Intelligent Connection System for Electric Vehicles Based on Beidou Positioning

Abstract

The present invention belongs to the technical field of electric vehicles, and discloses an intelligent connection system for electric vehicles based on Beidou positioning; it includes: obtaining Beidou positioning data and driving state data; performing multi-source fusion correction on the Beidou positioning data and conducting trajectory optimization processing to obtain the position trajectory of the two-wheeled electric vehicle; extracting features from the position trajectory of the two-wheeled electric vehicle to generate feature vectors; performing spatio-temporal correlation analysis on the feature vectors to generate a driving behavior model; conducting safety risk assessment on the driving behavior model to obtain a risk warning index; performing classification processing on the risk warning index to obtain a classified safety warning strategy; obtaining environmental data; performing semantic analysis on the environmental data to construct a knowledge graph; identifying dynamic obstacles in the knowledge graph and making path planning decisions to construct an intelligent navigation strategy; controlling the two-wheeled electric vehicle based on the classified safety warning strategy and the intelligent navigation strategy, significantly improving the riding safety and travel efficiency of the electric vehicle.

Description

Technical Field

[0001] The present invention relates to the technical field of electric vehicles, and more specifically, to an intelligent connection system for electric vehicles based on Beidou positioning. Background Art

[0002] With the rapid development of urban transportation and the wide application of two-wheeled electric vehicles in short-distance travel, the demand for the intelligence and safety of electric vehicles has become increasingly prominent. As a high-precision positioning technology, the Beidou satellite navigation system has the advantages of global coverage, high precision, anti-interference, etc., and has gradually become an important technical support in the field of intelligent transportation. However, the existing intelligent connection systems for electric vehicles based on Beidou positioning still face many challenges in practical applications.

[0003] The existing technologies have deficiencies in the positioning and intelligent connection control of two-wheeled electric vehicles, which limit their safety and intelligence levels in complex urban traffic environments. First, traditional positioning technologies rely on a single GPS signal and are easily affected by weak signal areas such as high-rise building blocks, tunnels, and bridges, resulting in a decrease in positioning accuracy, trajectory jumps or losses, and it is difficult to meet the requirements of high-precision navigation and safety control. Second, the existing systems have insufficient ability to extract and model the dynamic characteristics of the driving behavior of electric vehicles, lack in-depth analysis of spatio-temporal correlation characteristics such as speed, acceleration, and direction changes, and cannot accurately identify driving behavior patterns, thus making it difficult to achieve personalized safety risk assessment and early warning. In addition, the existing technologies have limited perception ability of the road environment and are difficult to obtain and analyze complex road scene information in real time, such as dynamic obstacles, traffic flow changes, road topology, etc., resulting in a lack of dynamic adaptability in path planning and difficulty in coping with situations such as sudden obstacles or traffic jams. In terms of safety control, the existing systems lack a hierarchical early warning and adaptive control mechanism and cannot dynamically adjust intervention strategies according to the risk level. For example, there is a lack of rapid braking or steering assistance in case of emergency risks, a lack of deceleration reminder and risk avoidance path coordination in medium risks, and inability to optimize navigation efficiency in low risks. These problems are manifested in actual situations as follows: when an electric vehicle is driving on complex urban roads, positioning drift leads to navigation errors, abnormal behaviors are not recognized in time and potential accidents are caused, path planning fails to avoid construction areas or congested sections, safety interventions are lagged or conflict with navigation, ultimately reducing the safety and travel experience of users.

[0004] In view of this, the present invention proposes an intelligent connection system for electric vehicles based on Beidou positioning to solve the above problems. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art and to achieve the above object, the present invention provides the following technical solution: An intelligent connection system for electric vehicles based on Beidou positioning, comprising:

[0006] A high-precision positioning module, which is used to obtain Beidou positioning data and driving state data of the two-wheeled electric vehicle; perform multi-source fusion correction on the Beidou positioning data and conduct trajectory optimization processing, so as to obtain a high-precision position trajectory of the two-wheeled electric vehicle;

[0007] A behavior model fitting module, which is used to extract dynamic features from the high-precision position trajectory of the two-wheeled electric vehicle to generate a multi-dimensional driving feature vector; perform spatio-temporal correlation analysis on the multi-dimensional driving feature vector, so as to generate a driving behavior model of the two-wheeled electric vehicle;

[0008] A safety policy evaluation module, which is used to evaluate the safety risks of the driving behavior model of the two-wheeled electric vehicle to obtain a risk warning index; perform grading processing on the risk warning index, so as to obtain a graded safety warning policy;

[0009] A knowledge graph construction module, which is used to obtain surrounding road environment data; perform real-time semantic analysis on the surrounding road environment data to construct a road scene knowledge graph;

[0010] A navigation policy fitting module, which is used to identify dynamic obstacles in the road scene knowledge graph and make path planning decisions to construct an intelligent navigation policy;

[0011] An adaptive control module, which adaptively controls the two-wheeled electric vehicle based on the graded safety warning policy and the intelligent navigation policy.

[0012] Further, the obtaining of Beidou positioning data and driving state data of the two-wheeled electric vehicle; performing multi-source fusion correction on the Beidou positioning data and conducting trajectory optimization processing, so as to obtain a high-precision position trajectory of the two-wheeled electric vehicle, includes:

[0013] Obtain Beidou positioning data and driving state data of the two-wheeled electric vehicle;

[0014] Perform signal quality evaluation on the Beidou positioning data to identify positioning points in weak signal areas;

[0015] Perform inertial navigation compensation on the positioning points in weak signal areas to eliminate signal occlusion errors;

[0016] Combine the data of the wheel speed sensor to perform odometer-assisted correction on the Beidou positioning data, so as to obtain corrected positioning data;

[0017] Perform Kalman filter smoothing processing on the corrected positioning data to generate a smoothed positioning trajectory;

[0018] Perform map matching and trajectory optimization processing on the smoothed positioning trajectory, so as to obtain a high-precision position trajectory of the two-wheeled electric vehicle.

[0019] Further, the performing of Kalman filter smoothing processing on the corrected positioning data includes:

[0020] Analyze the noise characteristics of the calibrated positioning data to obtain a positioning noise model;

[0021] Design adaptive Kalman filter parameters according to the positioning noise model to obtain optimized filter parameters;

[0022] Apply a Kalman filter with optimized filter parameters to the calibrated positioning data to generate a preliminary smoothed trajectory;

[0023] Detect abnormal points in the preliminary smoothed trajectory and mark the trajectory jump points;

[0024] Perform corrective interpolation on the trajectory jump points based on historical trajectory data to generate a continuous and smoothed trajectory;

[0025] Perform time-window sliding averaging on the continuous and smoothed trajectory to generate a smoothed positioning trajectory.

[0026] Furthermore, perform dynamic feature extraction on the position trajectory of the high-precision two-wheeled electric vehicle to generate a multi-dimensional driving feature vector; perform spatio-temporal correlation analysis on the multi-dimensional driving feature vector to generate a driving behavior model for the two-wheeled electric vehicle, including:

[0027] Segment the position trajectory of the high-precision two-wheeled electric vehicle by time window to generate multiple trajectory segments;

[0028] Calculate the average speed, maximum speed, acceleration distribution, angular velocity, and driving direction change for multiple trajectory segments to obtain multiple dynamic feature parameters;

[0029] Perform feature dimensionality reduction processing based on multiple dynamic feature parameters to generate a multi-dimensional driving feature vector;

[0030] Perform clustering analysis on the multi-dimensional driving feature vector to obtain driving behavior classification data;

[0031] Perform spatio-temporal correlation analysis on the multi-dimensional driving feature vector to generate spatio-temporal correlation features;

[0032] Construct a driving behavior model for the two-wheeled electric vehicle based on the driving behavior classification data and spatio-temporal correlation features.

