Vehicle networking data new energy automatic driving vehicle traffic behavior supervision and service system

Abstract

The invention discloses an Internet of Vehicles data new energy automatic driving vehicle traffic behavior supervision and service system, which comprises a data acquisition layer, a network transmission layer, a data processing and analysis layer and an application service layer, and adopts an end-edge-cloud collaborative architecture. The data acquisition layer comprehensively acquires multi-source data of vehicles, roads, traffic and the like; the network transmission layer ensures safe transmission of data through various communication technologies and encryption means; the data processing and analysis layer realizes data intelligent analysis by means of edge computing and a cloud platform in combination with an AI algorithm model; and the application service layer provides differentiated services for governments, enterprises and users. According to the method, data islands are broken, the transformation from post-event tracing to beforehand early warning and in-event intervention of a supervision mode is realized, the driving behavior analysis scientificity, the safety early warning perspectiveness and the carbon emission management accuracy are improved, the government is helped to improve the treatment efficiency, the enterprise is helped to reduce the cost and improve the efficiency, the user is helped to improve the travel experience, and the method has remarkable technical, social and economic values.

Description

Technical Field

[0001] This invention relates to the fields of vehicle networking, new energy vehicles and autonomous driving technology, specifically a vehicle networking data traffic behavior monitoring and service system for new energy autonomous vehicles. Background Technology

[0002] With the rapid development of new energy vehicles and autonomous driving technologies, the Internet of Vehicles (IoV), as a core supporting technology, is driving the transformation of the transportation sector towards intelligence and connectivity. my country has established a three-tiered (national, local, and enterprise) big data supervision platform system for new energy vehicles, connecting over 11 million vehicles and enabling the collection and interconnection of data on vehicle operating status, battery information, and driving behavior.

[0003] The system relies on several key technologies for operation: In terms of data acquisition and communication transmission, data is collected through vehicle-mounted terminals (T-Box), roadside units (RSU), and multi-source sensors, and 5G / C-V2X and other communication technologies and encryption methods are used to ensure transmission security; in terms of big data processing and analysis, distributed storage and stream processing technologies are used, combined with AI and machine learning algorithms to achieve functions such as driving behavior analysis and safety hazard prediction; cloud-edge-device collaborative computing optimizes resource allocation and reduces response latency; and data security and privacy protection are achieved through encrypted transmission, blockchain and other technologies to build a protective system.

[0004] Based on the above technologies, the system has been applied to scenarios such as product quality supervision and traceability, proactive prevention and control of operational safety, driving behavior analysis and evaluation, precise management and traceability of carbon emissions, and optimization of public services and intelligent transportation.

[0005] However, existing technologies still have many shortcomings: First, data silos and barriers are prominent issues, with inconsistent data standards across departments, platforms, and car manufacturers, and an interoperability and sharing rate of less than 40%. Unclear data ownership and a lack of benefit distribution mechanisms make data integration difficult. Second, there are risks of data security and privacy leaks. Sensitive data such as high-precision geographic information and user personal information are at risk of excessive collection, misuse, or leakage, and security standards and regulatory mechanisms are inadequate. Third, there are limitations in technological maturity and reliability. Sensor accuracy and communication reliability decline in harsh environments, AI model generalization ability is insufficient, and key technologies are highly dependent on external sources. Fourth, laws, regulations, and standards are lagging behind. There are gaps in the determination of liability for autonomous driving accidents, and standards for data interfaces and communication protocols are not uniform. Regulation lags behind technological innovation. Fifth, there are infrastructure and cost challenges. The coverage of intelligent infrastructure is insufficient and uneven, and the costs of transformation and maintenance are high. Sixth, there is a crisis of social acceptance and trust. The public has doubts about data privacy and the safety of autonomous driving, AI algorithm decision-making is not transparent, and a digital divide exists.

[0006] Therefore, it is necessary to propose a vehicle-to-everything (V2X) data-based traffic behavior monitoring and service system for new energy autonomous vehicles to address the aforementioned technical issues. Summary of the Invention

[0007] The purpose of this invention is to provide a vehicle-to-everything (V2X) data-based traffic behavior monitoring and service system for new energy autonomous vehicles, thereby solving the aforementioned technical problems.

[0008] To achieve the above objectives, the present invention provides the following technical solution:

[0009] The Internet of Vehicles (IoV) data-driven traffic behavior monitoring and service system for new energy and autonomous vehicles includes a data acquisition layer, a network transmission layer, a data processing and analysis layer, and an application service layer that are connected in sequence. The data acquisition layer, network transmission layer, and data processing and analysis layer adopt an edge-cloud collaborative architecture to achieve real-time monitoring, safety warnings, behavior analysis, and efficient services for new energy and autonomous vehicles.

[0010] The data acquisition layer is used to collect raw data on vehicle status, road environment, traffic events, and supplementary data sources.

[0011] The network transport layer is used to achieve low-latency, highly reliable transmission and secure encryption of data between layers;

[0012] The data processing and analysis layer is used to store, clean, process in real time, and perform intelligent analysis on the collected data to generate the analysis results required for supervision and services.

[0013] The application service layer is used to provide targeted supervision, operation and personalized services to different users.

[0014] The data acquisition layer includes vehicle-mounted terminals, roadside units and roadside sensing devices, and other data sources. The vehicle-mounted terminal is a T-Box / OBU integrated inside the vehicle, which is connected to the vehicle's power system, battery management system, autonomous driving domain controller, GPS / BeiDou positioning module, and inertial measurement unit via CAN bus, dedicated wiring harness or wireless interface to collect vehicle speed, acceleration, latitude and longitude, heading angle, turn signal status, braking status, battery data and fault codes.

