A data acquisition system and method

Abstract

The present invention discloses a data acquisition system and method, relating to the technical field of data acquisition. It includes constructing a three-dimensional state space based on vehicle states, mapping real-time vehicle states into vehicle state codes; collecting road condition features to construct a road condition feature matrix; reducing the dimension of the vehicle state code vector to a two-dimensional vector, performing a Kronecker product operation with the road condition feature matrix to generate a policy matrix; retaining the first n principal components after performing singular value decomposition on the policy matrix to form a compressed policy vector, and dynamically determining the acquisition frequency; linearly mapping the element values of the policy matrix to generate a grayscale map, and determining the adjustment amplitude of the acquisition frequency using the change trend of the grayscale values. The present invention improves the timeliness, stability and accuracy of the data acquisition system, and provides more efficient and reliable data support for the intelligent transportation system and the field of autonomous driving.

Description

Technical Field

[0001] The present invention relates to the technical field of data acquisition, and particularly to a data acquisition system and method. Background Art

[0002] In the context of the rapid development of modern intelligent transportation systems and autonomous driving technologies, the data acquisition system of vehicles has become one of the key technologies for vehicle operation state monitoring, road condition identification, and driving safety control. Existing data acquisition systems mainly rely on a fixed sampling frequency to collect vehicle state parameters (such as speed, acceleration, engine speed, etc.) and road condition information (such as slope, slipperiness, traffic flow, etc.). Such systems generally use a sensor array to perform periodic sampling on the target information and combine signal processing algorithms to extract features from the collected data. However, the traditional fixed-frequency acquisition method has the following problems: on the one hand, a lower sampling frequency may lead to insufficient data accuracy, affecting the rapid response to sudden road conditions and emergency states; on the other hand, blindly increasing the sampling frequency can enhance the timeliness and integrity of the data, but at the same time will greatly increase the pressure of data transmission and processing, resulting in waste of computing resources and even affecting the real-time performance of the system.

[0003] In the prior art, some studies have attempted to analyze vehicle states and road condition characteristics through machine learning models or expert systems and introduced a dynamic sampling mechanism to a certain extent to optimize the data acquisition process. For example, a method based on a time series prediction model has been proposed, which uses historical data to predict the future road condition change trend and adjusts the sampling frequency based on the prediction result. However, such methods have obvious deficiencies: firstly, the performance of the time series prediction model depends on a large amount of high-quality historical data. Once facing sudden situations (such as sudden vehicle failures, extreme weather, or sudden traffic events), the prediction accuracy of the model often fails to meet the actual requirements; secondly, some methods only rely on a single variable (such as speed or acceleration) when adjusting parameters, and do not fully consider the comprehensive influence of vehicle states, road condition characteristics, and their correlations, resulting in easy misjudgment under complex working conditions; in addition, existing methods mostly use fixed thresholds or static adjustment strategies in the process of sampling frequency adjustment, lacking an adaptive control mechanism for the frequency change amplitude, and it is difficult to achieve the best balance between data accuracy and system overhead.

[0004] Therefore, how to reasonably adjust the sampling frequency according to the dynamic changes of vehicle states and road condition characteristics to improve the effectiveness of data acquisition and the stability of the system has become an important research direction in this field. Summary of the Invention

[0005] In view of the problems existing in the above-mentioned traditional fixed-frequency acquisition method, the present invention is proposed.

[0006] Therefore, the problem to be solved by the present invention is how to improve the timeliness, stability and accuracy of the data acquisition system.

[0007] To solve the above technical problems, the present invention provides the following technical solutions:

[0008] In a first aspect, the present invention provides a data acquisition system, which includes a vehicle state dynamic coding generation module that constructs a three-dimensional state space based on the vehicle state and maps the real-time vehicle state to a vehicle state code; a road condition feature matrix modeling module that collects road condition features to construct a road condition feature matrix; a collection frequency adjustment module that reduces the vehicle state coding vector to a two-dimensional vector, performs a Kronecker product operation with the road condition feature matrix to generate a policy matrix; retains the first n principal components after performing a singular value decomposition on the policy matrix to form a compressed policy vector, and dynamically determines the collection frequency; a grayscale map-driven collection module that linearly maps the policy matrix element values to generate a grayscale map, and determines the adjustment amplitude of the collection frequency using the change trend of the grayscale values.

[0009] As a preferred solution of the data acquisition system of the present invention, wherein: the vehicle state includes the battery SOC value, vehicle speed, and driving mode; the road condition features include road surface gradient, bump index, and curve radius parameter of the curve.

