical driving path library. Each path in the typical driving path library contains information such as path features, usage frequency, and applicable scenarios. Combining the vehicle health index model HI and driving behavior data from in-vehicle sensors, including throttle pedal position, braking force, steering angular velocity, shift frequency, etc., an association analysis model between driving habits and vehicle states is established. Apply multiple linear regression and regularization methods (such as Ridge regression or Lasso regression) to estimate the parameters of the association analysis model: ; where is the driving behavior feature, is the regression coefficient, is the error term. By calculating the influence coefficient of each driving behavior feature on the vehicle health index, the influence of different driving habits on the vehicle health status is quantified, and the driving-vehicle state mapping matrix M_map is generated. This matrix represents the influence weight of different driving behavior patterns on the health status of each vehicle system. Sequential pattern mining techniques, such as the PrefixSpan algorithm or the GSP (generalized sequential pattern) algorithm, are applied to the historical trajectory data of large-scale vehicle groups (thousands to tens of thousands of vehicles) to identify frequently occurring spatiotemporal sequence patterns. Frequently occurring spatiotemporal sequence patterns are defined as sequences whose support is not less than the minimum support threshold sup: Through mining, high-frequency travel routes (such as commuting routes, commercial area access routes) and time windows (such as morning and evening peaks, weekend leisure time) in the city are identified to obtain the spatiotemporal activity pattern set P. Based on the spatiotemporal activity pattern set P, combined with the time series analysis method, a hybrid prediction model is constructed. This model combines the ARIMA (autoregressive integrated moving average) model to capture linear time dependencies and the neural network to capture nonlinear patterns: The hybrid prediction model is used to predict traffic flow in different time periods and different areas, and a dynamic traffic distribution map is generated. The dynamic traffic distribution map intuitively displays the distribution of urban traffic flow in the form of a heat map. Using the graph neural network (GNN), the interaction behavior between vehicles is modeled as a graph structure G=(V, E, A), where V is the node set (vehicles), E is the edge set (interaction relationship between vehicles), and A is the adjacency matrix. The weight of the edge is obtained based on the weighted sum of the frequency, duration and similarity of vehicle interactions; The graph neural network updates the node representation through the message passing mechanism. By training the graph neural network, the vehicle social network structure is obtained. In the vehicle social network structure, key nodes with high influence are identified. The centrality indicators are defined including degree centrality, betweenness centrality and eigenvector centrality; Apply the Louvain algorithm or the InfoMap algorithm for community detection to identify closely connected groups of vehicles: Initialization: Each node is a community; Local movement: moving the node to the adjacent community where the modularity gain is maximized; Community aggregation: merging nodes belonging to the same community into super nodes; Repeat the steps until modularity no longer increases significantly; modularity is the weighted sum of centrality indicators; Then, the vehicle group influence distribution is obtained, which reflects the status and influence of the vehicle in the social network. The typical driving path library, the driving-vehicle state mapping matrix M_map, the dynamic traffic distribution map and the vehicle group influence distribution are integrated, and a multi-level knowledge representation model is constructed using the knowledge graph: G_kn=(E, R, T) consists of entity set E, relationship set R and triple set T, where: The entity set E includes vehicles, drivers, road sections, time windows, charging stations, etc.; The relation set R includes semantic relations such as “travels on”, “belongs to”, “affects”, and “is located in”; The set of triples represents the relationship between entities; The multi-level knowledge representation model is stored in a graph database (such as Neo4j), supports query languages ​​such as SPARQL, and can perform complex semantic queries and reasoning. Through knowledge representation learning, such as TransE, ComplEx and other algorithms, entities and relationships are embedded into a low-dimensional vector space; A spatiotemporal trajectory knowledge graph is formed, which integrates multi-source heterogeneous data, provides a unified knowledge representation of vehicle behavior patterns, and provides a data basis for subsequent energy optimization and collaborative decision-making.

