Method and system for supplementing vehicle position loss data
By using pre-trained vector models and support vector machines to fill the vehicle position data in the Internet of Vehicles environment, the problem of vehicle position data packet loss in the Internet of Vehicles is solved, and high-precision and fast data filling are achieved, which meets the Traffic management needs.
Patent Information
- Application Number
- CN202310820431.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-07-05
- Publication Date
- 2025-01-07
AI Technical Summary
In the environment of Internet of Vehicles, due to the increase in the propagation characteristics of wireless communication and the increase in vehicle density, the packet loss of vehicle position data is difficult to achieve high-precision and rapid data replenishment, and cannot meet the needs of traffic management.
The vehicle position data is pre-processed using a pre-trained vector model, and the lost data is filled through the support vector machine and nonlinear constraint approximation equation. The model parameters are adjusted to ensure that the accuracy of the completed data is within the preset range, providing a A method and system for filling out lost data in vehicle locations.
It improves the accuracy and generalization ability of vehicle position data, meets the requirements for high precision in the Internet of Vehicles, is simple and fast to calculate, and realizes real-time online data filling.
Smart Images
Figure CN119272123A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of vehicle position missing data supplementation, and in particular to a vehicle position missing data supplementation method and system. Background Art
[0002] In the Internet of Vehicles environment, due to the propagation characteristics of wireless communication, two wireless transceivers may not be able to perceive each other and send data packets at the same time, resulting in packet conflicts at the receiving end within the reach of the two transmitters. This is the famous hidden terminal problem. Since there is no central entity to coordinate the transmission, and the transmitter is not in the same "cell" (vehicles can perceive each other), hidden nodes will seriously affect the success rate of packet reception. In addition, vehicle density also has an important impact on the successful reception performance of vehicle location broadcast messages. When other conditions in the scene remain unchanged, as the vehicle density increases, the probability of successfully receiving the vehicle location decreases significantly. The above unreliable factors will cause the vehicle location data to be lost, resulting in the loss of information related to the vehicle location.
[0003] Research on traditional data completion processing mainly includes filling in missing data information based on historical change trends and data from different time periods using weighted estimation prediction methods. For example, the linear difference method is used to use the differences in the data relationship matrix to find out the connection between the data, and the missing data is filled based on the adjacent data difference matrix; the probability method is used to study the similarity of data from different time periods, and the missing data is filled using the laws of the data. These methods are based on rough processing of historical data, with low accuracy, and cannot meet the high-precision requirements of vehicle location data for traffic management. With the emergence of support vector machines, this type of method effectively improves the accuracy of location data by predicting historical data, but it still requires the extraction of historical data, which is complicated and takes a long time to process, and cannot meet the needs of efficient and accurate processing of traffic information in the Internet of Vehicles. Summary of the invention
[0004] The purpose of the present invention is to provide a method and system for completing lost vehicle position data, so as to solve the technical problem of how to simply and quickly complete vehicle position data in real time.
[0005] On the one hand, a method for completing missing vehicle position data is provided, comprising:
[0006] Acquire vehicle position data, and preprocess the vehicle position data to obtain sample data;
[0007] Input the preprocessed sample data into a pre-trained vector model to obtain the vehicle position data complement data; wherein the vector model is used to use the historical data of the vehicle position as the past observation value to estimate the corresponding missing value to complement the missing data;
[0008] Determine whether the accuracy of the completed data of the vehicle position data is within a preset allowable accuracy range. If not, adjust the parameters of the vector model, and pre-train the adjusted vector model through the sample data until the accuracy of the completed data is within the preset allowable accuracy range. If within the preset allowable accuracy range, output the completed data of the vehicle position data at this time as the final completed data of the vehicle position data.
[0009] Preferably, the preprocessing of the vehicle position data specifically includes:
[0010] The vehicle position data is input into a preset support vector machine to determine whether the vehicle position data meets the preset learning conditions. If the preset learning conditions are met, the vehicle position data is output as sample data. If the preset learning conditions are not met, the vehicle position data is reacquired to generate corresponding learnable sample data.