[0010] As a preferred solution of the data acquisition system of the present invention, wherein: after obtaining the vehicle state, it is classified according to a set value to form a classification index; after constructing a three-dimensional state space according to the classification index, the vehicle state is represented by a state vector where respectively correspond to the SOC interval, vehicle speed classification, and driving mode.

[0011] As a preferred solution of the data acquisition system of the present invention, wherein: the state vector is mapped to a 6-bit binary vehicle state code, where the first 2 bits represent the SOC interval, the middle 2 bits represent the vehicle speed classification, and the last 2 bits represent the driving mode.

[0012] As a preferred solution of the data acquisition system of the present invention, wherein: the vehicle state coding vector is input into an adaptive projection matrix to generate a two-dimensional vector; wherein, the weight of the first row of the projection matrix is dynamically adjusted according to the SOC value.

[0013] As a preferred embodiment of the data acquisition system of the present invention, the construction of the adaptive projection matrix includes: calculating the influence degree between different state parameters according to each component of the vehicle state coding vector to form a state feature gradient matrix; each element of the state feature gradient matrix represents the relative change degree between two state parameters; according to the change trend of the state feature gradient matrix, extracting the most significant change direction as the main projection direction, and selecting the first two principal directions to construct the projection matrix.

[0014] As a preferred embodiment of the data acquisition system of the present invention, the dynamic adjustment according to the SOC value includes: calculating the adjustment amplitude of the vehicle state projection vector when the SOC value changes, and adjusting the main direction of the projection matrix according to the change amplitude of the projection vector.

[0015] As a preferred embodiment of the data acquisition system of the present invention, the dynamic confirmation of the acquisition frequency includes: calculating the feature mean value according to the compression strategy vector; if the feature mean value is within the set threshold range, maintaining the current acquisition frequency; if the feature mean value is lower than the lowest threshold, reducing the acquisition frequency; if the feature mean value is higher than the highest threshold, increasing the acquisition frequency.

[0016] As a preferred embodiment of the data acquisition system of the present invention, the determination of the adjustment amplitude of the acquisition frequency by using the gray value change trend includes: calculating the regional texture complexity of the gray scale map, analyzing the local gray change pattern, and adjusting the acquisition parameters to different-dimensional sampling modes according to the texture direction characteristics of the gray scale map.

[0017] In a second aspect, the present invention provides a data acquisition method, which includes: constructing a three-dimensional state space based on the vehicle state, mapping the real-time vehicle state into a vehicle state code; collecting road condition features to construct a road condition feature matrix; reducing the vehicle state coding vector to a two-dimensional vector, performing a Kronecker product operation with the road condition feature matrix to generate a policy matrix; performing singular value decomposition on the policy matrix and retaining the first n principal components to form a compression policy vector, and dynamically confirming the acquisition frequency; linearly mapping the element values of the policy matrix to generate a gray scale map, and determining the adjustment amplitude of the acquisition frequency by using the gray value change trend.

[0018] In a third aspect, the present invention provides a computer device, including a memory and a processor, where the memory stores a computer program, and: when the computer program instructions are executed by the processor, the data acquisition system described in the first aspect of the present invention is implemented.

[0019] Fourthly, the present invention provides a computer-readable storage medium, on which a computer program is stored, wherein: when the computer program instructions are executed by a processor, a data acquisition system as described in the first aspect of the present invention is implemented.

[0020] The beneficial effects of the present invention are as follows: by constructing a vehicle state dynamic coding generation module and a road condition feature matrix modeling module, the vehicle state and road condition features are mapped to a high-dimensional state space, and the Kronecker product operation of the coding vector and the road condition feature matrix is used to achieve efficient compression of data features. By introducing a gray scale map driving mechanism and dynamically determining the adjustment amplitude of the sampling frequency using the change trend of gray values, the response ability of the system to emergencies and complex road conditions is further improved. The present invention enhances the timeliness, stability, and accuracy of a data acquisition system, providing more efficient and reliable data support for the intelligent transportation system and the field of autonomous driving. Description of the Drawings

[0021] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention, and those of ordinary skill in the art can obtain other drawings without creative efforts based on these drawings.

[0022] Figure 1 It is a structural diagram of a data acquisition system provided by an embodiment of the present invention.

[0023] Figure 2 It is a schematic structural diagram of an acquisition frequency adjustment module in a data acquisition system provided by an embodiment of the present invention.

[0024] Figure 3 It is a schematic flow diagram of a data acquisition method provided by an embodiment of the present invention.