[0033] Furthermore, perform safety risk assessment on the driving behavior model of the two-wheeled electric vehicle to obtain a risk warning index; perform grading processing on the risk warning index to obtain a graded safety warning strategy, including:

[0034] Train the driving behavior model of the two-wheeled electric vehicle with historical data to obtain a safety risk assessment model;

[0035] Based on the safety risk assessment model, perform risk prediction on real-time driving data to obtain a risk warning index;

[0036] Perform threshold division processing on the risk warning index to obtain multiple levels of risk levels;

[0037] Formulate corresponding warning and intervention strategies for different risk levels, thereby obtaining a hierarchical safety warning strategy.

[0038] Furthermore, obtain the surrounding road environment data; perform real-time semantic analysis on the surrounding road environment data to construct a road scene knowledge graph, including:

[0039] Receive real-time data of the surrounding road environment to obtain the surrounding road environment data;

[0040] Extract multi-modal features from the surrounding road environment data to generate environmental feature data;

[0041] Perform semantic segmentation analysis on the environmental feature data to generate road scene semantic labels;

[0042] Perform relationship reasoning modeling on the road scene semantic labels to construct a road scene knowledge graph.

[0043] Furthermore, the performing semantic segmentation analysis on the environmental feature data to generate road scene semantic labels includes:

[0044] Identify road types, lane lines, traffic signs, and traffic lights based on the environmental feature data, that is, the identification results;

[0045] Evaluate the confidence of the identification results and screen the high-confidence identification results;

[0046] Perform road topology structure analysis based on the high-confidence identification results to obtain road network structure data;

[0047] Evaluate the traffic flow status of the road network structure data to obtain real-time traffic status data;

[0048] Perform trend prediction on the real-time traffic status data based on historical data to obtain traffic flow prediction results;

[0049] Perform semantic annotation on the road network structure data and the traffic flow prediction results to generate road scene semantic labels.

[0050] Furthermore, the performing dynamic obstacle identification on the road scene knowledge graph and making path planning decisions to construct an intelligent navigation strategy includes:

[0051] Perform real-time query on the road scene knowledge graph to extract surrounding potential obstacle data;

[0052] Perform target path planning for the current position of the two-wheeled electric vehicle to obtain the expected driving path;

[0053] Dynamically identify and track potential surrounding obstacles based on the expected driving path data to obtain obstacle risk assessment data;

[0054] Calculate the collision risk based on the obstacle risk assessment data to generate a collision risk assessment value;

[0055] Make real-time path optimization decisions based on the collision risk assessment value to construct an intelligent navigation strategy.

[0056] Furthermore, the adaptive control of the two-wheeled electric vehicle based on the hierarchical safety warning strategy and the intelligent navigation strategy includes:

[0057] Construct an intelligent interface for the two-wheeled electric vehicle control system to generate an intelligent control framework;

[0058] Map safety intervention rules to the intelligent control framework based on the hierarchical safety warning strategy to construct a safety intervention mechanism;

[0059] Map navigation assistance rules to the intelligent control framework based on the intelligent navigation strategy to construct a navigation assistance mechanism;

[0060] Coordinate the priorities of the safety intervention mechanism and the navigation assistance mechanism;

[0061] The priority coordination is specifically: identify the safety risk level based on the current situation;

[0062] The safety risk levels include: emergency risk, medium risk, and low risk;

[0063] When the safety risk level is an emergency risk, the priority of the safety intervention mechanism is higher than that of the navigation assistance mechanism, and emergency braking or steering assistance is executed;

[0064] When the safety risk level is a medium risk, the safety intervention mechanism and the navigation assistance mechanism work together to execute deceleration reminders and provide advice on evasive paths;

[0065] When the safety risk level is a low risk, the priority of the navigation assistance mechanism is higher than that of the safety intervention mechanism, and the optimal path navigation is normally executed.

[0066] The technical effects and advantages of the electric vehicle intelligent connection system based on Beidou positioning of the present invention:

[0067] The present invention significantly improves the Beidou positioning accuracy through a multi-source fusion calibration technology, solving the problem of inaccurate positioning of traditional electric vehicles in complex urban environments; realizes intelligent analysis of riding behaviors through dynamic feature extraction and behavior modeling, providing a reliable basis for safety warnings; realizes in-depth understanding of road environments through real-time semantic analysis and knowledge graph construction, laying a foundation for intelligent navigation; significantly improves the riding safety and travel efficiency of electric vehicles through hierarchical safety warnings and dynamic path planning; and realizes intelligent control of human-machine collaboration through an adaptive control system, greatly enhancing the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 It is a schematic diagram of the electric vehicle intelligent connection system based on Beidou positioning of the present invention;

[0069] Figure 2 It is a logical schematic diagram for the present invention to obtain a high-precision electric vehicle position trajectory;

[0070] Figure 3 It is a logical schematic diagram for the present invention to generate an electric vehicle driving behavior model;

[0071] Figure 4 It is a logical schematic diagram for the present invention to obtain a hierarchical safety warning strategy. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0072] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0073] Embodiment 1

[0074] Please refer to Figures 1 to 4 As shown, the electric vehicle intelligent connection system based on Beidou positioning in this embodiment includes:

[0075] A high-precision positioning module, configured to obtain Beidou positioning data and driving state data of the electric vehicle; perform multi-source fusion calibration on the Beidou positioning data and perform trajectory optimization processing, so as to obtain a high-precision electric vehicle position trajectory;

[0076] A behavior model fitting module, configured to perform dynamic feature extraction on the high-precision electric vehicle position trajectory to generate a multi-dimensional driving feature vector; perform spatio-temporal correlation analysis on the multi-dimensional driving feature vector, so as to generate an electric vehicle driving behavior model;

[0077] A security policy evaluation module for performing security risk assessment on the electric vehicle driving behavior model to obtain a risk warning index, and performing classification processing on the risk warning index to obtain a classified security warning policy;

[0078] A knowledge graph construction module for obtaining surrounding road environment data and performing real-time semantic analysis on the surrounding road environment data to construct a road scene knowledge graph;

[0079] A navigation strategy fitting module for identifying dynamic obstacles in the road scene knowledge graph, making path planning decisions, and constructing an intelligent navigation strategy;

[0080] An adaptive control module for adaptively controlling the electric vehicle based on the classified security warning policy and the intelligent navigation strategy.

[0081] The present invention improves the Beidou positioning accuracy through multi-source fusion correction and trajectory optimization processing, solves the problem of inaccurate positioning of traditional electric vehicles, and after obtaining a high-precision electric vehicle position trajectory, provides a reliable data basis for subsequent analysis. Dynamic feature extraction and spatio-temporal correlation analysis can deeply understand the driving characteristics and change rules of electric vehicles, providing a basis for security warning. Security risk assessment and classification processing realize early warning of potential dangers, improve the riding safety of electric vehicles. The road scene knowledge graph constructed by real-time semantic analysis provides environmental perception ability for intelligent navigation, enhances the environmental adaptability of electric vehicles. Dynamic obstacle recognition and path planning decisions can avoid risks in advance and reduce the accident rate. The adaptive control system realizes intelligent management of electric vehicles according to the risk level and navigation strategy, greatly improving the user riding experience and safety.

[0082] In an embodiment of the present invention, refer to Figure 1 , which is a schematic diagram of the step flow of an electric vehicle intelligent connection system based on Beidou positioning according to the present invention. In this example, the electric vehicle intelligent connection system based on Beidou positioning includes:

[0083] A high-precision positioning module for obtaining Beidou positioning data and driving state data of the electric vehicle, performing multi-source fusion correction on the Beidou positioning data, and performing trajectory optimization processing to obtain a high-precision electric vehicle position trajectory;

[0084] In this embodiment, the Beidou positioning module installed on the electric vehicle is used to collect the geographical location information of the electric vehicle in real time, including data such as longitude, latitude, altitude, speed, and direction. At the same time, driving state data is collected from the in-vehicle sensor system of the electric vehicle, and these data include: wheel speed, motor power, battery state, tilt sensor data, handlebar steering angle, etc. The Beidou positioning data is fused with other positioning data sources (such as inertial navigation systems, wheel speed sensors, Bluetooth beacons, etc.), and the Kalman filtering algorithm is used to optimize the fusion of multi-source data to improve the positioning accuracy. Aiming at the problem that Beidou signals are easily blocked or affected by multipath effects in complex environments such as urban canyons and tunnels, inertial navigation compensation and map matching technologies are used to optimize the trajectory and eliminate positioning jumps and drifts. Through these processes, a high-precision, continuous and smooth electric vehicle position trajectory is obtained, providing a reliable data basis for subsequent analysis and applications.