[0011] Preferably, the pre-trained vector model specifically includes:
[0012]
[0013] x(τ)=[P(τ-1),P(τ-1),…,P(τ-s)] T
[0014] y=P(τ)
[0015] Among them, f(x) means using historical data to fill the missing data, ω and b are the parameters in the nonlinear constraint approximation, x(τ) is the vehicle position at time τ, represents the preset filling model parameters, P(τ) represents the missing position data on the road at time τ to be filled, y represents the latitude in the vehicle's location information, x represents the longitude in the vehicle's location information, τ represents the time, and T represents the time value.
[0016] Preferably, the vector model is trained by the following process:
[0017] Acquire historical data of vehicle positions, determine the relationship between position data in different time periods according to the time series corresponding to the historical data of vehicle positions, and obtain a corresponding data set, wherein the data set represents the height division corresponding to the historical data and the data in different time periods;
[0018] The data set is input into the vector model to obtain an output result, and the accuracy of the output result is determined. When the accuracy of the output result is within an allowable range, the vector model at this time is output as the trained vector model.
[0019] Preferably, it also includes:
[0020] If the accuracy of the output result is not within the allowed range, the training parameters of the vector model are adjusted and optimized, and the corresponding data set is re-acquired and input into the vector model until the accuracy of the output result is within the allowed range.
[0021] On the other hand, a vehicle position missing data completion system is also provided, which is used to implement the vehicle position missing data completion method, comprising:
[0022] A preprocessing module is used to obtain vehicle position data and preprocess the vehicle position data to obtain sample data;
[0023] A data completion model is used to input the preprocessed sample data into a pre-trained vector model to obtain the completion data of the vehicle position data; wherein the vector model is used to complete the missing data by using the historical data of the vehicle position as the missing value corresponding to the past observation value estimate;
[0024] It is also used to determine whether the accuracy of the completed data of the vehicle position data is within the preset allowable accuracy range. If it is not within the preset allowable accuracy range, the parameters of the vector model are adjusted, and the adjusted vector model is pre-trained with the sample data until the accuracy of the completed data is within the preset allowable accuracy range; if it is within the preset allowable accuracy range, the completed data of the vehicle position data at this time is output as the final completed data of the vehicle position data.
[0025] Preferably, the preprocessing module is specifically used to input the vehicle position data into a preset support vector machine, determine whether the vehicle position data meets the preset learning conditions, if the preset learning conditions are met, output the vehicle position data as sample data, if the preset learning conditions are not met, re-acquire the vehicle position data to generate corresponding learnable sample data.
[0026] Preferably, it also includes:
[0027] A pre-training module is used to perform training through the following process, obtain historical data of vehicle positions, determine the relationship between position data in different time periods according to the time series corresponding to the historical data of vehicle positions, and obtain a corresponding data set, wherein the data set represents the height division corresponding to the historical data and the data in different time periods;
[0028] The data set is input into the vector model to obtain an output result, and the accuracy of the output result is determined. When the accuracy of the output result is within an allowable range, the vector model at this time is output as the trained vector model.
[0029] Preferably, the pre-trained vector model specifically includes:
[0030]
[0031] x(τ)=[P(τ-1),P(τ-1),…,P(τ-s)] T
[0032] y=P(τ)
[0033] Among them, f(x) means using historical data to fill the missing data, ω and b are the parameters in the nonlinear constraint approximation, x(τ) is the vehicle position at time τ, represents the preset filling model parameters, P(τ) represents the missing position data on the road at time τ to be filled, y represents the latitude in the vehicle's location information, x represents the longitude in the vehicle's location information, τ represents the time, and T represents the time value.
[0034] Preferably, the pre-training module is also used to adjust and optimize the training parameters of the vector model if the accuracy of the output result is not within the allowed range, re-obtain the corresponding data set and input it into the vector model until the accuracy of the output result is within the allowed range.