[0027] In a specific embodiment, the execution process of the energy optimization module specifically includes the following steps: (1) Based on the spatiotemporal trajectory knowledge graph, the energy consumption of commercial vehicle driving routes is analyzed to obtain a road section energy consumption feature library. In addition, a multi-factor energy consumption prediction model is established in combination with terrain, weather and traffic condition data to obtain accurate energy consumption estimation results; (2) calculating the vehicle's cruising range based on the precise energy consumption estimation result and the battery status information, obtaining a mileage prediction curve, and providing personalized energy-saving driving advice to the driver based on the mileage prediction curve to obtain a real-time driving optimization strategy; (3) Using the urban charging facility distribution data and the real-time status information of charging stations, combined with the spatiotemporal trajectory knowledge graph, an intelligent charging recommendation system is constructed to obtain the optimal charging station selection plan. Based on the grid load and electricity price fluctuation data, the optimal charging time period is recommended to obtain an economic charging strategy. (4) Coordinated scheduling of charging demands of multiple vehicles to avoid competition for local charging resources, obtain a balanced charging resource allocation plan, and optimize the charging resource sharing mechanism between vehicles based on the game theory model to obtain a collaborative charging network; (5) Integrate the real-time driving optimization strategy, the economic charging strategy, and the collaborative charging network to construct a hierarchical decision support system to obtain a dynamic energy management strategy network.

[0028] Specifically, based on the spatiotemporal trajectory knowledge graph, the commercial vehicle energy consumption data of each road section s in the urban road network at different time periods t is extracted. Through statistical analysis, the statistical characteristics of the road section energy consumption, such as the mean, variance, and quantile, are calculated, and the physical properties of the road section (length, slope, curvature, etc.) and traffic characteristics (speed limit, number of traffic lights, etc.) are associated to build a road section energy consumption feature library. Combined with the terrain data provided by the digital elevation model (DEM), the weather data (temperature, humidity, wind speed, precipitation, etc.) provided by the meteorological department, and real-time traffic condition data (average speed, congestion level, etc.), machine learning methods such as the gradient boosting decision tree (GBDT) are used to construct a multi-factor energy consumption prediction model. The multi-factor energy consumption prediction model gradually improves the prediction performance by integrating multiple weak learners (decision trees): ; in is the multi-factor energy consumption prediction model for the mth iteration, is the multi-factor energy consumption prediction model for the m-1th iteration, is the mth decision tree, is the step size of the mth iteration, is the input of the model. The training objective of the multi-factor energy consumption prediction model is to minimize the loss function: ; Where E_true is the actual energy consumption, E_pred is the predicted energy consumption, is the regularization parameter, is a regularization term. After training, the model can accurately estimate the energy consumption of future trips based on the input route planning, current environmental conditions and vehicle status, and obtain the accurate energy consumption estimation result E_es. Based on the accurate energy consumption estimation result E_es and the battery status information (SOC, SOH, etc.) provided by the battery management system (BMS), a hybrid model combining the physical battery model and the data-driven method is applied to calculate the vehicle endurance. The physical battery model part considers the battery capacity, discharge efficiency and temperature effects: ; Where C_ava is the available power, C_total is the total capacity, f(T) is the temperature effect function (Gaussian function of temperature T, etc.), and η_dis is the discharge efficiency. The data-driven method uses historical data to learn the energy consumption characteristics under different working conditions: ; in is a correction function (which can be a weighted function) taking into account the driving mode ,terrain and traffic conditions The impact of is the mileage. Through calculation, a mileage prediction curve is generated, which represents the expected mileage under different speeds and road conditions. Based on the mileage prediction curve and combined with the driver's historical