[0085] The behavior model fitting module is used to extract dynamic features from the high-precision electric vehicle position trajectory to generate a multi-dimensional driving feature vector; perform spatio-temporal correlation analysis on the multi-dimensional driving feature vector, thereby generating an electric vehicle driving behavior model;

[0086] In this embodiment, for the obtained high-precision electric vehicle position trajectory, the method of a sliding time window is adopted to extract the dynamic features of the electric vehicle's driving. The dynamic features include: average speed, speed change rate, acceleration distribution, turning radius, and steering frequency. Through feature engineering methods, these original features are converted into multi-dimensional driving feature vectors, and each multi-dimensional driving feature vector represents the driving features of the electric vehicle within a specific time window. These multi-dimensional driving feature vectors are input into the spatio-temporal correlation analysis model. The spatio-temporal correlation analysis model adopts the architecture of a recurrent neural network (RNN) or a long short-term memory network (LSTM) to learn the correlation of the electric vehicle's driving features in the time and space dimensions. Through deep learning training, the model can identify typical and abnormal driving patterns of the electric vehicle, such as sudden acceleration, sharp turns, and frequent lane changes. Finally, a model that can describe the driving behavior characteristics of the electric vehicle, that is, the electric vehicle driving behavior model, is obtained.

[0087] The safety policy evaluation module is used to conduct a safety risk assessment on the electric vehicle driving behavior model to obtain a risk warning index; perform classification processing on the risk warning index, thereby obtaining a classified safety warning policy;

[0088] In this embodiment, the electric vehicle driving behavior model is used as the input and processed by the safety risk assessment algorithm. Based on historical accident data and expert knowledge, this algorithm quantitatively evaluates the risk levels of various driving behaviors. The evaluation takes into account multiple factors, including: whether the speed is too fast, whether the turn is too sharp, whether lane changes are frequent, whether there are characteristics of fatigued riding, etc. Through the risk scoring mechanism, a risk warning index is assigned to each behavior pattern, and this index reflects the risk level of the current driving behavior that may lead to an accident. The risk warning index is classified, usually divided into three levels: high risk, medium risk, and low risk. For different risk levels, corresponding warning strategies are formulated. For example, for high risk, it triggers an audible and visual alarm and actively decelerates; for medium risk, it emits a prompt tone and displays a warning message; for low risk, it only displays a prompt on the APP interface, etc. This hierarchical safety warning strategy can take corresponding intervention measures according to different risk levels and improve the riding safety of electric vehicles.

[0089] The knowledge graph construction module is used to obtain the surrounding road environment data; perform real-time semantic analysis on the surrounding road environment data and construct a road scene knowledge graph;

[0090] In this embodiment, the surrounding road environment data of the electric vehicle is obtained through multiple channels. These data sources include: road network data provided by online map services, real-time traffic conditions released by traffic management departments, weather information provided by meteorological departments, and road condition information shared by other connected vehicles, etc. The collected road environment data is preprocessed, including data cleaning, format conversion, and standardization processing. Natural language processing and computer vision technologies are used to perform semantic analysis on the road environment data. For example, identify road types (main roads, side roads, bicycle lanes, etc.), traffic signs (speed limits, bans, warnings, etc.), and road conditions (congestion, construction, accidents, etc.). The semantic analysis results are constructed into a structured knowledge graph, which describes the various elements of the road environment and their relationships. Through this knowledge graph, the system can understand complex road scenes and provide semantic-level support for subsequent navigation decisions.

[0091] The navigation strategy fitting module is used to identify dynamic obstacles in the road scene knowledge graph, make path planning decisions, and construct an intelligent navigation strategy;

[0092] In this embodiment, based on the constructed road scene knowledge graph, the system can identify various obstacles in the surrounding environment. These obstacles include static obstacles (such as roadblocks, potholes, buildings, etc.) and dynamic obstacles (such as pedestrians, other vehicles, emergencies, etc.). For the identified obstacles, the system will evaluate their potential risks to the electric vehicle's driving. The evaluation considers factors such as the type, distance, relative speed, and moving trend of the obstacles, and calculates a collision risk index. Based on the current location, destination, and road scene knowledge graph, the system plans the optimal driving path. This path not only considers the shortest distance but also takes into account various factors such as safety, comfort, and energy consumption. Integrate the obstacle recognition results and path planning results into a comprehensive intelligent navigation strategy. This strategy includes: recommended driving routes, suggested driving speeds, turning prompts, danger warnings, etc., providing comprehensive navigation assistance for riders.

[0093] The adaptive control module adaptively controls the electric vehicle based on the hierarchical safety warning strategy and the intelligent navigation strategy.

[0094] In this embodiment, taking the hierarchical safety warning strategy and the intelligent navigation strategy as inputs, according to different risk levels and navigation requirements, automatically adjust the driving parameters of the electric vehicle. For example, automatically reduce the maximum speed limit in high-risk situations, optimize the motor power output in complex road conditions, and start an energy-saving mode when the battery is low. Adopt a human-machine collaborative control method, which not only retains the rider's active control right but also provides assistance or intervention when necessary. For example, when a possible collision risk is detected, the system will issue a warning and prepare for assisted braking, but the final decision-making power still lies with the rider. Integrate the adaptive control system with the aforementioned positioning, analysis, warning, navigation, and other modules to build a complete intelligent networked safety system for electric vehicles. It realizes the intelligent management of electric vehicles and greatly improves the riding safety and user experience.

[0095] In this embodiment, refer to Figure 2 , which is a schematic diagram of the detailed implementation steps for "obtaining Beidou positioning data and driving state data of the electric vehicle; performing multi-source fusion correction on the Beidou positioning data and conducting trajectory optimization processing to obtain a high-precision electric vehicle position trajectory". The detailed implementation steps include:

[0096] Obtain Beidou positioning data and driving state data of the electric vehicle;

[0097] Evaluate the signal quality of the Beidou positioning data and identify the positioning points in weak signal areas;

[0098] Perform inertial navigation compensation on the positioning points in weak signal areas to eliminate signal occlusion errors;

[0099] Combine the data of the wheel speed sensor to perform odometer-assisted correction on the Beidou positioning data to obtain corrected positioning data;

[0100] Perform Kalman filtering and smoothing on the calibrated positioning data to generate a smoothed positioning trajectory;

[0101] Perform map matching and trajectory optimization on the smoothed positioning trajectory to obtain a high-precision electric vehicle position trajectory.

[0102] In this embodiment, real-time Beidou positioning data, including information such as longitude, latitude, altitude, and timestamp, is collected through a Beidou receiver installed on the electric vehicle. At the same time, driving state data, including wheel speed, motor power, battery state, inclination angle, acceleration, etc., is collected from the sensor network of the electric vehicle. Quality assessment of the collected Beidou signals is performed, and indicators such as signal-to-noise ratio (SNR) and geometric dilution of precision (GDOP) are calculated. By setting thresholds, positioning points in weak signal areas with poor signal quality are identified. These points usually appear in areas with severe signal blockage such as high-rise building clusters, tunnels, and underground garages. For the identified positioning points in weak signal areas, the inertial navigation system is activated for compensation. The inertial navigation system is based on sensors such as accelerometers and gyroscopes, and calculates the position change of the electric vehicle through integral calculation.

[0103] Specifically, the accelerometer collects the linear acceleration data of the three axes (X, Y, Z) of the electric vehicle, and the gyroscope collects the angular velocity data of the three axes of the electric vehicle. The attitude angles (yaw angle, pitch angle, roll angle) of the vehicle are obtained through integral of the gyroscope data; Attitude angle(t) = Attitude angle(t - 1) + Angular velocity × Time interval. The acceleration of the vehicle body coordinate system measured by the accelerometer is transformed into the global coordinate system, and a rotation matrix is constructed using the attitude angle for coordinate transformation. The acceleration in the global coordinate system is integrated over time to obtain the velocity: Velocity(t) = Velocity(t - 1) + Acceleration × Time interval; The velocity is integrated over time to obtain the position change:

[0104] Position(t) = Position(t - 1) + Velocity × Time interval; The zero velocity update (ZUPT) technology is applied to correct the velocity drift when the vehicle is detected to be stationary, and periodic calibration is performed in combination with the Beidou positioning data to prevent error accumulation caused by long-term integration.