[0015] The roadside units and roadside sensing devices are deployed at key road nodes. The roadside sensing devices include cameras, lidar, millimeter-wave radar, and weather sensors, which are connected to the roadside computing unit (MEC) via cables or optical fibers to collect data on traffic flow, events, pedestrians, non-motorized vehicles, and weather conditions. The roadside unit (RSU) is connected to the MEC via a wired network and communicates with the on-board unit (OBU) and cloud platform via a wireless network. Other data sources include charging piles and traffic signal control systems. Charging piles upload charging data via Ethernet or wireless networks, and the traffic signal control system sends traffic light status information via a dedicated protocol.

[0016] The network transmission layer adopts a transmission method combining vehicle-to-everything (V2X) communication technology and wide-area communication technology. The V2X communication technology is C-V2X, including LTE-V2X and 5G-V2X, or the DSRC protocol, used for short-range, low-latency, and highly reliable direct communication between the vehicle-mounted OBU and the roadside RSU to transmit safety warning information. The wide-area communication technology is 5G / 4G or Ethernet, used for communication between the vehicle-mounted T-Box and the roadside MEC unit and the cloud platform to transmit monitoring and non-real-time data. The data transmission process adopts TLS / SSL encryption protocol and authentication mechanism.

[0017] The data processing and analysis layer includes a cloud-based control platform and edge computing nodes (MECs). The edge computing nodes are deployed in a data center near the roadside and connect to the RSUs and sensors via a high-speed network. They process latency-sensitive computing tasks, including obstacle recognition, traffic light status interpretation, and cooperative collision avoidance decision-making, and send the results to the vehicles via the RSUs. The cloud-based control platform adopts a distributed architecture, uses Hadoop HDFS for massive data storage, and employs Apache Kafka, Spark Streaming, or Flink stream processing frameworks for real-time data processing. It also incorporates AI algorithm models for intelligent analysis of the cleaned data.

[0018] The AI ​​algorithm model includes a driving behavior analysis model, a safety warning model, and an energy consumption assessment and carbon footprint tracking model. The driving behavior analysis model identifies undesirable driving behaviors such as rapid acceleration, rapid deceleration, and sharp turns by analyzing time-series data of vehicle acceleration, deceleration, and steering angular velocity. The safety warning model predicts accident risks based on historical data and real-time traffic conditions, or receives abnormal event information from roadside units through vehicle-road cooperation and issues warnings, with an accuracy rate of over 80% in predicting major accidents seven days in advance. The energy consumption assessment and carbon footprint tracking model dynamically analyzes the energy consumption data of new energy vehicles and calculates carbon emissions using the formula carbon emissions = ∫(real-time power × real-time carbon intensity) dt.

[0019] The application service layer includes a government regulatory platform, an enterprise operation platform, and a user service platform. The government regulatory platform provides traffic management, industry and information technology departments with web-based or large-screen visualization interfaces to achieve functions such as vehicle operation safety monitoring, traffic violation evidence collection, carbon emission supervision, and emergency plan management. The enterprise operation platform provides SaaS services to car manufacturers, logistics companies, taxi companies, etc., and outputs data analysis reports on vehicle condition, driving behavior, energy consumption, and efficiency through API interfaces or web portals for fleet management, maintenance warnings, insurance assessments, and optimized dispatching. The user service platform provides drivers with real-time traffic conditions, dangerous road section warnings, personalized driving scores, charging pile recommendations and reservations, and green travel points services through mobile apps or vehicle-mounted applications.

[0020] The Internet of Vehicles (IoV) data-driven new energy autonomous vehicle traffic behavior monitoring and service system also includes a data security and privacy protection module. It adopts encrypted data transmission and blockchain technology to ensure data authenticity and tamper-proof, constructs a three-layer information security transmission architecture of terminal-vehicle-cloud, and establishes a transparent data use and notification consent mechanism.

[0021] Preferably, the sensing technology route of the roadside sensing device and the vehicle terminal adopts a combination of single-vehicle intelligence and vehicle-road cooperation mode, or chooses to adopt a mode with vehicle-side intelligence as the main focus and roadside intelligence as the auxiliary focus, or a mode with roadside intelligence as the main focus and vehicle-side intelligence as the auxiliary focus.

[0022] The cloud control platform's storage architecture adopts a centralized cloud, regional private cloud, or hybrid cloud deployment mode. In the hybrid cloud mode, sensitive data is stored locally while simultaneously utilizing public cloud elastic computing resources.

[0023] The regulatory models of government-regulated platforms include government-led construction of city-level or national-level regulatory platforms for enterprises to access; or enterprises building their own platforms and accepting government regulation, with the latter providing interfaces to autonomous driving companies, integrating data, and supporting their own operations and government regulation.

[0024] The present invention has the following beneficial effects:

[0025] 1. Regulatory efficiency has undergone a fundamental transformation, shifting from traditional post-event tracing to pre-event warning and in-event intervention. The accuracy rate of early warning for major accidents, which is 7 days in advance, exceeds 80%. After adopting this system, the traffic management department of a certain city has improved the efficiency of inspecting violations by ride-hailing vehicles and freight vehicles by 50% and reduced labor costs by 40%.

[0026] 2. The technology has achieved a qualitative leap in effectiveness. A driving behavior scoring model based on hundreds of feature parameters has been established to quantitatively assess drivers' safety awareness and operating habits. After a commercial vehicle management system was connected to the model, the overall accident rate of the fleet decreased by 25%, and vehicle energy consumption decreased by 8%. By dynamically collecting real-time energy consumption data, accurate carbon emission tracking has been achieved, providing a reliable data foundation for carbon trading. The cloud-edge-device collaborative architecture reduces response latency and improves system processing efficiency.