driving habits data, a personalized energy-saving driving recommendation model S_eco is obtained. This model analyzes the difference between the current driving behavior and the optimal energy-efficient driving mode in real time, and uses reinforcement learning methods to optimize the driving strategy: ; in is the value (value) of the updated state-action value function, is the value of the state-action value function before the update, is the learning rate, r is the immediate reward (related to energy efficiency), γ is the discount factor, The value generated by the action taken for the new state. S_eco provides real-time suggestions to the driver, including the best speed, best acceleration, energy recovery strategy, etc., to form a real-time driving optimization strategy. Using the geographical distribution data of the city's charging infrastructure and the real-time operating status information of the charging station (the number of idle charging piles, queue conditions, charging rate, etc.), combined with the vehicle location and trip plan in the spatiotemporal trajectory knowledge graph, a smart charging recommendation model is constructed. The smart charging recommendation model uses a multi-objective optimization algorithm and considers multiple factors: Distance cost: d(v,c) is the distance from vehicle v to charging station c; Time cost: t_wait(c)+t_charge(v,c) is the waiting time and charging time; Price cost: p(c, t) is the electricity price of charging station c at time t; Facility quality: q(c) is the charging station facility quality score; The optimization objective function is to minimize the weighted sum of the above factors; Dynamically adjust according to user preferences. By solving the optimization problem, the optimal charging station selection scheme is obtained. Combined with the grid load data and time-of-use electricity price policy, the cost-effectiveness of different charging periods is analyzed. The dynamic programming algorithm is used to solve the optimal charging time arrangement: f[i]=min{f[i-1]+cost(i), f[i-2]+cost(i-1, i),...}; Where f[i] represents the minimum charging cost in the first i time periods, and cost(i) represents the cost of charging in the i-th time period. Through optimization, the most economical charging time recommendation is generated to form an economic charging strategy. By constructing a regional charging demand prediction model M_demand, the spatiotemporal distribution of charging demand for multiple vehicles is analyzed. The prediction model is based on historical data and the current vehicle status, and uses time series prediction methods (such as ARIMA or Prophet) to predict the distribution of charging demand in future time periods. Based on the prediction results, the charging demand balance scheduling algorithm A_balance is defined, which guides vehicles to charge at off-peak times through price incentives and time window allocation to avoid congestion of charging resources at specific times and regions. The constrained optimization problem solved by the algorithm is: ; ; Where D_t is the charging demand in period t, μ is the average demand level, and D_total is the total demand. Through algorithm optimization, a balanced charging resource allocation scheme is obtained. Based on the game theory model, a charging resource sharing mechanism between vehicles is defined. This mechanism regards private charging pile owners and charging demand parties as game participants, and realizes efficient allocation of resources through incentive mechanisms. The shared pricing model is based on demand elasticity and cost structure: ; Where p is the sharing price, c is the basic cost, Δ is the price adjustment function, D is the demand, and S is the supply. At the same time, a credit evaluation system is defined to record the historical behavior of participants, provide discounts to high-credit users, and form a collaborative charging network. Integrate real-time driving optimization strategy, economic charging strategy and collaborative charging network to build a hierarchical decision support model, which contains three levels: Perception layer: collects vehicle status, environmental conditions, and user preferences; Decision-making layer: Make decisions based on multi-objective optimization and reinforcement learning; Execution layer: converts decisions into specific control instructions and user suggestions; The model adopts a hierarchical reinforcement learning framework. The upper-level strategy sets goals and the lower-level strategy performs specific tasks. It can adaptively adjust the energy management strategy according to dynamic factors such as vehicle status, driving habits, energy prices, and charging infrastructure distribution, forming a dynamic energy management strategy network to provide commercial vehicle users with comprehensive energy optimization solutions.