[0105] Fuse the inertial navigation results with Beidou positioning to eliminate positioning errors caused by signal occlusion. Obtain the driving distance data of the electric vehicle from the wheel speed sensor, and fuse these data with the Beidou positioning results to achieve odometer-assisted calibration. This calibration is applicable to scenarios of short-distance and straight-line driving, and can effectively improve the continuity and smoothness of positioning. Input the calibrated positioning data obtained from the foregoing processing into a Kalman filter for smoothing. The Kalman filter can comprehensively consider the historical position and current observations, predict the most likely true position, effectively eliminate random noise and mutation points, and generate a smooth positioning trajectory. Match the smooth positioning trajectory with a high-precision electronic map to constrain the trajectory points on the actual road network. Through the map matching algorithm, correct the positioning points deviating from the road and perform trajectory optimization processing to finally obtain the position trajectory of the electric vehicle with high precision and conforming to the actual road conditions.

[0106] Specifically, load high-precision electronic map data, including information such as road network topology and road attributes, construct a spatial index structure (such as an R-tree) to accelerate spatial queries. For each smooth positioning point, find the road segments within a certain radius around it in the electronic map, calculate the vertical distance from the positioning point to each candidate road segment, and screen the road segments with relatively short distances as candidates. Calculate the matching probability for each candidate road segment, considering the following factors:

[0107] The distance from the positioning point to the road segment (the closer the higher the probability);

[0108] The consistency between the driving direction and the road direction (the more consistent the higher the probability);

[0109] The rationality of the vehicle speed and the road speed limit;

[0110] The continuity with the previous matching result;

[0111] Use the Hidden Markov Model (HMM) to construct the transition probability of the trajectory point sequence, apply the Viterbi algorithm to find the globally optimal matching road sequence to ensure the continuity and rationality of the trajectory. Project the matched positioning points onto the corresponding road segments to obtain the constrained trajectory points, apply the spline interpolation algorithm to smooth the projected trajectory to eliminate the sawtooth effect, detect and correct abnormal jumps in the trajectory (such as sudden road switches), correct unreasonable matching results according to the road topology relationship (such as skipping necessary nodes), and apply the least squares method to globally optimize the trajectory to balance the positioning accuracy and road constraints.

[0112] In this embodiment, the specific steps of "performing Kalman filter smoothing on the calibrated positioning data to generate a smooth positioning trajectory" are as follows:

[0113] Conduct noise characteristic analysis on the calibrated positioning data to obtain the positioning noise model;

[0114] Design adaptive Kalman filter parameters according to the positioning noise model to obtain optimized filtering parameters;

[0115] Apply the Kalman filter with optimized filtering parameters to the corrected positioning data to generate a preliminary smooth trajectory;

[0116] Perform outlier detection on the preliminary smooth trajectory and mark the trajectory jump points;

[0117] Perform correction interpolation on the trajectory jump points based on historical trajectory data to generate a continuous smooth trajectory;

[0118] Perform time window moving average processing on the continuous smooth trajectory to generate a smooth positioning trajectory.

[0119] In this embodiment, statistical analysis is performed on the corrected positioning data to calculate parameters such as the mean, variance, and distribution characteristics of the noise. These statistical characteristics are modeled into a mathematical form to obtain a positioning noise model. The positioning noise model describes the random characteristics and distribution laws of the positioning noise. Based on the obtained positioning noise model, the optimal Kalman filter parameters are calculated, including the process noise covariance matrix Q and the measurement noise covariance matrix R. These parameters will be adaptively adjusted according to different noise characteristics and environmental conditions to obtain the best filtering effect. The corrected positioning data is input into the Kalman filter configured with optimized filtering parameters for processing. The Kalman filter recursively estimates the true position state of the electric vehicle through two stages of prediction and update, and outputs a preliminary smooth trajectory. Outlier detection is performed on each point in the preliminary smooth trajectory to identify jump points with excessive deviation from the previous and subsequent points. Specifically, the preliminary smooth trajectory is composed of a series of trajectory points, that is, a trajectory point sequence. Calculate the local curvature and the rate of change of speed , define a sliding local window, traverse the trajectory point sequence, and calculate the adaptive fluctuation index based on the local window ; where and are the standard deviations of the rate of change of speed and the local curvature within the local window respectively, and then calculate the fluctuation index ;

[0120] Based on the local curvature , the rate of change of speed and the adaptive fluctuation index, calculate the multi-scale outlier ; where is the mean of the fluctuation index within the local window , is the standard deviation of the fluctuation index within the local window ;

[0121] For trajectory point i, if the maximum value of the multi-scale outliers within its local window ; where is the sensitivity parameter, is the standard deviation of the local trajectory segment. The sensitivity parameter is an adjustable coefficient used to control the strictness of anomaly detection and is set according to the noise level of the trajectory data and application requirements, usually taking values between 2.0 and 3.0. The standard deviation of the local trajectory segment measures the degree of change of the trajectory in a specific area, that is, a local window (such as k1 points before and after) is taken around the corresponding point, and the standard deviation of the positions of these points is calculated;

[0122] These jump points are usually the results of abnormal positioning caused by signal interference, multipath effects, or temporary occlusion. For the identified jump points, historical trajectory data is used for correction. Interpolation algorithms (such as linear interpolation, spline interpolation, etc.) are used to replace the abnormal points to make the trajectory more continuous and smooth. For the corrected trajectory, time window sliding average processing is applied to further eliminate high-frequency noise and fine jitter. This processing can retain the main features of the trajectory while improving the smoothness of the trajectory, and finally obtain a high-quality smooth positioning trajectory.

[0123] In this embodiment, referring to Figure 3 , it is a schematic diagram of the detailed implementation steps for "extracting dynamic features from the high-precision electric vehicle position trajectory to generate a multi-dimensional driving feature vector; performing spatio-temporal correlation analysis on the multi-dimensional driving feature vector to generate an electric vehicle driving behavior model". In this embodiment, the detailed implementation steps include:

[0124] Segment the high-precision electric vehicle position trajectory by time window to generate multiple trajectory segments;

[0125] Calculate the average speed, maximum speed, acceleration distribution, angular velocity, and driving direction change for multiple trajectory segments to obtain multiple dynamic feature parameters;

[0126] Perform feature dimensionality reduction processing based on multiple dynamic feature parameters to generate a multi-dimensional driving feature vector;

[0127] Perform clustering analysis on the multi-dimensional driving feature vector to obtain driving behavior classification data;

[0128] Perform spatio-temporal correlation analysis on the multi-dimensional driving feature vector to generate spatio-temporal correlation features;

[0129] Construct an electric vehicle driving behavior model based on the driving behavior classification data and spatio-temporal correlation features.

[0130] In this embodiment, the high-precision electric vehicle position trajectory is segmented at fixed time intervals (such as 5 seconds, 10 seconds, etc.) or fixed distance intervals to obtain a series of trajectory segments. These trajectory segments represent the driving conditions of the electric vehicle at different time periods. For each trajectory segment, dynamic characteristic parameters of the electric vehicle are calculated, including: average speed, maximum speed, acceleration distribution, angular velocity (turning rate), and change in driving direction. These parameters comprehensively describe the motion state of the electric vehicle during this time period. Since the calculated dynamic characteristic parameters have a high dimension, dimensionality reduction processing is required. Methods such as principal component analysis (PCA) or autoencoders are used to map the high-dimensional features to a low-dimensional space, extract the most representative feature combinations, and form a multi-dimensional driving feature vector. Cluster analysis is performed on the generated multi-dimensional driving feature vector using clustering algorithms such as K-means and DBSCAN to group similar driving behaviors into one category. Through cluster analysis, typical driving behavior patterns are identified, such as straight-line driving at a constant speed, acceleration, deceleration, turning, sudden braking, etc., to form driving behavior classification data. The multi-dimensional driving feature vector is input into a spatio-temporal correlation analysis model. The spatio-temporal correlation analysis model adopts a recurrent neural network (RNN) or long short-term memory network (LSTM) architecture and can learn the time-dependent relationships in sequence data. Through spatio-temporal correlation analysis, correlation features of the electric vehicle driving behavior in the time and space dimensions are extracted, such as behavior transition probabilities and typical behavior sequences. The aforementioned driving behavior classification data and spatio-temporal correlation features are integrated into a unified behavior model. This model includes both the classification of static behaviors and the modeling of the temporal changes in behaviors, and can comprehensively describe the driving behavior characteristics of the electric vehicle.