[0027] 3. Significant social and economic value: For government regulators, it innovates the online + offline supervision model, improving governance efficiency and scientific decision-making capabilities; for operating companies, it reduces maintenance costs through predictive maintenance, optimizes insurance costs based on driving behavior scores (safe drivers can receive up to 30% premium discounts), and improves product design with the help of real operational data; for end users, it provides personalized safety services, enhancing travel safety and experience.

[0028] 4. Break down information silos and achieve cross-departmental, cross-level, and cross-enterprise data sharing and business collaboration through unified data standards and a three-tiered platform system at the national, local, and enterprise levels. This enables unified network dispatch and command during major event security operations. Attached Figure Description

[0029] Figure 1 This is a block diagram illustrating the component principle of the present invention. Detailed Implementation

[0030] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0031] The embodiments of the present invention will be described in detail below with reference to the accompanying drawings, but the present invention can be implemented in many different ways as defined and covered by the claims.

[0032] The embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0033] This embodiment provides a technical solution:

[0034] like Figure 1 As shown, the Internet of Vehicles (IoV) data-based traffic behavior monitoring and service system for new energy autonomous vehicles includes a data acquisition layer, a network transmission layer, a data processing and analysis layer, and an application service layer that are connected in sequence. The data acquisition layer, network transmission layer, and data processing and analysis layer adopt an end-edge-cloud collaborative architecture to realize real-time monitoring, safety warning, behavior analysis, and efficient services for new energy and autonomous vehicles.

[0035] The data acquisition layer is used to collect raw data on vehicle status, road environment, traffic events, and supplementary data sources.

[0036] The network transport layer is used to achieve low-latency, highly reliable transmission and secure encryption of data between layers;

[0037] The data processing and analysis layer is used to store, clean, process in real time, and perform intelligent analysis on the collected data to generate the analysis results required for supervision and services.

[0038] The application service layer is used to provide targeted supervision, operation and personalized services to different users.

[0039] The data acquisition layer includes vehicle-mounted terminals, roadside units and roadside sensing devices, and other data sources. The vehicle-mounted terminal is a T-Box / OBU integrated inside the vehicle, which is connected to the vehicle's power system, battery management system, autonomous driving domain controller, GPS / BeiDou positioning module, and inertial measurement unit via CAN bus, dedicated wiring harness or wireless interface to collect vehicle speed, acceleration, latitude and longitude, heading angle, turn signal status, braking status, battery data and fault codes.

[0040] The roadside units and roadside sensing devices are deployed at key road nodes. The roadside sensing devices include cameras, lidar, millimeter-wave radar, and weather sensors, which are connected to the roadside computing unit (MEC) via cables or optical fibers to collect data on traffic flow, events, pedestrians, non-motorized vehicles, and weather conditions. The roadside unit (RSU) is connected to the MEC via a wired network and communicates with the on-board unit (OBU) and cloud platform via a wireless network. Other data sources include charging piles and traffic signal control systems. Charging piles upload charging data via Ethernet or wireless networks, and the traffic signal control system sends traffic light status information via a dedicated protocol.

[0041] The network transmission layer adopts a transmission method combining vehicle-to-everything (V2X) communication technology and wide-area communication technology. The V2X communication technology is C-V2X, including LTE-V2X and 5G-V2X, or the DSRC protocol, used for short-range, low-latency, and highly reliable direct communication between the vehicle-mounted OBU and the roadside RSU to transmit safety warning information. The wide-area communication technology is 5G / 4G or Ethernet, used for communication between the vehicle-mounted T-Box and the roadside MEC unit and the cloud platform to transmit monitoring and non-real-time data. The data transmission process adopts TLS / SSL encryption protocol and authentication mechanism.

[0042] The data processing and analysis layer includes a cloud-based control platform and edge computing nodes (MECs). The edge computing nodes are deployed in a data center near the roadside and connect to the RSUs and sensors via a high-speed network. They process latency-sensitive computing tasks, including obstacle recognition, traffic light status interpretation, and cooperative collision avoidance decision-making, and send the results to the vehicles via the RSUs. The cloud-based control platform adopts a distributed architecture, uses Hadoop HDFS for massive data storage, and employs Apache Kafka, Spark Streaming, or Flink stream processing frameworks for real-time data processing. It also incorporates AI algorithm models for intelligent analysis of the cleaned data.

[0043] The AI ​​algorithm model includes a driving behavior analysis model, a safety warning model, and an energy consumption assessment and carbon footprint tracking model. The driving behavior analysis model identifies undesirable driving behaviors such as rapid acceleration, rapid deceleration, and sharp turns by analyzing time-series data of vehicle acceleration, deceleration, and steering angular velocity. The safety warning model predicts accident risks based on historical data and real-time traffic conditions, or receives abnormal event information from roadside units through vehicle-road cooperation and issues warnings, with an accuracy rate of over 80% in predicting major accidents seven days in advance. The energy consumption assessment and carbon footprint tracking model dynamically analyzes the energy consumption data of new energy vehicles and calculates carbon emissions using the formula carbon emissions = ∫(real-time power × real-time carbon intensity) dt.

[0044] The application service layer includes a government regulatory platform, an enterprise operation platform, and a user service platform. The government regulatory platform provides traffic management, industry and information technology departments with web-based or large-screen visualization interfaces to achieve functions such as vehicle operation safety monitoring, traffic violation evidence collection, carbon emission supervision, and emergency plan management. The enterprise operation platform provides SaaS services to car manufacturers, logistics companies, taxi companies, etc., and outputs data analysis reports on vehicle condition, driving behavior, energy consumption, and efficiency through API interfaces or web portals for fleet management, maintenance warnings, insurance assessments, and optimized dispatching. The user service platform provides drivers with real-time traffic conditions, dangerous road section warnings, personalized driving scores, charging pile recommendations and reservations, and green travel points services through mobile apps or vehicle-mounted applications.