[0029] In a specific embodiment, the execution process of the multi-level vehicle connection module 105 specifically includes the following steps: (1) Based on the vehicle wireless communication technology, a direct communication link between vehicles is established to obtain a vehicle communication topology network, and a data exchange protocol is defined based on the vehicle communication topology network to obtain an efficient information sharing mechanism; (2) Integrate the vehicle’s own sensor data and the perception information shared by surrounding vehicles to achieve collaborative environmental perception and obtain enhanced perception results. Based on the enhanced perception results, a dynamic map of the local environment is constructed to obtain a shared situation model. (3) calculating the relative position relationship between vehicles based on the high-precision spatiotemporal positioning data to obtain a convoy formation condition set, and defining an adaptive convoy organization algorithm based on the convoy formation condition set to obtain an intelligent convoy formation strategy; (4) Combining the shared situation model and the dynamic energy management strategy network, defining a distributed decision-making mechanism so that each vehicle can make a global optimal decision based on local information, thereby obtaining a collaborative decision-making framework; (5) In response to communication interruption and node failure, a fault-tolerant mechanism and a backup decision-making strategy are defined to obtain a robustness assurance solution. The efficient information sharing mechanism, the intelligent vehicle fleet formation strategy, the collaborative decision-making framework, and the robustness assurance solution are integrated to obtain a multi-level vehicle-connected collaborative framework.