[0131] In this embodiment, refer to Figure 4 , which is a schematic diagram of the detailed implementation step process for "performing a safety risk assessment on the driving behavior model of a two-wheeled electric vehicle to obtain a risk warning index; performing a grading process on the risk warning index to obtain a graded safety warning strategy". In this embodiment, its detailed implementation steps include:

[0132] Train the driving behavior model of the electric vehicle with historical data to obtain a safety risk assessment model;

[0133] Based on the safety risk assessment model, perform a risk prediction on real-time driving data to obtain a risk warning index;

[0134] Perform a threshold division process on the risk warning index to obtain multiple risk levels;

[0135] Develop corresponding warning and intervention strategies for different risk levels to obtain a graded safety warning strategy.

[0136] In this embodiment, a large amount of historical driving data is collected, including normal driving data and accident / hazard data. These data are labeled and classified into categories such as "safe", "potential risk", and "high risk". The labeled historical data is input into a machine learning model for training, and the model can adopt algorithms such as random forest, support vector machine, or deep neural network. Through training, the model learns the correlation between various driving behaviors and safety risks, and forms a safety risk assessment model. The real-time collected electric vehicle driving data is input into the trained safety risk assessment model. The model predicts possible safety risks based on the current driving behavior characteristics and outputs a risk warning index. This index is usually a value between 0 and 1, indicating the possibility of a safety problem caused by the current behavior. According to historical data analysis and expert experience, a threshold for the risk warning index is set, and the risk level is divided into multiple levels, such as low risk (0 - 0.3), medium risk (0.3 - 0.7), and high risk (0.7 - 1). Different risk levels correspond to different warning response mechanisms. For different risk levels, corresponding warning and intervention strategies are formulated. For low-risk situations, only a prompt message may be displayed on the APP interface; for medium-risk situations, the system will issue an audible warning and suggest deceleration; for high-risk situations, the system will trigger an audible and visual alarm and may take active intervention measures, such as limiting the maximum speed or assisting in braking. These classification strategies constitute a complete safety warning system.

[0137] In this embodiment, "obtaining surrounding road environment data; performing real-time semantic analysis on the surrounding road environment data to construct a road scene knowledge graph" includes the following steps:

[0138] Receiving real-time data of the surrounding road environment to obtain the surrounding road environment data;

[0139] Performing multi-modal feature extraction on the surrounding road environment data to generate environmental feature data;

[0140] Performing semantic segmentation analysis on the environmental feature data to generate road scene semantic labels;

[0141] Performing relationship reasoning modeling on the road scene semantic labels to construct a road scene knowledge graph.

[0142] In this embodiment, environmental data around electric vehicles is obtained through multiple channels, including: road network data provided by online map services, real-time traffic conditions released by traffic management departments, weather information provided by meteorological departments, and road condition information shared by other connected vehicles, etc. The system can also access urban Internet of Things facilities, such as intelligent traffic lights, roadside units, etc., to obtain richer environmental information. Multimodal feature extraction is performed on the collected environmental data to process different types of data sources. For map data, features such as road network topology, road type, intersection location, etc. are extracted; for traffic condition data, features such as congestion level, average vehicle speed, traffic accident information, etc. are extracted; for weather data, features such as temperature, precipitation, visibility, etc. are extracted. These multimodal features together constitute an environmental feature dataset. Semantic segmentation technology is used to deeply analyze the environmental feature data to identify each element and its semantic meaning in the road scene. For example, roads are classified into motor lanes, non-motor lanes, and sidewalks; facilities such as traffic signs, traffic lights, and roadblocks are identified; special areas such as congestion areas, construction areas, and accident areas are marked. These semantic labels provide a basis for subsequent scene understanding. Based on the semantic labels, a knowledge graph of the road scene is constructed. This knowledge graph not only contains each element in the scene but also the relationships between elements, such as "an intersection connects two roads", "a non-motor lane is located next to a motor lane", "there is a traffic light 100 meters ahead", etc.

[0143] Specifically, integrate the semantic labels of the road scene from semantic segmentation analysis, standardize the labels, unify the label formats and meanings from different sources, remove redundant and low-confidence semantic labels, extract key entities from the semantic labels, such as roads, intersections, traffic signs, lanes, etc., assign a unique identifier to each entity, record its spatial location, type, and attributes, and construct an entity hierarchy (e.g., a main road contains multiple lanes); analyze the spatial relationships between entities (such as adjacent, containment, connection, etc.), identify functional relationships (such as a traffic light controls a certain intersection), establish temporal relationships (such as the start and end times of a construction area), and formalize these relationships as edges in the knowledge graph. Construct a connectivity graph of the road network to represent how road segments are connected, define turning restrictions and traffic rules at intersections, establish a multi-level road network structure (expressways, main roads, secondary roads, etc.), add attribute information to entities and relationships (such as the speed limit value and road condition status of a road), define inference rules (such as "if it is raining and the road is a gravel road, then it is defined as a high-risk road section"), establish spatio-temporal constraint rules (such as "the congestion probability of this road section is high during the morning and evening rush hours on weekdays"), store the constructed knowledge graph in the format of a graph database, establish a spatial index and a semantic index to support efficient query, and implement an incremental update mechanism to support the dynamic integration of real-time data. Through this structured knowledge representation, the system can deeply understand the road environment and provide support for navigation decisions.

[0144] In this embodiment, the specific steps of "performing semantic segmentation analysis on environmental feature data to generate road scene semantic labels" are as follows:

[0145] Identify road types, lane lines, traffic signs, and traffic lights based on environmental feature data;

[0146] Evaluate the confidence of the recognition results and screen out the recognition results with high confidence;

[0147] Perform road topology structure analysis based on the high-confidence recognition results to obtain road network structure data;

[0148] Evaluate the traffic flow status of the road network structure data to obtain real-time traffic status data;

[0149] Perform trend prediction on the real-time traffic status data based on historical data to obtain traffic flow prediction results;

[0150] Perform semantic annotation on the road network structure data and traffic flow prediction results to generate road scene semantic labels.

[0151] In this embodiment, computer vision and pattern recognition technologies are used to identify key road elements from environmental feature data. For road types, distinguish main roads, secondary roads, branch roads, non-motor vehicle lanes, etc.; for lane lines, identify solid lines, dashed lines, guiding arrows, etc.; for traffic signs, identify various signs such as speed limits, prohibitions, warnings, and instructions; for traffic lights, identify their positions, states, and control areas. Calculate a confidence score for each recognition result to reflect the reliability of the recognition.