[0045] The Internet of Vehicles (IoV) data-driven new energy autonomous vehicle traffic behavior monitoring and service system also includes a data security and privacy protection module. It adopts encrypted data transmission and blockchain technology to ensure data authenticity and tamper-proof, constructs a three-layer information security transmission architecture of terminal-vehicle-cloud, and establishes a transparent data use and consent mechanism.

[0046] Preferably, the sensing technology route of the roadside sensing device and the vehicle terminal adopts a combination of single-vehicle intelligence and vehicle-road cooperation mode, or chooses to adopt a mode with vehicle-side intelligence as the main focus and roadside intelligence as the auxiliary focus, or a mode with roadside intelligence as the main focus and vehicle-side intelligence as the auxiliary focus.

[0047] The cloud control platform's storage architecture adopts a centralized cloud, regional private cloud, or hybrid cloud deployment mode. In the hybrid cloud mode, sensitive data is stored locally while simultaneously utilizing public cloud elastic computing resources.

[0048] The regulatory models of government-regulated platforms include government-led construction of city-level or national-level regulatory platforms for enterprises to access; or enterprises building their own platforms and accepting government regulation, with the latter providing interfaces to autonomous driving companies, integrating data, and supporting their own operations and government regulation.

[0049] The following is a complete explanation of the core algorithm formulas and implementation steps of the AI ​​algorithm model:

[0050] I. Basic Data Preprocessing

[0051] Input the raw time series data:

[0052] t = [t1, t2, ..., t n / / Timestamp sequence

[0053] v = [v1, v2, ..., v] n / / Vehicle speed sequence (km / h)

[0054] a = [a1, a2, ..., a n / / Acceleration sequence (m / s) 2 )

[0055] θ = [θ1, θ2, ..., θ n / / Steering wheel angle sequence (°)

[0056] GPS = [(lat1, lon1), ...] / / Location sequence

[0057] brake_flag = [0 / 1, ...] / / Braking state sequence

[0058] II. Core Feature Engineering and Algorithm Formulas

[0059] 1. Instantaneous dynamic feature extraction

[0060] Acceleration calculation (finite difference method):

[0061] a_t=(v_t-v_{t-1}) / (Δt)×(1000 / 3600) / / Convert to m / s 2

[0062] Where Δt = t_i - t_{i-1} (unit: seconds)

[0063] Jerk (Jerk) calculation:

[0064] j_t=(a_t-a_{t-1}) / Δt / / Unit: m / s 3

[0065] The Jerk value directly reflects the smoothness of operation and is a core indicator for rapid acceleration and deceleration.

[0066] Calculation of lateral acceleration:

[0067] a_lateral_t=v_t 2 ×sin(Δθ_t) / R_t

[0068] Where R_t is the turning radius, Δθ_t = θ_t - θ_{t-1}2. Key Behavior Recognition Algorithm

[0069] Formula for determining rapid acceleration:

[0070]

[0071] Formula for determining sudden deceleration / sudden braking:

[0072]

[0073] Frequent lane change identification (based on position and heading):

[0074] Lateral displacement fluctuation rate = θ(Δy_i) / average vehicle speed

[0075] Where Δy_i = |lon_i - fitted curve(lat_i)| × C / / C is the latitude-longitude distance conversion factor

[0076] Entropy value of heading angle change: H(θ)=-∑p(θ_i)log p(θ_i)

[0077] Where p(θ_i) is the probability of the heading angle in the i-th interval.

[0078] 3. Fatigue driving detection model

[0079] Steering wheel micro-correction frequency analysis:

[0080] Micro-corrected signal = high-pass filter (θ_t, f_cutoff = 0.3Hz)

[0081] Correction frequency = count(zero crossing) / time window

[0082] Correction magnitude entropy = -∑p(A_i)log p(A_i) / / A_i is the correction magnitude level

[0083] Lane keeping ability indicators:

[0084] Lane Departure Index = (Deviation Area / Travel Distance) × 100%

[0085] Area of ​​deviation = ∫|LateralOffset(t)|dt

[0086] Where LateralOffset(t) is the distance from the vehicle center to the lane centerline. III. Comprehensive Scoring Model

[0087] 1. Feature normalization processing

[0088] X_norm=(X_raw-μ) / σ / / Z-score standardization

[0089] or

[0090] X_norm = (X_raw - X_min) / (X_max - X_min) / / Min-Max normalization

[0091] 2. Multi-dimensional scoring function

[0092] Principal component analysis and weight determination:

[0093] 1. Construct the feature matrix F = [f1, f2, ..., f_m] / / m features, n samples

[0094] 2. Calculate the covariance matrix ∑=cov(F)

[0095] 3. Eigenvalue decomposition: ∑=QΛQ T

[0096] 4. Select the top k principal components with a cumulative contribution rate > 85%.

[0097] 5. Weight vector W = [w1, w2, ..., w_k] = proportion of eigenvalues

[0098] Overall score calculation:

[0099] Driving safety score = 100 - ∑(w_i×S_i×α_i) / / Total score out of 100

[0100] in:

[0101] S_i is the weighted number of the i-th type of violation.

[0102] α_i is the severity coefficient (example value):

[0103] - Rapid acceleration: α = 1.2

[0104] -Emergency braking: α = 1.5

[0105] -Speeding: α = 1.8

[0106] - Frequent lane changes: α = 1.0

[0107] - Drowsy driving: α = 2.0

[0108] 3. Improved machine learning-based model, Gradient Boosting Tree (GBDT) risk prediction:

[0109]

[0110] Long Short-Term Memory (LSTM) Sequence Pattern Recognition:

[0111] IV. Implementation Steps and Processing Flow

[0112] ---------------------------------------------------------------

[0113] Step 1: Data Acquisition and Cleaning

[0114] - Obtain the raw signal from the CAN bus, with a sampling frequency ≥10Hz.