[0030] Specifically, based on wireless communication technologies such as DSRC (dedicated short-range communication, 5.9GHz frequency band) and C-V2X (cellular vehicle-to-everything, based on 4G / 5G technology), a multi-mode communication system supporting vehicle-to-vehicle (V2V), vehicle-to-infrastructure (V2I) and vehicle-to-cloud (V2C) is established. The DSRC communication module uses the IEEE 802.11p protocol to provide low-latency communication with a bandwidth of 10MHz, a data rate of 6Mbps and a communication range of about 300m, which is suitable for safety-critical applications; the C-V2X communication module uses 3GPP Release14 and above standards to provide a wider coverage and higher reliability, which is suitable for non-safety-critical applications. The system dynamically selects the communication mode that best suits the current scenario to form an adaptive hybrid communication architecture. Based on the vehicle location and communication connection status, a vehicle communication topology network G_co=(V, E, W) is constructed, where V is the vehicle node set, E is the communication connection set, and W is the connection quality weight set. The connection quality weight w_ij is calculated based on the signal strength (RSSI), signal-to-noise ratio (SNR), packet loss rate (PER) and delay, which can be a weighted sum; through periodic topology updates, the real-time and accurate communication network status is maintained. According to the characteristics of the vehicle communication topology network, a lightweight, low-latency data exchange protocol is defined. The protocol includes: Message format definition: Defines a standardized message structure based on ASN.1 (Abstract Syntax Notation), including fields such as message type, timestamp, sender ID, priority, and payload; Transmission priority strategy: Assign priorities based on the urgency and importance of the message content, such as collision warning > brake warning > road condition information > comfort application; Congestion control mechanism: Adaptively adjust transmission parameters based on channel busyness (CBR), including transmission power, transmission rate, and message generation rate; Security authentication mechanism: Use the Elliptic Curve Digital Signature Algorithm (ECDSA) to ensure message authenticity and integrity; The protocol implements an efficient and reliable information sharing mechanism M_sharing, which supports low-latency transmission of safety-critical information and high-throughput transmission of non-critical information. It integrates the vehicle's own sensor data (including radar, camera, lidar, etc.) and the perception information shared by surrounding vehicles through V2V communication to expand the perception range and eliminate blind spots. Apply the Bayesian sensor fusion framework for multi-source data integration: ; Where X is the environmental state (quantized value), is the observation result from the 1st sensor to the nth sensor; For the The optimal state estimation is obtained by maximum a posteriori probability (MAP) estimation: ; In the implementation, Kalman filtering, particle filtering and other algorithms are used to process continuous state variables, and evidence theory (Dempster-Shafer theory) is used to process discrete state variables to achieve noise suppression and conflict resolution, and obtain enhanced perception results. Based on the enhanced perception results, a local environment dynamic map centered on the vehicle and covering a range of 300-500 meters around it is constructed. The map contains the following levels: Static layer: road geometry, lane lines, traffic signs, permanent obstacles, etc. Semi-static layer: temporary road works, traffic control, weather conditions, etc. Dynamic layer: dynamic targets such as surrounding vehicles, pedestrians, bicycles, animals, etc. The map is represented by a unified local coordinate system and supports an update frequency of more than 10Hz to ensure the real-time status of the environment. By structuring local map information into a scene understanding model, including object recognition, behavior prediction and risk assessment, a shared situational model is formed to support advanced decision-making functions. Based on high-precision spatiotemporal positioning data, the relative position, relative speed and driving direction between adjacent vehicles are calculated. Define the necessary conditions for the formation of a convoy: Spatial proximity: The distance between vehicles is within a safe range; Speed ​​consistency: speed difference is within the threshold range; Direction consistency: driving direction is basically consistent; Communication quality: The communication link quality is greater than the preset threshold; These conditions constitute the convoy formation condition set. Based on the convoy formation condition set, an adaptive convoy organization algorithm is defined. The algorithm contains the following core components: Team initialization: The team leader is selected based on the leader election algorithm. Usually, the team leader is the one with the best communication quality or the one at the front. Fleet joining / leaving: Define smooth joining and leaving protocols to ensure fleet structure stability; Fleet split / merge: Supports dynamic splitting and merging of fleets based on traffic conditions and destination information; Distance control: Adaptively adjust the distance between vehicles according to vehicle speed, road conditions and communication quality. Common control strategies include constant time distance (CTH) or constant safety distance (CSD); Distributed optimization methods, such as consensus-based distributed model predictive control (DMPC), are used in the algorithm implementation. Through algorithm optimization, intelligent fleet formation strategies are formed to support dynamic, safe, and efficient fleet organization and control. A distributed decision-making mechanism is defined by combining a shared situation model and a dynamic energy management strategy network. This mechanism is based on a consensus algorithm and game theory methods, enabling each vehicle to coordinate and make globally optimized decisions based on local information and limited neighbor communication, including: Distributed consensus algorithms are used for state synchronization and intent coordination, and are typically implemented as average consistency algorithms. Distributed optimization algorithms are used to solve global resource allocation problems, such as ADMM (Alternating Direction Method of Multipliers): The distributed decision-making process considers multiple goals, including safety, energy efficiency, travel efficiency, and ride comfort. These goals are balanced through multi-objective optimization methods to form a collaborative decision-making framework, so that each vehicle can make decisions close to the global optimality based on local information. For abnormal situations such as communication interruption, node failure, and malicious attacks, a robustness protection scheme with multiple layers of protection is defined. The scheme includes: Redundant communication paths: Use multi-hop communication and network coding technology to ensure that information can still be transmitted when some links fail; Decision degradation strategy: Define a hierarchical decision framework to smoothly degrade to a more conservative decision mode when communication quality degrades; Anomaly detection mechanism: Apply statistical methods and machine learning techniques to detect abnormal behaviors and malicious attacks; Local backup strategy: pre-configure backup decision strategies based on local information to ensure basic security when the collaborative system fails; Security authentication mechanism: Use PKI (public key infrastructure) and blockchain technology to ensure communication security and data integrity; Through these mechanisms, the system can maintain basic functions in the face of various abnormal situations and ensure the safety of vehicles and passengers. Integrate efficient information sharing mechanisms, intelligent fleet formation strategies, collaborative decision-making frameworks and robustness assurance solutions to build a multi-level vehicle-connected collaborative framework. The framework consists of four levels: Communication layer: handles information exchange between vehicles and ensures the efficiency and reliability of data transmission; Perception layer: Integrate multi-source perception data and build a shared environment model; Decision-making layer: Collaborative decision-making optimization based on environmental models and energy strategies; Security layer: Provides robustness protection to deal with abnormal situations; The framework adopts a modular definition, supports independent upgrades and expansions of different functions, adapts to application scenarios under different penetration rates and technical conditions, and provides comprehensive collaborative support for commercial vehicle intelligent travel, from simple information sharing to complex collaborative control.