[0152] Specifically, for object detection models based on CNN / YOLO, etc., the model directly outputs the preliminary confidence scores of the detection boxes. The initial confidence value = P(Object) × IoU, where P(Object) is the probability that the box contains the target, and IoU is the intersection over union of the predicted box and the ground truth box. Output probability values for each category (such as "speed limit sign", "dashed line"), that is, category probabilities; then calculate the comprehensive confidence = initial confidence value × category probability;

[0153] Define a semantic segmentation model. For each pixel, the model outputs the probability distribution belonging to each category. The regional confidence = the average category probability of all pixels in the target area, or use a weighted method: regional confidence = ∑(category probability of pixel j × weight corresponding to pixel j) / ∑weights corresponding to pixel j;

[0154] Calibrate the original confidence using Platt scaling or isotonic regression. After calibration, the confidence = f(original confidence), where f is the calibration function, and the calibration function is learned based on the true detection performance on the validation set; evaluate the clarity, integrity, occlusion degree, etc. of the target, and then calculate the clarity score = the weighted sum of the contrast and gradient intensity of the target area; integrity score = the integrity ratio of the target area (the untruncated or unoccluded part); occlusion score = 1 - occlusion ratio;

[0155] Consider factors such as lighting conditions, weather conditions, viewing angles, etc. The lighting score = the suitability score calculated based on the luminance histogram of the target area; the weather impact coefficient = the visibility assessment based on the current weather conditions (rain, fog, snow, etc.); then calculate the adjusted confidence (confidence score) = calibrated confidence × feature quality coefficient × environmental factor coefficient;

[0156] The feature quality coefficient = w1 × clarity score + w2 × integrity score + w3 × occlusion score;

[0157] The environmental factor coefficient = w4 × lighting score × w5 × weather impact coefficient; w1 to w5 are the corresponding weight coefficients.

[0158] Set a confidence threshold (such as 0.85), and only retain the recognition results higher than the threshold to ensure the accuracy of subsequent analysis. For elements with low confidence but high importance, the system may obtain supplementary information from other data sources (such as map data). Based on the filtered high-confidence recognition results, analyze the topological structure of the road network. Construct a road connection relationship graph, including information such as the connection method between road segments, intersection types, and turning restrictions. This topological structure analysis can provide basic data support for path planning. Combine real-time traffic data and historical statistical data to evaluate the traffic flow status of the current road network. Calculate the congestion index, average vehicle speed, and travel time of each road segment to form real-time traffic status data. These data reflect the real-time operation of the road network. Use time series analysis and machine learning methods to perform short-term prediction on the real-time traffic status. Predict the traffic flow change trend in the next 15 minutes, 30 minutes, or even longer, providing a forward-looking reference for path planning. Label and integrate the road network structure data and traffic flow prediction results at the semantic level. For example, label "high congestion risk road segments", "road segments suitable for electric vehicles", "routes with high safety factors", etc. These semantic labels enable the system to understand the road environment at a higher level and provide semantic support for subsequent navigation decisions.

[0159] In this embodiment, "perform dynamic obstacle recognition on the road scene knowledge graph, make path planning decisions, and construct an intelligent navigation strategy" includes the following steps:

[0160] Perform real-time queries on the road scene knowledge graph to extract surrounding potential obstacle data;

[0161] Perform target path planning for the current position of the electric vehicle to obtain the expected driving path;

[0162] Based on the surrounding potential obstacle data, perform dynamic obstacle recognition and tracking on the expected driving path to obtain obstacle risk assessment data;

[0163] Calculate the collision risk for the obstacle risk assessment data to generate a collision risk assessment value;

[0164] Based on the collision risk assessment value, make real-time path optimization decisions to construct an intelligent navigation strategy.

[0165] In this embodiment, the system performs real-time queries on the constructed road scene knowledge graph to obtain structured information about the surrounding environment of the electric vehicle. Potential obstacle information that may affect the driving safety of the electric vehicle is extracted from the knowledge graph, such as road construction areas, temporary obstacles, congested sections, etc. This information forms the surrounding potential obstacle data, providing input for subsequent analysis. Based on the current position and destination of the electric vehicle, the system performs preliminary path planning. The planning process considers various factors, such as the shortest distance, the least time consumption, and a high safety factor. The system may generate multiple candidate paths and select the optimal path as the expected driving path according to user preferences and the current situation. Along the expected driving path, the system performs dynamic obstacle recognition and tracking. These obstacles include static obstacles (such as roadblocks, potholes) and dynamic obstacles (such as pedestrians, other vehicles). The system uses computer vision and sensor fusion technology to detect these obstacles in real time and predict their movement trajectories. Through this analysis, the system generates obstacle risk assessment data, describing the positions, movement characteristics, and potential risks of each obstacle. Based on the obstacle risk assessment data, the system calculates the collision risk between the electric vehicle and each obstacle. The calculation considers factors such as relative distance, relative speed, and movement direction to generate a quantified collision risk assessment value.

[0166] Specifically, based on the prediction, predict its movement trajectory, generate the spatio-temporal trajectory probability distribution of the electric vehicle, reflecting the uncertainty of position prediction; establish a time-space grid, divide the future time into several time segments, and for each time segment, calculate the spatial overlap probability between the electric vehicle and each obstacle. Specifically, simplify the electric vehicle and the obstacle into a rectangle or an ellipse, which is represented by the center point position, the length, width and orientation angle; use a two-dimensional Gaussian distribution to represent the uncertainty of the positions of the electric vehicle and the obstacle, which is described by the mean vector and the covariance matrix. Taking the rectangle as an example, calculate the projection intervals of the corresponding two rectangles on each projection axis. If there is an overlap on all axes, the two objects overlap, and the overlap index = min(overlap length / projection length of rectangle 1 on this axis, overlap length / projection length of rectangle 2 on this axis); then calculate the collision probability = Σ(probability of the electric vehicle position in the time segment × probability of the obstacle position × spatial overlap index).

[0167] Combine the collision probability with the collision severity to calculate the collision risk assessment value. The collision severity considers factors such as relative speed, collision angle, and obstacle mass / hardness (quantified weighted summation); the collision risk assessment value = collision probability × collision severity.

[0168] This collision risk assessment method based on a probability model can comprehensively consider the relative motion relationship between the electric vehicle and the obstacle, and provide a decision-making basis for the intelligent navigation system through the quantified risk assessment value, effectively improving the driving safety of the electric vehicle.

[0169] This assessment value reflects the possibility and severity of a collision between the electric vehicle and the obstacle. According to the calculated collision risk assessment value, the system performs real-time path optimization. For high-risk areas, the system will plan a detour path; for medium-risk areas, the system may recommend reducing speed to pass through; for low-risk areas, the system will maintain the original path. Through this dynamic optimization, the system constructs an intelligent navigation strategy to provide safe and efficient route guidance for the rider.

[0170] In this embodiment, "adaptive control of the two-wheeled electric vehicle based on a hierarchical safety warning strategy and an intelligent navigation strategy" includes the following steps:

[0171] Construct an intelligent interface for the electric vehicle control system to generate an intelligent control framework;

[0172] Based on the hierarchical safety warning strategy, map the safety intervention rules to the intelligent control framework to construct a safety intervention mechanism;

[0173] Based on the intelligent navigation strategy, map the navigation assistance rules to the intelligent control framework to construct a navigation assistance mechanism;

[0174] Coordinate the priorities of the safety intervention mechanism and the navigation assistance mechanism to construct an electric vehicle;

[0175] The priority coordination is specifically as follows: identifying the safety risk level based on the current situation;

[0176] The safety risk levels include: emergency risk, medium risk, and low risk;

[0177] When the safety risk level is an emergency risk, the priority of the safety intervention mechanism is higher than that of the navigation assistance mechanism, and emergency braking or steering assistance is executed;

[0178] When the safety risk level is a medium risk, the safety intervention mechanism and the navigation assistance mechanism work together, execute deceleration reminders and provide suggestions for risk avoidance paths;

[0179] When the safety risk level is a low risk, the priority of the navigation assistance mechanism is higher than that of the safety intervention mechanism, and the optimal path navigation is normally executed.