[0035] In summary, the implementation of the embodiments of the present invention has the following beneficial effects:
[0036] Compared with the traditional method of using the existing historical trend of the same attribute to complete the data, the method and system for completing the missing vehicle location data provided by the present invention have better effects in terms of accuracy and generalization ability, and can meet the high-precision requirements for vehicle location data in Internet of Vehicles security applications and traffic management. In addition, since the present invention does not need to extract historical trend data, the calculation is simple and fast, which meets the requirements for real-time and online completion of vehicle location data in the Internet of Vehicles environment. BRIEF DESCRIPTION OF THE DRAWINGS
[0037] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, without paying creative labor, other drawings obtained based on these drawings still belong to the scope of the present invention.
[0038] Figure 1The main flow diagram of a method for completing missing vehicle position data in an embodiment of the present invention is shown.
[0039] Figure 2 Schematic diagram of a vector model in an embodiment of the present invention.
[0040] Figure 3 Schematic diagram of a system for completing missing vehicle position data in an embodiment of the present invention. DETAILED DESCRIPTION
[0041] In order to make the objectives, technical solutions and advantages of the present invention more clear, the present invention will be further described in detail below with reference to the accompanying drawings.
[0042] like Figure 1 FIG. 1 is a schematic diagram of an embodiment of a method for completing missing vehicle position data provided by the present invention. In this embodiment, the method comprises the following steps:
[0043] Step S1, obtain vehicle position data, and preprocess the vehicle position data to obtain sample data; understandably, the sample data is input into the support vector machine, and first determine whether the sample can be learned. If so, use the LS-SVM-BDCA model for training, otherwise re-prepare learnable sample data.
[0044] In a specific embodiment, the preprocessing of the vehicle position data specifically includes inputting the vehicle position data into a preset support vector machine, determining whether the vehicle position data meets preset learning conditions, and if so, outputting the vehicle position data as sample data; if not, reacquiring the vehicle position data to generate corresponding learnable sample data.
[0045] Step S2, input the pre-processed sample data into the pre-trained vector model to obtain the completion data of the vehicle position data; wherein, the vector model is used to use the historical data of the vehicle position as the missing value corresponding to the past observation value estimation to complete the missing data; it is understandable that the traditional data completion algorithm research mainly uses historical data to predict and estimate the missing data, which is essentially based on the specific functional relationship between the missing data and the historical data, and uses the historical data to infer the missing data. This traditional method is very similar to the function approximation problem, so solving the data completion problem can be transformed into solving the approximation problem. Based on this feature, in this embodiment, the vehicle position historical data is studied, and the relationship between the position data in different time periods is found according to the changing time series of the position data, and a data set is established for the given vehicle position original data. The historical data and the data in different time periods are highly divided, and the vehicle position data missing completion problem can be transformed into the non-linear function approximation problem in the support vector machine using LS-SVM-BDCA.
[0046] In a specific embodiment, in the existing method, in the SVM algorithm, in order to process the vector set that is difficult to divide in the low-dimensional space, a nonlinear function is used to map it to a high-dimensional or infinite-dimensional space for division, but the complexity of the calculation is greatly increased. In order to solve this problem, a kernel function is usually introduced to simplify the calculation. Depending on the different kernel functions introduced, the SVM algorithm also has great differences. Common kernel functions are mainly the following:
[0047] Polynomial kernel function: K(x,x i )=[(x·x i )+1] d , where d is the order of the polynomial.
[0048] Linear kernel function: K(x,x i )=x·x i .
[0049] Neural network kernel function (Sigmoid): K(x,x i ) = tanh[v(x·x i )+c], where v and c are constants.
[0050] Gaussian Radial Basis Function (RBF): Where σ is the width function of the RBF function.
[0051] The calculation results of the above kernel functions are different due to different data characteristics. Considering the regularity and stability of vehicle position data, historical data can be used to fill in the missing data to meet the requirements of data accuracy under the premise that the data fluctuation is not large. The most commonly used Gaussian radial basis kernel function is selected as the kernel function for algorithm verification. Whether it is a small sample or a large sample, high dimension or low dimension, the Gaussian radial basis kernel function is applicable. Compared with the polynomial kernel function, the Gaussian radial basis kernel function requires fewer parameters to be determined, and the number of kernel function parameters directly affects the complexity of the function. In addition, when the order of the polynomial is relatively high, the element values of the kernel matrix will tend to infinity or infinitesimal, while the Gaussian radial basis kernel function will reduce the difficulty of numerical calculation. Simplifying the calculation process of nonlinear inequality constraints can ensure the accuracy and precision requirements of filling in the vehicle position number.