[0031] This application builds a complete set of intelligent networking solutions for commercial vehicles by fusing and processing the Beidou multimodal positioning signals obtained by commercial vehicles and enhancing their accuracy, combined with comprehensive monitoring and analysis of vehicle status. This solution achieves a leap from single-vehicle intelligence to group collaboration, organically combining vehicle positioning, status monitoring, behavior analysis, energy management and collaborative control to form a closed-loop intelligent travel ecosystem. The system uses advanced technologies such as multi-source data fusion, machine learning, and distributed decision-making to greatly improve the positioning accuracy, energy efficiency, and travel experience of commercial vehicles. At the same time, through deep integration with urban intelligent transportation systems, it promotes the coordinated development of commercial vehicles and urban transportation infrastructure.

[0032] The present application also provides a commercial vehicle intelligent connection device based on Beidou positioning, which includes a memory and a processor. The memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, the processor executes the steps of the commercial vehicle intelligent connection system based on Beidou positioning in the above-mentioned embodiments.

[0033] The present application also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are executed on a computer, the computer executes the steps of the commercial vehicle intelligent connection system based on Beidou positioning.

[0034] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0035] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program codes.

[0036] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.

[0037] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technical users in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. The commercial vehicle intelligent connection system based on Beidou positioning is characterized by: include: The data acquisition module is used to perform fusion processing and precision enhancement on the Beidou multi-modal positioning signals acquired by commercial vehicles to obtain high-precision spatiotemporal positioning data; Intelligent analysis module, used to collect and intelligently analyze the state parameters of the power system, battery system and on-board equipment of commercial vehicles in multiple dimensions to obtain the vehicle health index model; An intelligent mining module, based on the high-precision spatiotemporal positioning data and the vehicle health index model, performs commercial vehicle group behavior pattern mining to obtain a spatiotemporal trajectory knowledge graph; An energy optimization module, which is used to optimize the energy efficiency and charge planning of commercial vehicles by using the spatiotemporal trajectory knowledge graph to obtain a dynamic energy management strategy network; The multi-level vehicle connection module builds a collaborative perception and distributed decision-making mechanism between vehicles based on high-precision spatiotemporal positioning data and a dynamic energy management strategy network, and obtains a multi-level vehicle connection collaboration framework; the multi-level vehicle connection collaboration framework is integrated with the urban intelligent transportation system to obtain a comprehensive solution for commercial vehicle intelligent travel.

2. The commercial vehicle intelligent connection system based on Beidou positioning according to claim 1 is characterized in that: The fusion processing and precision enhancement of the Beidou multi-modal positioning signals obtained by the commercial vehicle to obtain high-precision spatiotemporal positioning data includes: Collecting Beidou navigation system multi-frequency signals and ground augmentation station differential signals to obtain an original positioning data set, and performing signal quality evaluation on the original positioning data set to obtain a signal reliability vector; Based on the signal reliability vector, weighted fusion is performed on multi-source positioning information to obtain a preliminary fused positioning result, and the positioning noise is suppressed by using a Kalman filter algorithm to obtain filtered positioning data; Combining the inertial navigation unit data, adaptively compensating the filtered positioning data to obtain a continuous positioning trajectory, and applying a trajectory prediction algorithm in the signal blocking area to obtain a complete positioning sequence; Performing map matching and constraint optimization on the complete positioning sequence to obtain road-level precise positioning data, and adaptively correcting the positioning accuracy in complex scenarios based on a deep learning model to obtain sub-meter positioning results; The sub-meter positioning results are subjected to spatiotemporal index construction and time stamp synchronization to obtain high-precision spatiotemporal positioning data.

3. The commercial vehicle intelligent connection system based on Beidou positioning according to claim 2 is characterized in that: The multi-dimensional collection and intelligent analysis of state parameters of the power system, battery system and on-board equipment of the commercial vehicle to obtain a vehicle health index model includes: Collecting power system parameters through a multi-source sensor network to obtain a power system state matrix, and performing abnormality detection and fault diagnosis on the power system state matrix to obtain a power system health vector; Monitor the key parameters of the battery pack to obtain the battery status data set, and evaluate the battery health status based on electrochemical impedance spectroscopy analysis technology to obtain the battery capacity decay model; Collecting the operating status data of the vehicle control unit, communication module and auxiliary equipment to obtain a set of equipment performance indicators, and constructing an equipment reliability evaluation network based on the equipment performance indicator set to obtain an equipment health score; Performing multi-level fusion on the power system health vector, the battery capacity attenuation model and the equipment health score to obtain a vehicle comprehensive state feature space, and applying clustering and classification algorithms in the vehicle comprehensive state feature space to obtain a vehicle state assessment result; Based on the vehicle status assessment results, a deep neural network prediction model is constructed to dynamically predict the remaining service life of key vehicle components, and combined with historical operating data to generate a vehicle health index model.

4. The commercial vehicle intelligent connection system based on Beidou positioning according to claim 3 is characterized in that: The commercial vehicle group behavior pattern mining is performed based on the high-precision spatiotemporal positioning data and the vehicle health index model to obtain a spatiotemporal trajectory knowledge graph, including: Performing trajectory segmentation and feature extraction on the high-precision spatiotemporal positioning data to obtain a vehicle movement pattern feature set, and performing trajectory clustering based on the vehicle movement pattern feature set to obtain a typical driving path library; Combining the vehicle health index model with the driving behavior data, establishing a correlation analysis model between driving habits and vehicle status, and obtaining a driving-vehicle status mapping matrix; Frequent pattern mining is performed on the spatiotemporal trajectory data of a large number of vehicles to identify high-frequency travel paths and time windows, obtain a spatiotemporal activity pattern set, and build an urban traffic flow prediction model based on the spatiotemporal activity pattern set to obtain a dynamic traffic flow distribution map; Use graph neural networks to model the interaction behaviors between vehicles, obtain the vehicle social network structure, identify key nodes and community structures in the vehicle social network structure, and obtain the vehicle group influence distribution; The typical driving path library, the driving-vehicle state mapping matrix, the dynamic traffic distribution map and the vehicle group influence distribution are integrated to construct a multi-level knowledge representation model and obtain a spatiotemporal trajectory knowledge graph.