[0025] In the figure: 100, vehicle state dynamic coding generation module; 200, road condition feature matrix modeling module; 300, acquisition frequency adjustment module; 400, gray scale map driven acquisition module; 310, coding dimension reduction unit; 320, policy matrix generation unit; 330, principal component extraction unit; 340, parameter dynamic confirmation unit. Detailed Embodiments

[0026] In order to make the above objects, features, and advantages of the present invention more obvious and understandable, the following will make a detailed description of the specific embodiments of the present invention in conjunction with the drawings of the specification.

[0027] In the following description, many specific details are set forth in order to provide a thorough understanding of the present invention. However, the present invention may be practiced in other ways different from those described herein. Those skilled in the art can make similar generalizations without departing from the spirit of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0028] As described in the above background art, in the prior art, some studies have attempted to analyze vehicle states and road condition features through machine learning models or expert systems, and to a certain extent, introduced a dynamic sampling mechanism to optimize the data collection process. For example, some studies have proposed a method based on a time series prediction model, which uses historical data to predict future road condition change trends and adjusts the sampling frequency based on the prediction results. However, such methods have obvious deficiencies: First, the performance of the time series prediction model depends on a large amount of high-quality historical data. Once faced with unexpected situations (such as sudden vehicle failures, extreme weather, or sudden traffic events), the prediction accuracy of the model often fails to meet the actual requirements; Second, some methods only rely on a single variable (such as speed or acceleration) when adjusting parameters, and do not fully consider the comprehensive effects of vehicle states, road condition features, and their correlations, resulting in easy misjudgment under complex working conditions; In addition, existing methods mostly adopt fixed thresholds or static adjustment strategies in the process of adjusting the sampling frequency, lacking an adaptive control mechanism for the amplitude of frequency change, and it is difficult to achieve the best balance between data accuracy and system overhead. Therefore, the present invention proposes a data collection system and method.

[0029] The following elaborates in detail on the implementation details of the technical solution of this application:

[0030] Refer to Figures 1 - 3 , Figure 1 shows a structural diagram of a data collection system according to an embodiment of the present application. Refer to Figure 1 As shown, the data collection system includes: a vehicle state dynamic coding generation module 100, which constructs a three-dimensional state space based on the vehicle state and maps the real-time vehicle state to a vehicle state code; a road condition feature matrix modeling module 200, which collects road condition features to construct a road condition feature matrix; a collection frequency adjustment module 300, which reduces the vehicle state coding vector to a two-dimensional vector, performs a Kronecker product operation with the road condition feature matrix to generate a policy matrix; after performing singular value decomposition on the policy matrix, retains the first n principal components to form a compressed policy vector, and dynamically determines the collection frequency; a grayscale map-driven collection module 400, which linearly maps the policy matrix element values to generate a grayscale map, and determines the adjustment amplitude of the collection frequency using the change trend of the grayscale values.

[0031] In an embodiment of the present invention, the vehicle state dynamic coding generation module 100 specifically includes:

[0032] Vehicle state data is a set of parameters that change dynamically during vehicle operation. In this embodiment, three key state variables are mainly concerned, namely the battery SOC value, vehicle speed, and driving mode. The SOC value represents the percentage of the remaining battery power at present, usually ranging from 0% to 100%, and is the core parameter for the energy management of new energy vehicles; the vehicle speed is the driving speed of the vehicle, and the unit is usually km / h (kilometers per hour). Different vehicle speed ranges correspond to different driving requirements, such as low-speed conditions, high-speed cruising, etc.; the driving mode refers to the current power output mode of the vehicle, such as pure electric mode, hybrid mode, engine mode, etc. There are significant differences in the energy distribution strategies under different modes.

[0033] The methods for obtaining vehicle state data include but are not limited to: Modern vehicles generally use the CAN bus to transmit the data of each vehicle sensor in real time. In this embodiment, the SOC, vehicle speed, and driving mode information are collected through the CAN interface and polled regularly to ensure the timeliness of the data. The SOC value is usually calculated by the BMS (Battery Management System) and stored in the ECU. The vehicle speed is measured by the wheel speed sensor, and the driving mode is determined by the power control unit and read through the standard diagnostic protocol (such as OBD-II or UDS).

[0034] The obtained vehicle state data may have time delay or noise. Therefore, in this embodiment, methods such as Kalman filtering are used for data smoothing processing to ensure data continuity and accuracy.

[0035] After obtaining the vehicle state, it is classified according to the set value to form a classification index.