[0052] In this embodiment, Figure 2As shown, the sample data is input into the support vector machine. First, it is determined whether the sample can be learned. If it can, the LS-SVM-BDCA model is used for training; otherwise, the learnable sample data is prepared again. For the training sample set of LS-SVM-BDCA, the output result needs to judge the accuracy of the data, and it can only be used within the allowable accuracy range. If the accuracy is insufficient, the training parameters need to be adjusted and optimized, and the LS-SVM-BDCA model training is performed again on the input learnable sample data. When filling in the data, the data filling model can only be used under the condition that both the parameter γ and the kernel width parameter σ2 are determined. To solve the data filling model, let P(t) be the missing data of the road vehicle position that needs to be filled at time t, then P(t - k) represents the missing vehicle position data corresponding to time t - k (k = 1, 2,..., m).
[0053] Among them, the pre-trained vector model specifically includes,
[0054]
[0055] x(τ) = [P(τ - 1), P(τ - 1), …, P(τ - s)] T
[0056] y = P(τ)
[0057] Among them, f(x) represents filling in the missing data using historical data, ω and b represent the parameters in the non-linear constraint approximation formula, x(τ) represents the vehicle position at time τ, represents the preset parameters of the filling model, P(τ) represents the missing position data on the road at time τ to be filled, y represents the dimension in the vehicle position information, x represents the longitude in the vehicle position information, τ represents the time, T represents the time value. Specifically, in the parameters, let P(τ) be the missing position data on the road at time τ to be filled, and P(τ - k) is the vehicle position data on the road at time τ - k (k = 1, 2,..., m). Consider the influence of the vehicle positions in the previous s (s < m) time periods (i.e., the vehicle positions at times τ - 1, τ - 2, …, τ - s) on the vehicle position at time τ. Therefore, a functional relationship or an approximation thereof needs to be found between P(τ) and the previous s vehicle positions (i.e., P(τ - 1), …, P(τ - s)). Let x(τ) = [P(τ - 1), P(τ - 1), …, P(τ - s)] T , y = P(τ), then the model for filling in the missing data using historical data is: Among them: ω and b are both parameters in the above non-linear constraint approximation formula. In the vehicle networking, the vehicle position information can be expressed as X = (X, Y, V). In the formula: X is the longitude; Y is the latitude; V is the speed.
[0058] Specifically, the vector model is trained through the following process: obtaining the historical data of the vehicle position, determining the relationship between the position data of different time periods according to the time series corresponding to the historical data of the vehicle position, and obtaining the corresponding data set, wherein the data set represents the height division corresponding to the historical data and the data of different time periods; inputting the data set into the vector model, obtaining the output result, and judging the accuracy of the output result, when the accuracy of the output result is within the allowable range, outputting the vector model at this time as the trained vector model. If the accuracy of the output result is not within the allowable range, the training parameters of the vector model are adjusted and optimized, and the corresponding data set is obtained again and input into the vector model until the accuracy of the output result is within the allowable range.
[0059] Step S3, determine whether the accuracy of the padding data of the vehicle position data is within the preset accuracy allowable range. If not, adjust the parameters of the vector model, and pre-train the adjusted vector model through the sample data until the accuracy of the padding data is within the preset accuracy allowable range; if it is within the preset accuracy allowable range, output the padding data of the vehicle position data at this time as the final padding data of the vehicle position data. It can be understood that this step is a process of accuracy adjustment, that is, if the padding data result processed by the pre-trained vector model meets the requirements, it is proved that the result is the final result and can be output. However, if this result does not meet the requirements, it is necessary to further train and adjust the vector model.