[0115] -Outlier detection and handling: using the 3σ principle or IQR method

[0116] Missing value imputation: linear interpolation or forward filling

[0117] Step 2: Time Window Segmentation

[0118] - Fixed window: Each analysis window lasts 60 seconds, with a 50% overlap.

[0119] - Event-triggered window: Automatically creates a variable-length window when an anomaly is detected.

[0120] Step 3: Feature Extraction

[0121] For each time window:

[0122] Calculate basic statistics: mean, variance, kurtosis, skewness

[0123] Calculate dynamic characteristics: Jerk integral, energy spectral density

[0124] Calculate the frequency domain characteristics: the main frequency components after FFT transform

[0125] Step 4: Behavior Recognition and Classification

[0126] - Rule Engine: Apply the above threshold judgments to generate preliminary labels.

[0127] - Machine learning classifiers: Fine-grained classification using pre-trained SVMs / random forests - Sequence models: LSTM for recognizing complex temporal dependency patterns

[0128] Step 5: Rating and Feedback Generation

[0129] - Calculate sub-scores for each dimension

[0130] - Weighted fusion generates a comprehensive score

[0131] - Generate personalized improvement suggestion report

[0132] V. Key Threshold Reference Table

[0133] Behavioral types Main detection indicators Threshold range Severity level rapid acceleration Acceleration a > <![CDATA[2.5-3.5m / s 2 ]]> high Rapid deceleration deceleration a < <![CDATA[.3.0--4.0m / s 2 ]]> high sharp turn Lateral acceleration > <![CDATA[0.4g(3.92m / s 2 )]]> middle speeding Speed ​​> Speed ​​Limit 10-20% speed limit Medium-high Frequent lane changes Number of lane changes per unit distance >0.1 times / km Low-medium Fatigue driving Micro-correction frequency change <30% of baseline value high

[0134] VI. Model Validation and Optimization

[0135] Cross-validation strategy:

[0136] Time-series cross-validation is used to prevent data leakage.

[0137] Training set:test set:validation set = 70:15:15

[0138] Final output: y_t = softmax(W_y·h_t + b_y) / / behavior classification

[0139] Performance metrics:

[0140] Precision = TP / (TP + FP) / / For high-risk behaviors

[0141] Recall rate = TP / (TP + FN) / / Ensure no serious behaviors are missed

[0142] F1 score = 2 × Precision × Recall / (Precision + Recall)

[0143] AUC-ROC area under the curve / / ​​Comprehensive evaluation of the model's discriminative ability

[0144] Online learning and adaptive learning:

[0145] Collect driver feedback (confirming false alarms)

[0146] Periodically update the threshold: threshold_new = α × threshold_old + (1-α) × threshold_actual

[0147] Where α is the smoothing factor (usually 0.8-0.9).

[0148] The energy consumption assessment and carbon footprint tracking model is based on the real-time power integral method to build a calculation system for the energy consumption and carbon footprint of new energy vehicles, abandoning the traditional average energy consumption coefficient method and achieving accurate calculation at the second level.

[0149] I. Core Algorithm Formula

[0150] 1. Instantaneous energy consumption calculation model

[0151] Vehicle total power demand function:

[0152] The formulas for calculating the power of each component are: P_total(t) = P_kinetic(t) + P_air(t) + P_roll(t) + P_grade(t) + P_accessory(t) - P_regen(t).

[0153] Power of kinetic energy change:

[0154] P_kinetic(t) = 1 / 2 × m × (v_t) 2 -v_{t-1} 2 ) / Δt

[0155] m: Gross vehicle mass (kg), including battery and load

[0156] v_t: Vehicle speed at time t (m / s)

[0157] Δt: Sampling interval (s), typically 0.1-1 seconds.

[0158] Air resistance power:

[0159] P_air(t)=1 / 2×ρ×C_d×A_f×v_t 3

[0160] ρ: Air density (kg / m³) 3 The default value is 1.225, which can be adjusted according to temperature and humidity.

[0161] C_d: Drag coefficient, an inherent parameter of the vehicle.

[0162] A_f: Vehicle frontal area (m²) 2 )

[0163] Rolling resistance power:

[0164] P_roll(t)=μ_r×m×g×cos(θ)×v_t

[0165] μ_r: Rolling resistance coefficient (0.01-0.015 for asphalt roads, 0.02 for wet and slippery roads)

[0166] g: acceleration due to gravity (9.81 m / s²) 2 )

[0167] θ: Road slope angle (°), from high-precision map or IMU

[0168] Slope resistance power:

[0169] P_grade(t)=m×g×sin(θ)×v_t

[0170] Attachment power:

[0171] P_accessory(t)=P_ac+P_light+P_infotainment+...

[0172] Air conditioner (P_ac): Related to set temperature difference and compressor power.

[0173] Other accessories are read via the CAN bus or estimated based on rated power.