[0036] It should be noted that in order to facilitate data mapping and coding, it is necessary to perform hierarchical processing on the SOC value, vehicle speed, and driving mode. The conventional method usually uses fixed threshold division. For example, the SOC value is divided into four levels at intervals of 25%, the vehicle speed is divided into several gears at intervals of 20 km / h, and the driving mode is classified according to the predefined mode. However, this fixed classification scheme cannot adapt to the changes under different vehicle models and different driving environments. Therefore, the present invention adopts an adaptive classification strategy:

[0037] SOC classification: Based on the historical SOC change curve and typical working conditions, the SOC classification threshold is adaptively adjusted to make the SOC classification more in line with the vehicle energy consumption characteristics. For example, in urban working conditions, the SOC change is small, and a finer classification can be adopted; in high-speed working conditions, the SOC change is large, and a coarser classification can be adopted.

[0038] Vehicle speed classification: The vehicle speed classification interval is adaptively determined. For example, in a congested environment, the low-speed interval can be divided into more levels to better reflect the influence of small speed changes, while in high-speed cruising, a larger interval is adopted to reduce the coding complexity.

[0039] Driving mode classification: By analyzing the vehicle operation data, the driving mode with the highest actual usage frequency is identified and grouped according to priority. For example, if a certain mode has a low usage frequency in a specific vehicle model, it can be merged into the adjacent mode to reduce the state space scale.

[0040] After completing the grading of state variables, the present invention constructs a three-dimensional state space, where: the X-axis represents the SOC interval, that is, discrete indexes are formed according to the graded SOC value range; the Y-axis represents the vehicle speed grading, that is, different vehicle speed gears correspond to different index values; the Z-axis represents the driving mode, that is, different driving modes correspond to different state indexes. The vehicle state is represented by the state vector where correspond to the SOC interval, vehicle speed grading, and driving mode respectively.

[0041] The construction method of the three-dimensional state space makes the representation of the vehicle state more intuitive and can be used for subsequent mapping and coding processing. Conventional technologies usually adopt the direct indexing method, that is, the index values of SOC, vehicle speed, and driving mode are simply combined to form a state vector. However, the present invention adopts a state vector optimization strategy, that is, considering the correlation between state variables, the state indexes are normalized and adjusted to reduce redundant states and improve coding efficiency.

[0042] The state vector is mapped to a 6-bit binary vehicle state code, where the first 2 bits represent the SOC interval, the middle 2 bits represent the vehicle speed grading, and the last 2 bits represent the driving mode.

[0043] Specifically, the specific mapping rules are as follows: the first 2 bits represent the SOC interval, that is, the index value after SOC grading is converted into 2-bit binary; the middle 2 bits represent the vehicle speed grading, that is, the vehicle speed grading index is converted into 2-bit binary; the last 2 bits represent the driving mode, that is, the driving mode index is converted into 2-bit binary.

[0044] Conventional methods usually adopt direct indexing mapping, that is, the state index value is converted into a binary number in a fixed format. However, the disadvantage of this method is that the information utilization rate is low, and some state indexes may not be fully encoded. The present invention adopts an optimal state index allocation strategy, that is, according to the vehicle historical operation data, the binary mapping method of the state indexes is adjusted, so that the high-frequency state indexes occupy a more compact binary coding space, while the low-frequency state indexes are assigned to more distant coding positions to improve the information utilization efficiency.

[0045] Optionally, to ensure the real-time performance and accuracy of the encoded data, the present invention adopts a dynamic encoding update strategy, that is: set the encoding update frequency to 10 Hz, which means 10 state encoding updates are performed per second to ensure timely reflection of state changes. When the vehicle state changes rapidly, such as a sudden drop in SOC, a sharp change in vehicle speed, or a drive mode switch, etc., the encoding update frequency is temporarily increased to 20 Hz to capture key state changes. When the vehicle is in a stable operating state, such as high-speed cruising or low-load operation, the encoding update frequency can be reduced to 5 Hz to reduce the computational overhead and data storage pressure. The encoding update mechanism of the present invention avoids the problems of information lag or data redundancy caused by a fixed update frequency in traditional methods, enabling the encoded data to more accurately reflect the real-time operating state of the vehicle.

[0046] In the embodiment of the present invention, the objective of the road condition feature matrix modeling module 200 is to construct a 3×3 road condition feature matrix that can accurately describe the road state by collecting and processing key road condition feature parameters. This matrix is used for optimizing the vehicle operating state, so as to more reasonably adjust the driving strategy and improve driving safety and comfort.

[0047] The road condition features include road surface slope, bump index, and curve radius of curvature parameters.

[0048] Specifically, the slope refers to the inclination degree of the road, usually expressed in terms of angle (θ, unit: °) or percentage (%). The present invention adopts the angle form, and the range is set to -15° to +15°, where a positive value indicates an uphill slope and a negative value indicates a downhill slope.