[0060] like Figure 3 As shown, an embodiment of the present invention further provides a vehicle position missing data completion system, which is used to implement the vehicle position missing data completion method, including:
[0061] A preprocessing module is used to obtain vehicle position data and preprocess the vehicle position data to obtain sample data;
[0062] A data completion model is used to input the preprocessed sample data into a pre-trained vector model to obtain the completion data of the vehicle position data; wherein the vector model is used to complete the missing data by using the historical data of the vehicle position as the missing value corresponding to the past observation value estimate;
[0063] It is also used to determine whether the accuracy of the completed data of the vehicle position data is within the preset allowable accuracy range. If it is not within the preset allowable accuracy range, the parameters of the vector model are adjusted, and the adjusted vector model is pre-trained with the sample data until the accuracy of the completed data is within the preset allowable accuracy range; if it is within the preset allowable accuracy range, the completed data of the vehicle position data at this time is output as the final completed data of the vehicle position data.
[0064] In a specific implementation, the preprocessing module is specifically used to input the vehicle position data into a preset support vector machine to determine whether the vehicle position data meets the preset learning conditions. If the preset learning conditions are met, the vehicle position data is output as sample data. If the preset learning conditions are not met, the vehicle position data is reacquired to generate corresponding learnable sample data.
[0065] Specifically, it also includes:
[0066] A pre-training module is used to perform training through the following process, obtain historical data of vehicle positions, determine the relationship between position data in different time periods according to the time series corresponding to the historical data of vehicle positions, and obtain a corresponding data set, wherein the data set represents the height division corresponding to the historical data and the data in different time periods;
[0067] Input the data set into the vector model, obtain the output result, and judge the accuracy of the output result. When the accuracy of the output result is within the allowed range, output the vector model at this time as the trained vector model. The pre-trained vector model specifically includes:
[0068]
[0069] x(τ)=[P(τ-1),P(τ-1),…,P(τ-s)] T
[0070] y=P(τ)
[0071] Among them, f(x) means using historical data to fill the missing data, ω and b are the parameters in the nonlinear constraint approximation, x(τ) is the vehicle position at time τ, represents the preset filling model parameters, P(τ) represents the missing position data on the road at time τ to be filled, y represents the latitude in the vehicle's location information, x represents the longitude in the vehicle's location information, τ represents the time, and T represents the time value.
[0072] More specifically, the pre-training module is also used to adjust and optimize the training parameters of the vector model if the accuracy of the output result is not within the allowed range, re-obtain the corresponding data set and input it into the vector model until the accuracy of the output result is within the allowed range.
[0073] It should be noted that the system described in the above embodiment corresponds to the method described in the above embodiment. Therefore, the undetailed parts of the system described in the above embodiment can be obtained by referring to the contents of the method described in the above embodiment, and will not be repeated here.
[0074] In summary, the implementation of the embodiments of the present invention has the following beneficial effects:
[0075] Compared with the traditional method of using the existing historical trend of the same attribute to complete the data, the method and system for completing the missing vehicle location data provided by the present invention have better effects in terms of accuracy and generalization ability, and can meet the high-precision requirements for vehicle location data in Internet of Vehicles security applications and traffic management. In addition, since the present invention does not need to extract historical trend data, the calculation is simple and fast, which meets the requirements for real-time and online completion of vehicle location data in the Internet of Vehicles environment.
[0076] The above disclosure is only the preferred embodiment of the present invention, which certainly cannot be used to limit the scope of the present invention. Therefore, equivalent changes made according to the claims of the present invention are still within the scope of the present invention.
Claims
1. A method for completing missing vehicle position data, characterized in that: include: Acquire vehicle position data, and preprocess the vehicle position data to obtain sample data; Input the preprocessed sample data into a pre-trained vector model to obtain the vehicle position data complement data; wherein the vector model is used to use the historical data of the vehicle position as the past observation value to estimate the corresponding missing value to complement the missing data; Determine whether the accuracy of the completed data of the vehicle position data is within a preset allowable accuracy range. If not, adjust the parameters of the vector model, and pre-train the adjusted vector model through the sample data until the accuracy of the completed data is within the preset allowable accuracy range. If within the preset allowable accuracy range, output the completed data of the vehicle position data at this time as the final completed data of the vehicle position data.
