An internet of vehicles signal data slicing method based on a travel strategy

Abstract

This invention discloses a method for slicing vehicle network signal data based on a travel strategy. This method collects spatiotemporal feature data during vehicle travel, uses machine learning algorithms to identify and segment vehicle travel, and selects representative samples for data slicing based on a travel selection strategy. The system of this invention includes functional units such as a data acquisition module, a travel identification module, a slicing processing module, an evaluation and feedback module, and a storage management module. The method and system provided by this invention can effectively solve the data processing efficiency problem in the context of big data in the Internet of Vehicles (IoV), significantly reducing storage and computational overhead while ensuring data representativeness, and has good practical value.

Description

Technical Field

[0001] This invention belongs to the field of vehicle-to-everything (V2X) big data processing technology, specifically relating to a V2X signal data slicing method based on travel strategies. This invention aims to solve the efficiency problem of storing and processing massive amounts of signal data in a V2X environment. Through intelligent signal data slicing technology, it achieves efficient management and utilization of signal data while ensuring representativeness. Background Technology

[0002] With the rapid development of vehicle-to-everything (V2X) technology, the signal data generated by vehicles during operation is experiencing explosive growth. Traditional data storage and processing methods face several major problems when dealing with such massive amounts of data, including: 1. High data storage costs: Due to the enormous volume of V2X signal data, a large amount of storage resources are required, and traditional storage solutions are insufficient to meet the demands for efficient read / write operations; 2. High data redundancy: The presence of a large amount of duplicate information leads to low data processing efficiency, increasing the time required for full data processing and consuming significant computing resources, making it difficult to guarantee the performance of services with high real-time requirements; 3. Difficulty in extracting data value: Fixed-period data collection lacks flexibility, and simple signal data sampling methods are insufficiently representative, failing to serve as backups or even replacements for the entire signal data set.

[0003] Therefore, there is an urgent need for an intelligent method for slicing vehicle network signal data, which can effectively reduce storage overhead while improving data computing efficiency. This invention is an innovative solution to the above problems. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a method for slicing vehicle network signal data based on travel strategy. This method can achieve an effect similar to using the full dataset by using only sliced ​​data without relying on full dataset calculations. The method includes the following steps: S1. Calculate and assign a travel ID to each signal based on the travel definition; S2. Construct a clustering model and select representative trips; S3. Develop an optimal itinerary segmentation strategy based on the selected representative itineraries; S4. Implement vehicle-to-everything (V2X) signal data slicing based on trip segmentation strategy.

[0005] Furthermore, S1, based on the trip definition, calculates and assigns a trip ID to each signal, innovatively constructing a dynamic trip identification mechanism. Through multi-source signal fusion and adaptive threshold determination, it achieves accurate driving trip segmentation. Specific steps include: 1) Multimodal signal acquisition: Through the collaborative work of the vehicle CAN bus and GPS module, real-time acquisition of multi-dimensional time-series data including ignition signal, engine speed, vehicle speed, latitude and longitude, etc., is carried out. The sampling frequency of key signals is not less than 10Hz to ensure high-precision capture of scenarios such as short-term parking. 2) Effective stroke dynamic determination: continuous ignition time > 2 minutes and displacement distance ≥ max (usually taken as 1km); 3) Spatiotemporal uniqueness encoding: A globally unique ID is generated for each valid journey. The encoding rules are as follows: ID = Date Hash (yyyyMMdd) ⊕ Geographic Area Code (GEO) ⊕ Serial Number (SEQ) Where: ⊕ represents the XOR operation, and the GEO encoding accuracy reaches the 100-meter grid level.

[0006] Furthermore, S2 constructs a clustering model and selects representative trips. It uses a combination of feature engineering and machine learning to select statistically representative typical trips from massive trip data. Specific steps include: 1) Multidimensional Feature Space Construction: Based on vehicle operating state parameters (including but not limited to vehicle speed, steering wheel angle, accelerator pedal opening / closing degree, battery temperature, maximum battery voltage, frequency of abnormal events, etc., exceeding 100 fields), an n-dimensional feature vector is constructed for each trip ID, forming a feature space Ω∈ℝⁿ. The selection of the feature dimension n satisfies:

[0007] Where N_{signal} is the total number of sampling points for a single-pass signal, and adaptive compression of the feature dimension is achieved through logarithmic transformation; 2) Improved clustering modeling: The K-means++ algorithm is used for clustering analysis of the feature space. Its core improvement lies in: • Initial cluster centers are selected using a probability density weighting method to avoid getting trapped in local optima; • A dynamic iteration termination mechanism is introduced, which terminates the calculation early when the change rate of the profile coefficient is ≤1% for three consecutive iterations; 3) Dual threshold screening mechanism: Intra-class compactness is ensured by using a silhouette coefficient s(i) ≥ 0.6, while an inter-class distance threshold is introduced:

[0008] Where c_p and c_q are different cluster centers, and σ_i is the standard deviation of the i-th cluster. High-quality running samples that can characterize both intra-cluster features and inter-cluster differences are selected through dual constraints.

[0009] Furthermore, S3 formulates an optimal trip segmentation strategy based on the selected representative trips, including: 1) Trip Quantity Statistics: Count the number N of valid trips selected through step S2 within the current processing cycle; 2) Slicing strategy formulation: Based on the number of runs N, the following slicing strategy is adopted: • When the number of trips N=1: Use a 30-minute threshold for data truncation. If the actual trip duration is less than 30 minutes, truncate according to the actual duration. • When the number of trips N=2: Data is truncated for each trip based on a 15-minute threshold. If the duration of a single trip is less than 15 minutes, the actual duration is used for truncating. • When the number of trips N≥3: 10 minutes of data are extracted from the first and last trips, and 10 minutes of data are also extracted from each trip in between. If the trip duration is less than 10 minutes, the actual duration is used for extraction.

[0010] Compared with existing technologies, this invention has the following advantages: it achieves a score result close to that of the full signal data using sliced ​​vehicle network signal data. Enterprises can reduce the consumption of computing and storage resources while ensuring the quality of the output indicators, which has significant practical value and promotional significance in the future big data environment. Attached Figure Description

[0011] Figure 1 This is a flowchart illustrating the steps of a vehicle-to-everything (V2X) signal data slicing method based on a travel strategy, as provided in an embodiment of this application.

[0012] Figure 2 This is a comprehensive driving score result based on full data and slice data, provided in one embodiment of this application.

[0013] Figure 3 This is a comparison chart of the relationship between comprehensive driving score results based on full data and slice data, provided in one embodiment of this application. Detailed Implementation

[0014] To make the technical solutions and effects of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0015] The invention aims to provide a method for slicing vehicle network signal data based on travel strategy (see details). Figure 1 This method consists of 6 steps, which will be implemented and introduced one by one.

[0016] Step 1: Calculate and assign a travel ID to each signal based on the travel definition. The specific process includes: 1) In the signal data acquisition stage, the vehicle's ignition status signal, engine shutdown status signal, and GPS positioning information are collected in real time via the vehicle's CAN bus to form a multi-dimensional time-series signal dataset. The signal sampling frequency is set to 10Hz to ensure high-precision capture of short-term start-stop operations of the vehicle; 2) Furthermore, the determination of valid trips employs a dynamic threshold determination mechanism to identify valid driving trips. The specific process includes: • Ignition event detection: Record the start time ts when the ignition signal changes from 0 to 1; • Flameout event detection: Record the end time te when the ignition signal changes from 1 to 0; • Distance calculation: The difference in vehicle odometer mileage between ts and te; 3) Furthermore, in the dynamic threshold determination process, a stroke is considered valid when the following conditions are met: • Continuous ignition time te - ts > 2 minutes; • Driving cluster S ≥ max(1km, 0.1 × historical average daily driving distance); • Engine shutdown lasting less than 3 minutes is not counted as trip termination; 4) Furthermore, the uniqueness coding generation process generates a globally unique ID for each valid process, with the coding rule as follows: ID = Date code (yyMMdd) + Geographic grid code (100m precision) + Serial number (0001−9999).