[0049] The present invention combines an in-vehicle IMU (Inertial Measurement Unit) with a real-time data fusion algorithm to enhance the accuracy of slope measurement. Compared with the traditional single GPS measurement method, this method can provide more stable slope data in complex environments such as tunnels and mountains. In addition, the update frequency of the slope data is adaptively adjusted: in sections where the slope changes slowly, the sampling interval is set to 1 s to reduce the computational burden; in sections where the slope changes sharply, the sampling interval is shortened to 0.2 s to capture slope mutation information.

[0050] Exemplarily, the vehicle-mounted IMU is installed at the vehicle chassis or the inertial center position to reduce measurement errors caused by vehicle jolts and body sway; a three-axis accelerometer is used to measure the acceleration of the vehicle in the front-rear direction, and the vehicle pitch rate is obtained by combining with a three-axis gyroscope. In a short time range (such as 0.1 second to 1 second), the IMU data is still relatively reliable, so the slope angle can be directly calculated through the quaternion attitude solution algorithm; dual-sensor complementary filtering is adopted, that is, the accelerometer data is used for low-frequency compensation (to reduce drift), and the gyroscope data is used for high-frequency compensation (to reduce instantaneous jitter). Since there is zero-bias drift in the IMU, error accumulation may occur during long-term calculations. Therefore, the present invention further introduces a GPS elevation change correction strategy: calculate the elevation change of the vehicle within a certain distance through GPS, and compare it with the slope angle measured by the IMU; adopt a sliding window algorithm to perform adaptive adjustment within a certain time range (such as the data in the past 5 seconds) to correct the IMU drift error. This can still maintain a high slope measurement accuracy in complex environments (such as tunnels, bridges, etc.).

[0051] The bump index is used to measure the unevenness of the road, which is measured by a road detector or ground radar. In this embodiment, the bump index is divided into levels 0-10, where level 0 represents a completely flat road and level 10 represents an extremely bumpy road surface.

[0052] The degree of curvature of a road is usually characterized by the radius of curvature. The larger the radius of curvature, the gentler the curve; the smaller the radius of curvature, the sharper the curve. The present invention calculates the curvature based on the GPS trajectory and adopts a dynamic grading method, only recording the radius of curvature ≥ 50m and ignoring small curvature fluctuations, so as to avoid unnecessary interference with the road condition assessment.

[0053] Furthermore, a 3×3 road condition feature matrix is constructed. Its main purpose is to map the slope, bump index, and curve curvature information into a unified data structure through matrix representation methods for subsequent calculations and optimizations. Among them, the first column represents the slope information, that is, the influence factors in different slope intervals; the second column represents the bump index, that is, the quantization value of the road unevenness; the third column represents the radius of curvature of the curve, which is used to describe the characteristics of the curve road condition.

[0054] Preferably, the slope influence factor , and its calculation method is , where is the slope angle. For example, in the case of a slope of ±5°, the slope influence factor = 1.25, and when the slope is ±10°, = 1.5, that is, the influence of the slope on the road condition feature matrix is amplified as the slope increases.

[0055] Since the dimensions of different road condition parameters are different (the slope is in degrees, the bump index is in the range of 0 - 10 levels, and the radius of curvature is in meters), the min-max normalization method needs to be used to map the matrix element values to the interval [0, 1] so that unified calculations and comparisons can be made between different variables.

[0056] It should be noted that in traditional road condition feature modeling methods, the slope, bump index, and radius of curvature are usually measured and processed independently, failing to form a unified data structure. This results in uneven weight distribution of different road condition parameters in subsequent optimization calculations, affecting the effectiveness of the data. The present invention constructs a 3×3 road condition feature matrix, fuses three key road condition information under the same matrix framework, and introduces a dynamic adjustment strategy, improving the self-adaptability and generalization ability of the matrix, thereby more reasonably adjusting the driving strategy, enhancing driving safety and comfort; and combines IMU and GPS for high-precision slope measurement, using dual-sensor complementary filtering to reduce errors, so that the slope data still maintains high accuracy in complex environments.

[0057] In the embodiment of the present invention, as Figure 2 shown, the acquisition frequency adjustment module 300 includes an encoding and dimensionality reduction unit 310: reducing the vehicle state encoding vector to a two-dimensional vector through a dynamic projection matrix, where the weight distribution of the first row of the dynamic projection matrix is dynamically adjusted according to the SOC value to reflect the influence of the remaining battery power on the acquisition frequency; a policy matrix generation unit 320: performing a Kronecker product operation on the reduced two-dimensional vector and the road condition feature matrix to generate a 6×6 policy matrix; a principal component extraction unit 330: performing singular value decomposition (SVD) on the policy matrix and retaining the first 3 principal components to form a compressed policy vector; a parameter dynamic confirmation unit 340: dynamically confirming the acquisition parameters based on the compressed policy vector to adjust the frequency of data acquisition and achieve adaptive control of the frequency.