[0180] In this embodiment, an intelligent interface layer of the electric vehicle control system is designed and implemented, including a hardware interface and a software interface. The hardware interface is connected to the key control units of the electric vehicle, such as the motor controller, braking system, display screen, etc.; the software interface provides standardized APIs, allowing high-level applications to interact with the underlying control system. Based on these interfaces, a modular and extensible intelligent control framework is constructed, providing a basic platform for subsequent safety intervention and navigation assistance functions. Map the hierarchical safety warning strategy obtained above into the intelligent control framework to form specific safety intervention rules. For example, when a high-risk situation is detected, the system may trigger motor power limitation or assist braking; when fatigue riding characteristics are detected, the system may issue a rest reminder, etc. These rules constitute the safety intervention mechanism of the electric vehicle, which can provide timely protection in dangerous situations. Translate the intelligent navigation strategy into specific navigation assistance rules and map them into the intelligent control framework. These rules define how the system provides navigation guidance, such as voice prompts, display screen instructions, vibration reminders, etc. The navigation assistance mechanism not only provides route guidance but also provides speed suggestions and energy consumption optimization strategies according to road conditions and battery status. Design a priority coordination mechanism to handle possible conflicts between safety intervention and navigation assistance. The system evaluates the current situation in real time and classifies the risk level into three levels: emergency risk, medium risk, and low risk. In the case of emergency risk (such as detecting a possible collision), the safety intervention mechanism has the highest priority, and the system will immediately execute emergency braking or steering assistance, temporarily ignoring the navigation suggestions. In the case of medium risk (such as approaching a congested section), the safety intervention mechanism and the navigation assistance mechanism work together, and the system will remind the rider to slow down and provide a risk avoidance path suggestion at the same time. In the case of low risk (such as normal riding), the navigation assistance mechanism dominates, and the system provides navigation guidance according to the optimal path, and the safety intervention mechanism is in a monitoring state. Through this priority coordination mechanism, the system can make the most appropriate decisions in different situations, constructing a comprehensive and intelligent electric vehicle networking safety system.

[0181] In this embodiment, the safety intervention mechanism includes:

[0182] Monitor the speed of the electric vehicle in real time, and when an overspeed behavior is detected, issue an audible and visual alarm prompt;

[0183] Monitor the steering angle of the electric vehicle in real time, and when a sharp turn behavior is detected, perform active deceleration intervention;

[0184] Monitor the braking state of the electric vehicle in real time, and when an emergency braking demand is detected, optimize the braking force distribution;

[0185] Monitor the battery state of the electric vehicle in real time, and when an abnormal battery power is detected, automatically enter the low-power safe home mode.

[0186] The safety intervention mechanism ensures cycling safety in various ways. The system monitors the driving speed of the electric vehicle in real time through a speed sensor and compares it with the road speed limit information. When it detects that the speed of the electric vehicle exceeds the safety threshold or the road speed limit, the system will immediately trigger an audible and visual alarm, reminding the cyclist to pay attention to decelerating through sound prompts and screen flashing. In case of continuous speeding, the system may further limit the motor output power to force the vehicle speed to decrease. The system uses a gyroscope and an accelerometer to monitor the steering angle and angular velocity of the electric vehicle. When it detects that the cyclist makes a sharp turn, the system will automatically intervene and perform deceleration intervention, reducing the motor output power to ensure stability and safety during the turning process. This intervention is particularly important on slippery roads or during high-speed turns, effectively preventing skidding and rollover accidents. The system monitors the braking operation and braking force requirements of the cyclist. When it detects an emergency braking requirement (such as when the cyclist suddenly brakes hard), the system will optimize the braking force distribution to ensure a reasonable distribution of the front and rear wheel braking forces, preventing wheel lock-up and skidding. In some high-end models, the system may also be equipped with an ABS (antilock braking system) function to further improve safety during emergency braking. The system continuously monitors the battery power, temperature, and health status of the electric vehicle. When it detects that the battery power is abnormally low or the battery temperature is abnormally high, the system will automatically enter a low-power safe return home mode. In this mode, the system will limit the maximum speed and acceleration performance, prioritize ensuring basic driving functions, and plan the most energy-saving route to the nearest charging station to ensure that the cyclist can safely reach the destination or charging point.

[0187] In this embodiment, the navigation assistance mechanism includes:

[0188] Calculating the optimal path based on Beidou positioning data and map data, and providing real-time navigation guidance through a voice and display interface;

[0189] Dynamically adjusting the path planning according to the current road congestion condition to avoid traffic congestion areas;

[0190] Intelligently planning the optimal path passing through a charging station according to the remaining battery power of the electric vehicle;

[0191] Learning the user's preferred routes based on historical cycling data and providing personalized path recommendations;

[0192] Providing safety cycling suggestions and road condition warnings based on weather and road condition data.

[0193] The navigation assistance mechanism provides comprehensive navigation support for riders. The system combines Beidou high-precision positioning data with detailed map data to calculate the optimal path from the current location to the destination. During the calculation process, the system takes into account a variety of factors, such as distance, estimated time, road type, safety factor, etc. After the calculation is completed, the system provides riders with clear navigation instructions through voice prompts and display interfaces, including turn prompts, distance information, estimated arrival time, etc. The system receives traffic status data in real time to understand the congestion of each section of the road. When severe congestion is detected on the scheduled route, the system automatically replans the route to guide riders to bypass the congested area and choose a more unobstructed alternative route. This dynamic adjustment can effectively reduce riding time and improve travel efficiency. The system monitors the battery power of the electric vehicle and estimates the drivable distance based on the remaining power and energy consumption model. When the system determines that the remaining power may not be able to support the completion of the entire journey, it will automatically plan a route passing through a charging station. The system will select the optimal charging station location to minimize the total travel time or the total energy consumption, ensuring that riders can complete charging and continue their journey safely and conveniently. The system records and analyzes the user's historical riding data to understand the user's route preferences, frequented places, and habitual travel times. Based on this data, the system learns the user's personal preferences and provides personalized route recommendations. For example, the system may prioritize routes that users frequently choose, or adjust the route planning algorithm based on the user's riding style (such as preference for flat roads or scenic roads). The system integrates weather forecasts and road condition data to provide cyclists with safe riding advice. In rainy and snowy weather or slippery roads, the system will prompt cyclists to slow down; in low visibility conditions, the system will recommend turning on the lights and keeping a safe distance; when road construction or accidents are detected ahead, the system will issue a road condition warning in advance to guide cyclists to pass safely or choose a detour.

[0194] Through the collaborative work of the above-mentioned safety intervention mechanism and navigation assistance mechanism, the present invention constructs a comprehensive and intelligent electric vehicle networking safety system. The system can not only provide accurate navigation guidance, but also actively intervene in dangerous situations to fully protect the safety of riders. The modular design and priority coordination mechanism of the system ensure that each functional module can make the most reasonable decision according to the actual situation, providing electric vehicle users with a safe, convenient and intelligent riding experience.

[0195] The intelligent connection system for electric vehicles based on Beidou positioning of the present invention significantly improves the Beidou positioning accuracy through multi-source fusion calibration technology, and solves the problem of inaccurate positioning of traditional electric vehicles in complex urban environments; through dynamic feature extraction and behavior modeling, it realizes the intelligent analysis of riding behaviors and provides a reliable basis for safety warnings; through real-time semantic analysis and knowledge graph construction, it realizes the in-depth understanding of road environments and lays a foundation for intelligent navigation; through hierarchical safety warnings and dynamic route planning, it significantly improves the riding safety and travel efficiency of electric vehicles; through an adaptive control system, it realizes the intelligent control of human-machine collaboration and greatly enhances the user experience.

[0196] The present invention is not only applicable to two-wheeled electric vehicles, but also can be extended to various types of small electric transportation tools such as electric bicycles and electric motorcycles. The modular design of the system enables it to have good scalability and adaptability, and can be customized according to the characteristics of different vehicle models.

[0197] Compared with the prior art, the present invention has significant innovation and progressiveness. Traditional electric vehicle navigation systems are mostly based on simple GPS positioning, with low accuracy and lack of in-depth analysis of riding behaviors and road environments; while the present invention realizes high-precision positioning and intelligent environment perception through the fusion of Beidou positioning and multi-source data, greatly improving the system performance. Traditional systems are mostly passive prompt types and lack the ability of active intervention; while the present invention can perform active intervention in dangerous situations through a safety intervention mechanism, effectively preventing accidents. Traditional systems mostly work independently and lack cooperation with other systems; while the present invention realizes information sharing and collaborative decision-making with traffic infrastructure and other vehicles through a knowledge graph and intelligent networking, laying a foundation for future intelligent transportation.

[0198] It should be noted that the above formulas are all dimensionless and take their numerical values for calculation. The formulas are obtained by collecting a large amount of data for software simulation to obtain a formula closest to the actual situation. The preset parameters and threshold selection in the formulas are set by those skilled in the art according to the actual situation.