5. The commercial vehicle intelligent connection system based on Beidou positioning according to claim 4 is characterized in that: The method utilizes the spatiotemporal trajectory knowledge graph to optimize commercial vehicle energy efficiency and charge planning, and obtains a dynamic energy management strategy network, including: Based on the spatiotemporal trajectory knowledge graph, the energy consumption of commercial vehicle driving routes is analyzed to obtain a road section energy consumption feature library, and a multi-factor energy consumption prediction model is established in combination with terrain, weather and traffic condition data to obtain accurate energy consumption estimation results; Calculate the vehicle's cruising range based on the precise energy consumption estimation result and the battery status information to obtain a mileage prediction curve, and provide personalized energy-saving driving advice to the driver based on the mileage prediction curve to obtain a real-time driving optimization strategy; Using the distribution data of urban charging facilities and the real-time status information of charging stations, combined with the knowledge graph of spatiotemporal trajectories, an intelligent charging recommendation model is constructed to obtain the optimal charging station selection plan. Based on the grid load and electricity price fluctuation data, the optimal charging time period is recommended to obtain an economic charging strategy. Coordinated scheduling of multiple vehicle charging demands to obtain a balanced charging resource allocation plan, and optimized the charging resource sharing mechanism between vehicles based on the game theory model to obtain a collaborative charging network; The real-time driving optimization strategy, the economic charging strategy and the collaborative charging network are integrated to construct a hierarchical decision support system and obtain a dynamic energy management strategy network.

6. The commercial vehicle intelligent connection system based on Beidou positioning according to claim 5 is characterized in that: Based on the high-precision spatiotemporal positioning data and the dynamic energy management strategy network, a vehicle-to-vehicle collaborative perception and distributed decision-making mechanism is constructed to obtain a multi-level vehicle-to-vehicle collaborative framework, including: Based on the vehicle wireless communication technology, a direct communication link between vehicles is established to obtain a vehicle communication topology network, and a data exchange protocol is defined according to the vehicle communication topology network to obtain an efficient information sharing mechanism; Integrate the vehicle's own sensor data and the perception information shared by surrounding vehicles to obtain enhanced perception results, and build a dynamic map of the local environment based on the enhanced perception results to obtain a shared situation model; According to the high-precision spatiotemporal positioning data, the relative position relationship between vehicles is calculated to obtain a set of conditions for forming a convoy, and based on the set of conditions for forming a convoy, an adaptive convoy organization algorithm is defined to obtain an intelligent convoy formation strategy; Combining the shared situation model and the dynamic energy management strategy network, a distributed decision-making mechanism is defined to enable each vehicle to make the global optimal decision based on local information, thus obtaining a collaborative decision-making framework; A fault-tolerant mechanism and a backup decision-making strategy are defined to obtain a robustness assurance solution, and the efficient information sharing mechanism, the intelligent vehicle fleet formation strategy, the collaborative decision-making framework and the robustness assurance solution are integrated to obtain a multi-level vehicle-connected collaborative framework.

7. A commercial vehicle intelligent connection device based on Beidou positioning, characterized in that: The commercial vehicle intelligent connection device based on Beidou positioning includes: a memory and at least one processor, wherein instructions are stored in the memory; The at least one processor calls the instructions in the memory so that the commercial vehicle intelligent connection device based on Beidou positioning executes the commercial vehicle intelligent connection system based on Beidou positioning as described in any one of claims 1-6.

8. A computer-readable storage medium having instructions stored thereon, characterized in that: When the instructions are executed by the processor, a commercial vehicle intelligent connection system based on Beidou positioning is implemented as described in any one of claims 1-6.

Images