[0058] Specifically, the main task of the encoding and dimensionality reduction unit 310 is to reduce the dimension of the high-dimensional vehicle state encoding vector through a dynamic projection matrix, enabling the system to capture the main features in a low-dimensional representation, thereby reducing the computational complexity and ensuring the retention of the core information. The specific operations are as follows.

[0059] Input the vehicle state encoding vector into the adaptive projection matrix to generate a two-dimensional vector; among them, the weight of the first row of the projection matrix is dynamically adjusted according to the SOC value.

[0060] Specifically, for each component of the vehicle state encoding vector, calculate the influence degree between different state parameters to form a state feature gradient matrix. Each element in the matrix represents the relative change degree between two state parameters, which is specifically normalized by the change rates of the two to measure the degree of their mutual influence. If the change trends of the two parameters are the same, a relatively high positive value is assigned; if the change trends are opposite, a relatively high negative value is assigned; if there is no obvious correlation, a value close to zero is assigned.

[0061] Decompose the state feature gradient matrix, extract the direction with the most significant change as the main projection direction, and select the first two principal directions to construct a projection matrix. These two directions respectively represent the main change trends of the state parameters, enabling the projection matrix to be dynamically adjusted according to the characteristics of the actual state data. When selecting the directions, corresponding weights are given to different directions according to the influence degree in the state feature gradient matrix to make it more conform to the real change situation of the vehicle state.

[0062] Use the constructed projection matrix to transform the vehicle state encoding vector, mapping the original high-dimensional vector to a two-dimensional vector space. During the mapping process, the contribution degree of each state component is determined by the weight of the projection matrix, ensuring that the projected two-dimensional vector can still maintain the main characteristics of the original state information.

[0063] The dynamic adjustment according to the SOC value includes: calculating the adjustment amplitude of the vehicle state projection vector when the SOC value changes, and adjusting the main direction of the projection matrix according to the change amplitude of the projection vector.

[0064] Exemplarily, partition the SOC value and calculate its change trend within the historical time window to determine whether the SOC value is in a state of rapid decline, stability, or rapid increase. According to different states, adjust the weight allocation method of the projection matrix. Specifically: when the SOC value is rapidly declining, increase the weights of energy-related states (such as vehicle speed and driving mode) in the projection matrix to enhance the influence of the energy consumption state; when the SOC value is stable, keep the direction of the original projection matrix unchanged to maintain the stability of the data acquisition strategy; when the SOC value is rapidly increasing, reduce the influence of energy-related states, making the projection matrix pay more attention to the change trends of other vehicle states. This dynamic adjustment method enables the projection matrix to be adaptively optimized according to the change trend of the SOC value, rather than relying on fixed weight settings.

[0065] It can be seen that traditional high-dimensional state vectors usually require complete feature modeling. This method reduces the data dimension to two dimensions through dimensionality reduction, improving the computational efficiency in a vehicle environment with limited computing resources; through the selection of the projection direction based on feature gradients, the information after dimensionality reduction can still highly summarize the vehicle state, ensuring the accuracy of the data acquisition strategy; different weight distributions are adopted in different SOC value intervals, enabling the data after dimensionality reduction to accurately reflect the vehicle's energy consumption state and optimizing the data acquisition decision under energy consumption-sensitive states. The prior art generally uses a fixed projection matrix for dimensionality reduction, ignoring the dynamic changes in the vehicle state. The dynamic projection matrix of the present invention not only enhances the adaptability of dimensionality reduction but also avoids the information loss that may be caused by traditional dimensionality reduction methods, making the data acquisition strategy more in line with the real-time vehicle working conditions.

[0066] The policy matrix generation unit 320 includes: performing operations on the two-dimensional state projection vector after dimensionality reduction and the road condition feature matrix, and combining the main features of the vehicle state and road condition information to generate a 6×6 policy matrix. Each element of this matrix represents the data acquisition strategy weight under a specific state and specific environmental conditions. The update frequency of the policy matrix is consistent with the update frequency of the vehicle state encoding to ensure that the data acquisition strategy can be optimized and adjusted at any time; when the vehicle enters different road conditions or the state changes drastically, the policy matrix can be adaptively adjusted to improve the real-time performance and accuracy of data acquisition.