[0017] Step 2: Construct a clustering model and select representative travel rules. The specific process includes: 1) Construct a feature space and use multi-dimensional feature vectors to represent driving behavior features, specifically including the following parameter categories: • Driving behavior characteristics: o Frequency of rapid acceleration (count of events with acceleration ≥ 2.5 m / s²); o Frequency of rapid deceleration (count of events with deceleration ≥ -2.5m / s²); o Average rate of change of steering wheel angle (degrees / second); o Other; • Spatiotemporal distribution characteristics: o Travel time distribution (morning peak / daytime / evening peak / nighttime period marking); o Road type distribution (proportion of expressways / urban roads / rural roads); o Standard deviation of mean speed (reflects the degree of speed fluctuation); o Other; • Vehicle status characteristics: o Average engine load rate; o Variance of brake pedal opening; o Battery voltage fluctuation range; o Other; 2) Further, cluster analysis will be implemented. Before the analysis, the data will be standardized, and the specific formula is as follows:

[0018] Where μ is the characteristic mean and σ is the standard deviation; 3) Furthermore, dimensionality was reduced to 5 dimensions using principal component analysis (PCA), retaining more than 90% of the feature information; 4) Further, begin representative trip screening by implementing a dual filtering standard: • Intra-class compactness standard: Profile coefficient ≥ 0.6; • Inter-class discrimination criterion: minimum inter-class distance ≥ 1.5 times the average intra-class distance; The selected trips are sorted by timestamp, and the three samples closest to the cluster center from each cluster are selected as representatives.

[0019] Step 3: Develop an optimal itinerary segmentation strategy based on the selected representative itineraries. Specific segmentation rules are as follows: • Single-pass processing, when N=1: o The baseline slice duration is set to 30 minutes; o If the total trip duration is ≤30 minutes, it will be sliced ​​into segments based on the actual duration; o If the total trip duration is >30 minutes, divide it into segments at equal intervals (e.g., a 35-minute trip → 1 30-minute segment + 1 5-minute segment); • Two-pass processing, when N=2: Each trip is allocated a base duration of 15 minutes; o Calculate slices independently for two runs, prioritizing ensuring that each run contains at least one complete slice; Example: Trip A lasts 20 minutes → 15 minutes + 5 minutes slice; Trip B lasts 10 minutes → whole segment slice; • Multi-run processing, when N≥3: Each trip is allocated a base duration of 10 minutes; Example: 3 trips (12 / 18 / 15 minutes) → 17-minute slice for the first trip, 10-minute slice for the middle trip, and 15-minute slice for the last trip; Furthermore, optimization processing is performed on detected short trips. When the trip duration T is less than the allocated duration, the following steps are taken: • If T ≥ 5 minutes: Generate a single slice based on the actual duration; • If T < 5 minutes: merge into the adjacent trip slice (time continuity must be satisfied).

[0020] Step 4: Initialize the structure and parameters of the linear model constructed using linear regression to generate a comprehensive scoring index y used to evaluate the final effect of the slices. Specifically, the linear model constructed using linear regression satisfies the formula: y = ω 1 x 1 + ω 2 x 2 + ⋯ + ω n x n + b in: y To provide a comprehensive driving score, x 1 … x n for Driving behavior characteristics ω 1… ω n For the corresponding weighting coefficients, b Offset To evaluate whether the trip-strategy-based vehicle-to-everything (V2X) data slicing method can effectively replace full-data sets, we use the comprehensive driving score index y as a benchmark to compare the differences between the two datasets. We collected full signal data from 3542 vehicles of a certain OEM within a single day, sliced ​​the full signal data based on the trip-strategy method, and calculated the comprehensive driving score results based on the full signal data and the comprehensive driving score results based on the sliced ​​signal data for each vehicle using the driving score model in step 4 (see [reference]). Figure 2 ).

[0021] Step 5: Compare the differences in the evaluation results. The specific indicator measurement standards and thresholds are as follows: • Pearson correlation coefficient: ρ≥0.8 • Difference control indicators: MAE≤8 ∧ RMSE≤12 When the data indicators reach the threshold requirements, it is considered that the driving behavior scores based on sliced ​​data and the driving behavior scores based on full data have a strong correlation, and the sliced ​​data meets the business requirements. In this verification of signal data from 3542 vehicles, we evaluated the correlation and difference of the data results. The specific evaluation results are as follows (refer to...). Figure 3 ): • Pearson correlation coefficient: ρ = 0.81 • Difference control indicators: MAE=7.62 ∧ RMSE=11.70 To ensure the model's predictive accuracy, if the model's prediction result is less than or equal to a preset threshold, it indicates that the model's training error is too large, requiring adjustment of the cross-training parameters and retraining. If the model's prediction result is greater than the preset threshold, it indicates that the model has achieved the expected accuracy. Based on the results of this sample data execution, both the comprehensive driving behavior score based on sliced ​​data and the comprehensive driving behavior score based on the full dataset met the threshold requirements in both correlation and difference dimensions, indicating that the two are interchangeable.