[0174] Regenerative braking power recovery:

[0175] P_regen(t)=η_regen×|P_brake(t)|×I_regen_flag(t)

[0176] 2. Battery-to-wheel efficiency model

[0177] Total battery output power:

[0178] E_battery=∫P_battery(t)dt=∫[P_total(t) / η_total(t)]dt The system's total efficiency function is:

[0179] η_total(t)=η_battery(t)×η_inverter(t)×η_motor(t)×η_transmission(t)

[0180] Battery efficiency: η_battery(t) = f(SOC, T_batt, I_batt), look up a table or fit a curve;

[0181] Motor efficiency: η_motor(t) = f(Torque, RPM), refer to the efficiency map;

[0182] 3. Equivalent power consumption and carbon emission conversion

[0183] Total power consumption for the trip:

[0184] E_trip=∑[P_battery(t)×Δt] / 3600 / / Unit: kWh

[0185] Equivalent energy consumption per 100 kilometers:

[0186] e_100km=(E_trip / D_trip)×100 / / Unit: kWh / 100km

[0187] Carbon emissions based on grid carbon intensity:

[0188] CO2_trip=E_trip×λ_grid(t,loc)

[0189] λ_grid(t, loc): Real-time marginal carbon emission factor of the power grid (gCO2 / kWh) in region loc at time t, which needs to be obtained dynamically;

[0190] Carbon emissions based on oil-electricity conversion (alternative options):

[0191] CO2_trip=E_trip×(ρ_fuel×CV_fuel×EF_fuel) / (η_powerplant×3600)

[0192] ρ_fuel: Gasoline density (~0.732 kg / L)

[0193] CV_fuel: Calorific value of gasoline (~44.4 MJ / kg)

[0194] EF_fuel: Gasoline emission factor (~69.3gCO2 / MJ)

[0195] η_powerplant: Power plant generation efficiency (~40%)

[0196] III. Model Implementation Steps

[0197] Step 1: Data Acquisition and Preprocessing

[0198] Input: Time series data stream {t, v, a, SOC, T_batt, I_batt, Torque, RPM, θ, ...}

[0199] Sampling rate: ≥10Hz

[0200] Anomaly handling: 5-point moving average filtering, 3σ outlier removal.

[0201] Step 2: Calculate power demand per second (edge ​​computing)

[0202] For each sampling point i: |

[0203] P_air_i=0.5*ρ*C_d*A_f*v_i 3

[0204] P_roll_i=μ_r*m*g*cos(θ_i)*v_i

[0205] P_grade_i=m*g*sin(θ_i)*v_i

[0206] P_kinetic_i=0.5*m*(v_i 2 -v_{i-1} 2 ) / Δt

[0207] P_accessory_i = lookup table (air conditioner setting, outside temperature) + constant value

[0208] P_regen_i=η_regen*|min(0, P_brake_i)|

[0209] IV. Key Parameter Table (Example Values)

[0210]

[0211] V. Model Validation and Calibration

[0212] Chassis dynamometer benchmark test:

[0213] Under the NEDC / WLTC cycle:

[0214] Actual energy consumption: E_dyno = 15.2 kWh / 100km

[0215] Model calculation: E_model = 15.8 kWh / 100km

[0216] Error: Δ = (15.8 - 15.2) / 15.2 = 3.9% < 5% (acceptable) Real-vehicle road verification:

[0217] Select 50 vehicles of the same model and record 1000 trips:

[0218] Model mean error: 4.2%

[0219] Standard deviation of error: 2.1%

[0220] Error < 8% within 95% confidence interval

[0221] VI. Special Treatment for Carbon Footprint Tracking

[0222] Charging source traceability:

[0223]

[0224] Carbon offset accounting:

[0225] Net carbon emissions = vehicle emissions - carbon credit offset

[0226] Carbon credits come from:

[0227] - Green electricity certificates (1 REC generated per MWh)

[0228] -Afforestation projects (calculated based on absorption)

[0229] - Carbon allowances purchased by enterprises

[0230] Full lifecycle extension (optional):

[0231] LCA carbon emissions = vehicle emissions + manufacturing emissions allocation + battery production emissions allocation

[0232] Manufacturing emissions allocation = Total emissions from vehicle manufacturing / Expected total mileage × Current mileage

[0233] Battery production emissions = Battery pack carbon footprint / Cycle life × Number of cycles

[0234] The present invention has the following beneficial effects:

[0235] 1. Regulatory efficiency has undergone a fundamental transformation, shifting from traditional post-event tracing to pre-event warning and in-event intervention. The accuracy rate of early warning for major accidents, which is 7 days in advance, exceeds 80%. After adopting this system, the traffic management department of a certain city has improved the efficiency of inspecting violations by ride-hailing vehicles and freight vehicles by 50% and reduced labor costs by 40%.

[0236] 2. The technology has achieved a qualitative leap in effectiveness. A driving behavior scoring model based on hundreds of feature parameters has been established to quantitatively assess drivers' safety awareness and operating habits. After a commercial vehicle management system was connected to the model, the overall accident rate of the fleet decreased by 25%, and vehicle energy consumption decreased by 8%. By dynamically collecting real-time energy consumption data, accurate carbon emission tracking has been achieved, providing a reliable data foundation for carbon trading. The cloud-edge-device collaborative architecture reduces response latency and improves system processing efficiency.

[0237] 3. Significant social and economic value: For government regulators, it innovates the online + offline supervision model, improving governance efficiency and scientific decision-making capabilities; for operating companies, it reduces maintenance costs through predictive maintenance, optimizes insurance costs based on driving behavior scores (safe drivers can receive up to 30% premium discounts), and improves product design with the help of real operational data; for end users, it provides personalized safety services, enhancing travel safety and experience.

[0238] 4. Break down information silos and achieve cross-departmental, cross-level, and cross-enterprise data sharing and business collaboration through unified data standards and a three-tiered platform system at the national, local, and enterprise levels. This enables unified network dispatch and command during major event security operations.