[0067] It should be noted that compared with formulating strategies based on only a single data source (such as vehicle state or road characteristics), the present invention integrates the information of both, ensuring the adaptability of the data acquisition strategy to different working conditions; the Kronecker product operation retains the coupling relationship between the state and environmental information in terms of structure and has a low computational complexity, making it suitable for the in-vehicle computing environment. When the vehicle enters an emergency (such as sudden braking, slippery road surface) or the road conditions change drastically (such as turning from an urban road to a highway), the policy matrix can be quickly adjusted, thereby improving the reliability and response speed of data acquisition.

[0068] The principal component extraction unit 330 includes: performing SVD decomposition on the policy matrix and extracting the first 3 principal components to reduce the data dimension and retain the key features. That is, the first 3 principal components with a cumulative contribution rate exceeding 95% are selected, so that the compressed policy vector can still represent the core information of the original policy matrix. The finally formed compressed policy vector is used for subsequent dynamic parameter confirmation to optimize the adaptive control of the data acquisition frequency.

[0069] The parameter dynamic confirmation unit 340 includes: dynamically confirming the acquisition frequency, including: calculating the feature mean value according to the compressed policy vector. If the feature mean value is within the set threshold range, the current acquisition frequency is maintained; if the feature mean value is lower than the lowest threshold, the acquisition frequency is reduced; if the feature mean value is higher than the highest threshold, the acquisition frequency is increased.

[0070] Compared with the prior art, the strategy matrix optimization and adaptive acquisition frequency adjustment scheme of the present invention ensure efficient and accurate acquisition of vehicle state data in an environment with limited computing resources.

[0071] In the embodiment of the present invention, the grayscale map-driven acquisition module 400 specifically includes: when the strategy matrix changes rapidly, linearly map the element values of the strategy matrix to generate a grayscale map, use the change trend of the grayscale values to determine the adjustment amplitude of the acquisition frequency. At the same time, for the regional distribution characteristics of the grayscale map, calculate the image gradient change rate, identify the mutation region and trigger local high-frequency sampling.

[0072] Among them, determining the adjustment amplitude of the acquisition frequency using the change trend of the grayscale values includes: generating a grayscale map corresponding to the element values of the strategy matrix, adopting a non-uniform mapping method, making the color scale change more sensitive in the low numerical range, and the color scale change more smooth in the high numerical range, so as to enhance the perception ability of the edge change region.

[0073] Exemplarily, the following mapping formula can be used:

[0074]

[0075] Among them, represents the grayscale value, is the element value at the corresponding position of the strategy matrix, and are the maximum and minimum element values of the strategy matrix respectively, is the segmentation threshold, and the low numerical range is linearly mapped to , and the high numerical range is mapped to (127, 255], realizing low-value sensitivity and high-value smoothness.

[0076] Since the numerical values in the strategy matrix may show large interval differences, direct linear mapping may cause loss of some details. Therefore, a non-uniform piecewise linear mapping method is adopted, so that the low numerical range (less changing region) has higher grayscale resolution and improves the perception ability of subtle changes; the high numerical range (more changing region) is smoothly mapped to avoid acquisition instability caused by grayscale mutations.

[0077] Furthermore, calculate the regional texture complexity of the grayscale map, and use a structure adaptive filter to calculate the grayscale gradient change rate. Exemplarily:

[0078]

[0079] Among them, and respectively represent the gradients of the grayscale map in the x and y directions, and the gradient calculation is performed by the Sobel operator or the Roberts operator. is the local texture complexity; the mean value of the texture change in the local area is calculated as follows:

[0080]

[0081] where is the texture complexity. If is relatively high, it indicates that the texture details are rich and there may be important state changes. is the total number of pixels in the local area; represents the gray-scale change rate at the pixel point . If , reduce the acquisition density in a certain direction and optimize the computing resources; if , then increase the global sampling frequency to capture potential important information; where and are the maximum and minimum values of the texture complexity threshold.

[0082] Furthermore, according to the texture direction characteristics of the map, the acquisition parameters are adjusted to sampling modes in different dimensions. For example, when the grayscale map presents directional textures (such as continuous patches or gradient distributions), reduce the acquisition frequency along the main texture direction and increase the sampling density in the vertical direction; when the grayscale map has no obvious structural features, increase the global acquisition frequency to enhance the data coverage.

[0083] During the data acquisition process, monitor the change trend of the texture complexity of the map. If the texture complexity is maintained in a certain stable interval, adopt the standard sampling mode; if the complexity changes suddenly, temporarily switch to the high-density sampling mode to ensure the data integrity of the mutation area; if the complexity decreases, reduce the redundant data acquisition and improve the system efficiency.