[0022] Step 6: Implement vehicle-to-everything (V2X) signal data slicing based on the trip segmentation strategy. Since the execution of data slicing is not the main focus of this invention, it will not be described in detail here.

[0023] It should be noted that the information interaction and execution process between the various units in the above-mentioned device and system are based on the same concept as the method embodiment of the present invention, and the specific details can be found in the description in the method embodiment of the present invention, and will not be repeated here.

[0024] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, which may include: read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0025] The foregoing has provided a detailed description of a vehicle network signal data slicing method based on travel strategy provided by the embodiments of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for slicing vehicle network signal data based on travel strategy, characterized in that, The method includes the following steps: S1. Calculate and assign a travel ID to each signal based on the travel definition; S2. Construct a clustering model and select representative travel rules; S3. Develop an optimal itinerary segmentation strategy based on the selected representative itineraries; S4. Implement vehicle-to-everything (V2X) signal data slicing based on trip segmentation strategy.

2. The method for calculating and assigning a travel ID to each signal based on travel definition as described in claim 1, characterized in that, Step S1, which calculates and assigns a travel ID to each signal based on the travel definition, includes: 1) Signal acquisition: Ignition status signal V_ign(t) and instantaneous speed v(t) are acquired in real time via the vehicle CAN bus; 2) Start and end detection: When V_ign(t) is detected to change from 0 to 1, record the start time t_s; when it changes from 1 to 0, record the end time t_e. 3) Validity determination: A valid trip ID is generated if and only if the following conditions are met: • Duration Δt = t_e - t_s ≥ 2 minutes; • The distance traveled is S = ∫_{t_s}^{t_e} v(t)dt ≥ S_min; • Maintain ignition state throughout: ∀t ∈ [t_s, t_e], V_ign(t)=1; in: • S_min = max(1km, 0.1×historical average travel distance); • Historical average trip distance is the arithmetic mean of the effective trip distances within the most recent 7 days.

3. The vehicle-to-everything (V2X) data slicing method based on trip strategy according to claim 1, characterized in that, Step S2, which involves constructing a clustering model and selecting representative travel rules, includes: 1) Feature Space Construction: Extract multi-dimensional feature vectors corresponding to each trip ID from the preprocessed vehicle network signal data, and construct the feature space Ω⊆ℝⁿ, where n is the number of feature dimensions; 2) Clustering Modeling: The K-means++ algorithm is used to cluster Ω, and the initial cluster centers are selected to satisfy: in: c_i represents the i-th cluster center, and the upper limit of the number of iterations is set to T_max; 3) Representative journey screening: Calculate the silhouette coefficient s(i) for each sample: in: a(i) is the average distance from sample i to other samples in the same cluster, and b(i) is the minimum average distance from sample i to the nearest neighbor sample. The travel data that satisfy s(i)≥0.6 are retained to form a representative travel set Ψ.

4. The method according to claim 1, characterized in that, Step S3 includes: 1) Trip Quantity Determination: Count the number N of valid trips filtered by S2 within the current processing cycle; 2) Dynamic slicing strategy generation: The slicing strategy is determined based on the mapping relationship between the number of trips N and the preset time threshold T. in: • Tmax = 30 minutes is the maximum slice duration for a single run; • Tmin = 10 minutes is the minimum effective slice duration; • T1 represents the actual duration of the current trip; 3) Based on the above trip segmentation rules, determine the number of trips the vehicle will make that day, and then determine the specific data extraction strategy based on the number of trips: • When N=1, 30 minutes of travel data are extracted; if the travel time is less than 30 minutes, the actual travel time is used. • When N=2, each trip is truncated to 15 minutes. If the duration of a single trip is less than 15 minutes, it is truncated to the actual trip duration. • When N>=3, data from the first, last, and middle trips are extracted. Data from each trip is extracted in 10-minute increments. If the trip duration is less than 10 minutes, the actual trip duration is used.

5. A computer-readable storage medium storing computer program instructions, characterized in that, When the instruction is executed, it implements the method as described in any one of claims 1.

6. An electronic device, comprising a memory and a processor, characterized in that, The memory stores a computer program product, and the processor executes the program product to implement the method as described in any one of claims 1.

Images