2. The method according to claim 1, characterized in that The preprocessing of the vehicle position data specifically includes: The vehicle position data is input into a preset support vector machine to determine whether the vehicle position data meets the preset learning conditions. If the preset learning conditions are met, the vehicle position data is output as sample data. If the preset learning conditions are not met, the vehicle position data is reacquired to generate corresponding learnable sample data.
3. The method according to claim 2, characterized in that The pre-trained vector model specifically includes: x(τ)=[P(τ-1),P(τ-1),…,P(τ-s)] T y=P(τ) Among them, f(x) means using historical data to fill the missing data, ω and b are the parameters in the nonlinear constraint approximation, x(τ) is the vehicle position at time τ, represents the preset filling model parameters, P(τ) represents the missing position data on the road at time τ to be filled, y represents the latitude in the vehicle's location information, x represents the longitude in the vehicle's location information, τ represents the time, and T represents the time value.
4. The method according to claim 3, characterized in that The vector model is trained through the following process, Acquire historical data of vehicle positions, determine the relationship between position data in different time periods according to the time series corresponding to the historical data of vehicle positions, and obtain a corresponding data set, wherein the data set represents the height division corresponding to the historical data and the data in different time periods; The data set is input into the vector model to obtain an output result, and the accuracy of the output result is determined. When the accuracy of the output result is within an allowable range, the vector model at this time is output as the trained vector model.
5. The method according to claim 4, characterized in that Also includes, If the accuracy of the output result is not within the allowed range, the training parameters of the vector model are adjusted and optimized, and the corresponding data set is re-acquired and input into the vector model until the accuracy of the output result is within the allowed range.
6. A system for completing missing vehicle position data, used to implement the method according to any one of claims 1 to 5, characterized in that: include: A preprocessing module is used to obtain vehicle position data and preprocess the vehicle position data to obtain sample data; A data completion model is used to input the preprocessed sample data into a pre-trained vector model to obtain the completion data of the vehicle position data; wherein the vector model is used to complete the missing data by using the historical data of the vehicle position as the missing value corresponding to the past observation value estimate; It is also used to determine whether the accuracy of the completed data of the vehicle position data is within the preset allowable accuracy range. If it is not within the preset allowable accuracy range, the parameters of the vector model are adjusted, and the adjusted vector model is pre-trained with the sample data until the accuracy of the completed data is within the preset allowable accuracy range; if it is within the preset allowable accuracy range, the completed data of the vehicle position data at this time is output as the final completed data of the vehicle position data.
7. The system according to claim 6, characterized in that The preprocessing module is specifically used to input the vehicle position data into a preset support vector machine, determine whether the vehicle position data meets the preset learning conditions, if the preset learning conditions are met, output the vehicle position data as sample data, if not, reacquire the vehicle position data to generate corresponding learnable sample data.
8. The system according to claim 7, characterized in that Also includes, A pre-training module is used to perform training through the following process, obtain historical data of vehicle positions, determine the relationship between position data in different time periods according to the time series corresponding to the historical data of vehicle positions, and obtain a corresponding data set, wherein the data set represents the height division corresponding to the historical data and the data in different time periods; The data set is input into the vector model to obtain an output result, and the accuracy of the output result is determined. When the accuracy of the output result is within an allowable range, the vector model at this time is output as the trained vector model.
9. The system according to claim 8, characterized in that The pre-trained vector model specifically includes: x(τ)=[P(τ-1),P(τ-1),…,P(τ-s)] T y=P(τ) Among them, f(x) means using historical data to fill the missing data, ω and b are the parameters in the nonlinear constraint approximation, x(τ) is the vehicle position at time τ, represents the preset filling model parameters, P(τ) represents the missing position data on the road at time τ to be filled, y represents the latitude in the vehicle's location information, x represents the longitude in the vehicle's location information, τ represents the time, and T represents the time value.
10. The system according to claim 9, characterized in that The pre-training module is also used to adjust and optimize the training parameters of the vector model if the accuracy of the output result is not within the allowed range, re-obtain the corresponding data set and input it into the vector model until the accuracy of the output result is within the allowed range.