[0199] Therefore, from any perspective, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, it is intended to cover all changes falling within the meaning and scope of the equivalent elements of the application documents within the present invention.

[0200] The above are only specific embodiments of the present invention, enabling those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but rather to the broadest scope consistent with the principles and novel features invented herein.

Claims

1. The electric vehicle intelligent connection system based on Beidou positioning is characterized by: include: High-precision positioning module, used to obtain Beidou positioning data and driving status data of two-wheeled electric vehicles; Perform multi-source fusion correction on Beidou positioning data and perform trajectory optimization processing to obtain a high-precision two-wheeled electric vehicle position trajectory; The behavior model fitting module is used to extract dynamic features of the high-precision two-wheeled electric vehicle position trajectory and generate a multi-dimensional driving feature vector; Performing spatiotemporal correlation analysis on multi-dimensional driving feature vectors to generate a two-wheeled electric vehicle driving behavior model; The safety strategy assessment module is used to conduct safety risk assessment on the driving behavior model of two-wheeled electric vehicles and obtain a risk warning index; The risk warning index is graded to obtain a graded safety warning strategy; The knowledge graph construction module is used to obtain surrounding road environment data; perform real-time semantic analysis on the surrounding road environment data and construct a road scene knowledge graph; The navigation strategy fitting module is used to dynamically identify obstacles on the road scene knowledge graph, make path planning decisions, and build an intelligent navigation strategy; The adaptive control module performs adaptive control on the two-wheeled electric vehicle based on the hierarchical safety warning strategy and intelligent navigation strategy.

2. The electric vehicle intelligent connection system based on Beidou positioning according to claim 1 is characterized in that: The method of obtaining Beidou positioning data and driving status data of the two-wheeled electric vehicle; performing multi-source fusion correction on the Beidou positioning data and performing trajectory optimization processing to obtain a high-precision position trajectory of the two-wheeled electric vehicle includes: Obtain Beidou positioning data and driving status data of two-wheeled electric vehicles; Evaluate the signal quality of Beidou positioning data and identify positioning points in weak signal areas; Perform inertial navigation compensation on positioning points in weak signal areas to eliminate signal blocking errors; Combine wheel speed sensor data to perform odometer-assisted correction on Beidou positioning data, thereby obtaining corrected positioning data; Perform Kalman filter smoothing on the corrected positioning data to generate a smooth positioning trajectory; The smooth positioning trajectory is subjected to map matching and trajectory optimization processing to obtain a high-precision position trajectory of the two-wheeled electric vehicle.

3. The electric vehicle intelligent connection system based on Beidou positioning according to claim 2 is characterized in that: The performing Kalman filter smoothing processing on the corrected positioning data comprises: Perform noise characteristic analysis on the corrected positioning data to obtain a positioning noise model; Design the adaptive Kalman filter parameters according to the positioning noise model to obtain the optimized filter parameters; Applying a Kalman filter with optimized filter parameters to the corrected positioning data to generate a preliminary smoothed trajectory; Perform outlier detection on the initial smooth trajectory and mark the trajectory jump points; Based on historical trajectory data, the trajectory jump points are corrected and interpolated to generate a continuous and smooth trajectory; The continuous smooth trajectory is processed by time window sliding average to generate a smooth positioning trajectory.

4. The electric vehicle intelligent connection system based on Beidou positioning according to claim 3 is characterized in that: The dynamic feature extraction of the high-precision two-wheeled electric vehicle position trajectory is performed to generate a multi-dimensional driving feature vector; The multi-dimensional driving feature vector is analyzed in time and space to generate a two-wheeled electric vehicle driving behavior model, including: The high-precision two-wheeled electric vehicle position trajectory is segmented into time windows to generate multiple trajectory segments; Calculate the average speed, maximum speed, acceleration distribution, angular velocity and driving direction change of multiple trajectory segments to obtain multiple dynamic characteristic parameters; Perform feature dimension reduction processing based on multiple dynamic feature parameters to generate a multi-dimensional driving feature vector; Perform cluster analysis on multi-dimensional driving feature vectors to obtain driving behavior classification data; Performing spatiotemporal correlation analysis on the multi-dimensional driving feature vector to generate spatiotemporal correlation features; A two-wheeled electric vehicle driving behavior model is constructed based on driving behavior classification data and spatiotemporal correlation characteristics.

5. The electric vehicle intelligent connection system based on Beidou positioning according to claim 4 is characterized in that: The safety risk assessment of the two-wheeled electric vehicle driving behavior model is performed to obtain a risk warning index; The risk warning index is graded to obtain a graded safety warning strategy, including: The historical data of the two-wheeled electric vehicle driving behavior model is trained to obtain a safety risk assessment model; Based on the safety risk assessment model, risk prediction is performed on real-time driving data to obtain a risk warning index; The risk warning index is divided into thresholds to obtain multi-level risk levels; Formulate corresponding early warning and intervention strategies for different risk levels, thus obtaining a graded safety early warning strategy.

6. The electric vehicle intelligent connection system based on Beidou positioning according to claim 5 is characterized in that: The obtaining of surrounding road environment data; performing real-time semantic analysis on the surrounding road environment data and constructing a road scene knowledge graph includes: Receive real-time data of surrounding road environment and obtain surrounding road environment data; Perform multimodal feature extraction on surrounding road environment data to generate environmental feature data; Perform semantic segmentation analysis on environmental feature data to generate road scene semantic labels; Perform relational reasoning modeling on road scene semantic labels and construct a road scene knowledge graph.

7. The electric vehicle intelligent connection system based on Beidou positioning according to claim 6 is characterized in that: The performing semantic segmentation analysis on the environmental feature data to generate a road scene semantic label includes: Identify road types, lane lines, traffic signs and signal lights based on environmental feature data, i.e., recognition results; Conduct confidence assessment on the recognition results and select high-confidence recognition results; Perform road topology analysis based on high-confidence recognition results to obtain road network structure data; Evaluate the traffic flow status of the road network structure data to obtain real-time traffic status data; Based on historical data, the trend of real-time traffic status data is predicted to obtain traffic flow prediction results; Semantic annotation is performed on road network structure data and traffic flow prediction results to generate road scene semantic labels.

8. The electric vehicle intelligent connection system based on Beidou positioning according to claim 7 is characterized in that: The method of dynamically identifying obstacles on the road scene knowledge graph, making path planning decisions, and building an intelligent navigation strategy includes: Conduct real-time queries on the road scene knowledge graph to extract surrounding potential obstacle data; Perform target path planning for the current position of the two-wheeled electric vehicle to obtain the expected driving path; Based on the surrounding potential obstacle data, dynamic obstacle identification and tracking are performed on the expected driving path to obtain obstacle risk assessment data; Calculate the collision risk of the obstacle risk assessment data to generate a collision risk assessment value; Make real-time path optimization decisions based on collision risk assessment values ​​and build intelligent navigation strategies.

9. The electric vehicle intelligent connection system based on Beidou positioning according to claim 8 is characterized in that: The adaptive control of the two-wheeled electric vehicle based on the hierarchical safety warning strategy and the intelligent navigation strategy includes: Construct intelligent interfaces for the control system of two-wheeled electric vehicles and generate intelligent control frameworks; Based on the hierarchical security early warning strategy, the security intervention rules of the intelligent control framework are mapped to build a security intervention mechanism; Based on the intelligent navigation strategy, the navigation assistance rules are mapped to the intelligent control framework to build a navigation assistance mechanism; Coordinate priorities for safety intervention mechanisms and navigation assistance mechanisms; The priority coordination specifically includes: identifying the security risk level based on the current situation; The security risk levels include: urgent risk, medium risk and low risk; When the safety risk level is an emergency risk, the safety intervention mechanism takes precedence over the navigation assistance mechanism and performs emergency braking or steering assistance; When the safety risk level is medium, the safety intervention mechanism works together with the navigation assistance mechanism to execute deceleration reminders and provide risk avoidance path suggestions; When the safety risk level is low, the navigation assistance mechanism takes priority over the safety intervention mechanism, and the optimal path navigation is performed normally.

Images