[0239] This system adopts a collaborative technical architecture of end-edge-cloud, and the implementation process is as follows:

[0240] Data collection: Onboard terminals and roadside equipment continuously collect data such as vehicle status and environmental perception;

[0241] Data transmission: Transmits data to edge nodes and the cloud with low latency and high reliability via C-V2X and 5G networks;

[0242] Data processing and analysis: Edge nodes handle security services with extremely high real-time requirements, while the cloud platform performs massive data storage, aggregation, and in-depth analysis;

[0243] Services and Applications: Providing analysis results to regulators, businesses, and users to achieve regulatory, operational, and service functions.

[0244] (II) Specific application scenarios: Early warning of road construction ahead and coordinated detour

[0245] Event perception and uploading (data acquisition and transmission layer): Roadside sensing devices (LiDAR and cameras located 300 meters ahead of the construction section) capture obstacles such as cones and construction vehicles. The YOLOv7 target detection algorithm built into the MEC identifies road construction events in real time. The MEC application generates standardized MAPEM and SPaT messages containing event type, latitude and longitude coordinates, affected lanes, suggested speed, and detour trajectory. The RSU periodically broadcasts this warning message through the PC5 interface.

[0246] The vehicle-mounted terminal receives and makes decisions. The vehicle's OBU communication module receives the warning message and transmits it to the decision-making and planning module for parsing. The decision-making and planning module calls the data from the perception fusion module and compares and verifies the forward environment information perceived by the vehicle's sensors with the MAPEM message. If the vehicle is located in the lane affected by construction, it calculates the safe lane-changing conditions. If the conditions are met, it generates a left lane-changing trajectory planning instruction. If the conditions are not met, it generates a deceleration instruction. If the event does not affect the vehicle, it saves the data and continues to monitor.

[0247] The vehicle executes and sends decision-making instructions to the vehicle control actuator, and the vehicle automatically performs actions such as deceleration and lane changing; after the vehicle HMI is triggered, the construction area is highlighted on the instrument panel, and the driver is reminded through voice and icons; the T-Box encrypts and uploads core information such as vehicle ID, reception time, event type, and response measures to the cloud control platform through the Uu interface.

[0248] The cloud platform receives data for cloud analytics and archiving (data processing and application service layer) and writes it to the HDFS distributed file system for permanent storage via Kafka message queues. The Spark Streaming stream processing program aggregates and analyzes data on similar events involving multiple vehicles, calculates the average speed change and congestion index at construction sites, and assesses the reach rate of warning messages and vehicle response rate. The analysis results are presented on the GIS map of the government supervision platform, and a "Temporary Traffic Incident Impact Report" is generated and sent to the road maintenance department.

[0249] (III) Key Computer Program Technology Implementation

[0250] Roadside MEC target detection module: The YOLOv7 algorithm is used to realize real-time identification of traffic events, ensuring the accuracy and real-time performance of event detection.

[0251] Cloud-based driving behavior analysis model training (SparkMLlib example): Historical vehicle CAN signal data is read from HDFS, and features such as acceleration and jerk are calculated. The time series data is divided into 5-second fixed-duration windows, and features such as max_jerk and avg_acceleration are extracted. Supervised learning is adopted, using expert-annotated rapid acceleration labels as the training basis. Gradient boosting tree (GBTClassifier) ​​or random forest (RandomForestClassifier) ​​is selected as the classifier. Hyperparameters are optimized through cross-validation and grid search. The trained model is saved as a model.pkl file and loaded by a stream processing program for real-time inference.

[0252] Vehicle-Road-Cloud Communication Message Encoding: SAE J2735 (DSRC) or CSAE53-2020 (China C-V2X Application Layer Standard) is adopted, and data interaction is achieved through standardized message sets such as MAPEM, BSM, and RSI.

[0253] This invention deeply integrates multiple technical fields such as vehicles, roads, cloud, networks, and maps to construct a comprehensive, safe, reliable, and flexible monitoring and service system. It effectively solves many shortcomings of existing technologies, promotes the transformation of traffic monitoring models towards modernization and refinement, and has broad application prospects and significant socio-economic value.

[0254] Table 1 below compares the regulatory effectiveness of traditional regulatory methods with that of the present invention:

[0255] Table 1 Regulatory Efficiency:

[0256]

[0257] Table 2 shows the comparison of the effects of innovation point enhancement.

[0258]

[0259] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A vehicle-to-everything (V2X) data-based traffic behavior monitoring and service system for new energy autonomous vehicles, characterized by: It includes a data acquisition layer, a network transmission layer, a data processing and analysis layer, and an application service layer that are connected in sequence. The data acquisition layer, network transmission layer, and data processing and analysis layer adopt an end-edge-cloud collaborative architecture to realize real-time monitoring, safety warning, behavior analysis, and efficient services for new energy and autonomous vehicles. The data acquisition layer is used to collect raw data on vehicle status, road environment, traffic events, and supplementary data sources. The network transport layer is used to achieve low-latency, highly reliable transmission and secure encryption of data between layers; The data processing and analysis layer is used to store, clean, process in real time, and perform intelligent analysis on the collected data to generate the analysis results required for supervision and services. The application service layer is used to provide targeted supervision, operation and personalized services to different users.

2. The vehicle-to-everything (V2X) data-based traffic behavior monitoring and service system for new energy autonomous vehicles as described in claim 1, characterized in that: The data acquisition layer includes an on-board terminal, a roadside unit and roadside sensing devices, and other data sources. The on-board terminal is a T-Box / OBU integrated inside the vehicle, which is connected to the vehicle's power system, battery management system, autonomous driving domain controller, GPS / BeiDou positioning module, and inertial measurement unit via a CAN bus, dedicated wiring harness, or wireless interface to collect vehicle speed, acceleration, latitude and longitude, heading angle, turn signal status, braking status, battery data, and fault codes. The roadside units and roadside sensing devices are deployed at key road nodes. The roadside sensing devices include cameras, lidar, millimeter-wave radar, and weather sensors, which are connected to the roadside computing unit (MEC) via cables or optical fibers to collect data on traffic flow, events, pedestrians, non-motorized vehicles, and weather conditions. The roadside unit (RSU) is connected to the MEC via a wired network and communicates with the on-board unit (OBU) and cloud platform via a wireless network. Other data sources include charging piles and traffic signal control systems. Charging piles upload charging data via Ethernet or wireless networks, and the traffic signal control system sends traffic light status information via a dedicated protocol.

3. The vehicle-to-everything (V2X) data-based traffic behavior monitoring and service system for new energy autonomous vehicles as described in claim 1, characterized in that: The network transmission layer adopts a transmission method combining vehicle-to-everything (V2X) communication technology and wide-area communication technology. The V2X communication technology is C-V2X, including LTE-V2X and 5G-V2X, or the DSRC protocol, used for short-range, low-latency, and highly reliable direct communication between the vehicle-mounted OBU and the roadside RSU to transmit safety warning information. The wide-area communication technology is 5G / 4G or Ethernet, used for communication between the vehicle-mounted T-Box and the roadside MEC unit and the cloud platform to transmit monitoring and non-real-time data. The data transmission process adopts TLS / SSL encryption protocol and identity authentication mechanism.

4. The vehicle-to-everything (V2X) data-based traffic behavior monitoring and service system for new energy autonomous vehicles as described in claim 1, characterized in that: The data processing and analysis layer includes a cloud control infrastructure platform and edge computing nodes (MECs). The edge computing nodes are deployed in a data center near the roadside and connect to the RSUs and sensors via a high-speed network. They process latency-sensitive computing tasks, including obstacle recognition, traffic light status interpretation, and cooperative collision avoidance decision-making, and send the results to the vehicles via the RSUs. The cloud control infrastructure platform adopts a distributed architecture, uses Hadoop HDFS for massive data storage, and uses Apache Kafka, Spark Streaming, or Flink stream processing frameworks for real-time data processing. It also has built-in AI algorithm models for intelligent analysis of the cleaned data.

5. The vehicle-to-everything (V2X) data-based traffic behavior monitoring and service system for new energy autonomous vehicles as described in claim 4, characterized in that: The AI ​​algorithm model includes a driving behavior analysis model, a safety warning model, and an energy consumption assessment and carbon footprint tracking model. The driving behavior analysis model identifies poor driving behaviors such as rapid acceleration, rapid deceleration, and sharp turns by analyzing time-series data of vehicle acceleration, deceleration, and steering angular velocity. The safety warning model predicts accident risks based on historical data and real-time traffic conditions, or receives abnormal event information from roadside units through vehicle-road cooperation and issues warnings, with a prediction accuracy of over 80% for major accidents seven days in advance. The energy consumption assessment and carbon footprint tracking model dynamically analyzes the energy consumption data of new energy vehicles and calculates carbon emissions using the formula carbon emissions = ∫(real-time power × real-time carbon intensity) dt. The driving behavior analysis model achieves a systematic quantitative assessment of driver operating characteristics through multi-dimensional data fusion and pattern recognition.

6. The vehicle-to-everything (V2X) data-based traffic behavior monitoring and service system for new energy autonomous vehicles as described in claim 1, characterized in that: The application service layer includes a government regulatory platform, an enterprise operation platform, and a user service platform. The government regulatory platform provides traffic management and industry and information technology departments with a web-based or large-screen visualization interface to realize functions such as vehicle operation safety monitoring, traffic violation evidence collection, carbon emission supervision, and emergency plan management. The enterprise operation platform provides SaaS services to car manufacturers, logistics companies, and taxi companies, outputting data analysis reports on vehicle status, driving behavior, energy consumption, and efficiency through API interfaces or web portals for fleet management, maintenance warnings, insurance assessments, and optimized dispatching. The user service platform provides drivers with real-time traffic conditions, dangerous road section warnings, personalized driving scores, charging pile recommendations and reservations, and green travel points services through mobile apps or vehicle-mounted applications.

7. The vehicle-to-everything (V2X) data-based traffic behavior monitoring and service system for new energy autonomous vehicles as described in claim 1, characterized in that: It also includes a data security and privacy protection module, which uses encrypted data transmission and blockchain technology to ensure data authenticity and tamper-proof, builds a three-layer information security transmission architecture of terminal-vehicle-cloud, and establishes a transparent data use and consent mechanism.

8. The vehicle-to-everything (V2X) data-based traffic behavior monitoring and service system for new energy autonomous vehicles as described in claim 2, characterized in that: The perception technology approach of the roadside sensing equipment and vehicle-mounted terminal adopts a combination of single-vehicle intelligence and vehicle-road cooperation, or chooses to adopt a mode with vehicle-end intelligence as the main focus and roadside intelligence as the auxiliary focus, or a mode with roadside intelligence as the main focus and vehicle-end intelligence as the auxiliary focus.

9. The vehicle-to-everything (V2X) data-based traffic behavior monitoring and service system for new energy autonomous vehicles as described in claim 4, characterized in that: The storage architecture of the cloud control platform adopts a central cloud, regional private cloud, or hybrid cloud deployment mode. In the hybrid cloud mode, sensitive data is stored locally while calling on public cloud elastic computing resources.

10. The vehicle-to-everything (V2X) data-based traffic behavior monitoring and service system for new energy autonomous vehicles as described in claim 6, characterized in that: The regulatory models of the government regulatory platform are as follows: the government takes the lead in building a city-level or national-level regulatory platform, which enterprises connect to; or enterprises build their own platforms and accept government supervision, with the enterprise-built platforms opening interfaces to autonomous driving companies, integrating data, and providing support for their own operations and government supervision.

Images