[0084] It should be noted that when the element values of the policy matrix change violently in a short period of time (for example, within adjacent time steps, the change amplitudes of multiple elements exceed the set threshold), the grayscale map-driven acquisition mode will be enabled. If the policy matrix changes slowly or is in a stable state, it will be executed according to the default sampling frequency set by the acquisition frequency adjustment module 300 without additional adjustment. The rapid change of the policy matrix usually corresponds to the violent fluctuation of the external environment or the system state, such as the behavior change of the target object, the sudden change of the sensor input signal, etc. These situations require a higher sampling frequency to ensure the data integrity. If the policy matrix is in a stable state, forcibly increasing the sampling frequency will increase the computing burden and lead to redundant data storage. Therefore, only start the grayscale map-driven acquisition module 400 when necessary to optimize the computing resources.

[0085] Further, this embodiment also provides a data acquisition method, as Figure 3 shown, including:

[0086] S1. Construct a three-dimensional state space based on the vehicle state, and map the real-time vehicle state into a vehicle state code;

[0087] S2. Collect road condition features to construct a road condition feature matrix;

[0088] S3. Reduce the vehicle state code vector to a two-dimensional vector, perform a Kronecker product operation with the road condition feature matrix to generate a policy matrix; after performing singular value decomposition on the policy matrix, retain the first n principal components to form a compressed policy vector, and dynamically confirm the acquisition frequency;

[0089] S4. Linearly map the element values of the policy matrix to generate a grayscale map, and use the change trend of the grayscale values to determine the adjustment range of the acquisition frequency.

[0090] This embodiment also provides a computer device, applicable to the situation of a data acquisition system, including a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the data acquisition system proposed in the above embodiment.

[0091] This computer device may be a terminal. This computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected through a system bus. Among them, the processor of this computer device is used to provide computing and control capabilities. The memory of this computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of this computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a carrier network, NFC (Near Field Communication), or other technologies. The display screen of this computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of this computer device may be a touch layer covered on the display screen, or a button, a trackball, or a touchpad provided on the housing of the computer device, or an external keyboard, touchpad, or mouse, etc.

[0092] This embodiment also provides a storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements the data acquisition system proposed in the above embodiment.

[0093] In summary, the present invention realizes the efficient compression of data features by constructing a vehicle state dynamic coding generation module and a road condition feature matrix modeling module, mapping vehicle states and road condition features to a high-dimensional state space, and using the Kronecker product operation of the coding vector and the road condition feature matrix. By introducing a gray scale map driving mechanism and dynamically determining the adjustment amplitude of the sampling frequency based on the change trend of gray scale values, the response ability of the system to emergencies and complex road conditions is further improved. The present invention enhances the timeliness, stability, and accuracy of the data acquisition system, providing more efficient and reliable data support for the intelligent transportation system and the field of autonomous driving.

[0094] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A data acquisition system, characterized in that: include: The vehicle state dynamic code generation module constructs a three-dimensional state space based on the vehicle state and maps the real-time vehicle state into the vehicle state code; The road condition feature matrix modeling module collects road condition features and constructs a road condition feature matrix; The acquisition frequency adjustment module reduces the dimension of the vehicle state encoding vector to a two-dimensional vector, performs a Kronecker product operation with the road condition feature matrix, and generates a strategy matrix; After performing singular value decomposition on the strategy matrix, the first n principal components are retained to form a compression strategy vector, and the acquisition frequency is dynamically confirmed; The grayscale map drives the acquisition module, linearly maps the element values ​​of the strategy matrix to generate a grayscale map, and uses the grayscale value change trend to determine the adjustment range of the acquisition frequency; After obtaining the vehicle state, the vehicle state is classified according to the set value to form a classification index; after constructing the three-dimensional state space according to the classification index, the state vector S = (S SOC ,S speed, S mode ) represents the vehicle state, where S SOC ,S speed, S mode The state vectors correspond to the SOC interval, the vehicle speed classification and the driving mode respectively; the state vector is mapped into a 6-bit binary vehicle state code, wherein the first 2 bits represent the SOC interval, the middle 2 bits represent the vehicle speed classification, and the last 2 bits represent the driving mode; The method of determining the adjustment range of the acquisition frequency by using the gray value change trend includes: generating a grayscale spectrum by using a non-uniform piecewise linear mapping method according to the element values ​​of the strategy matrix; Calculate the regional texture complexity of the grayscale map, use the structure adaptive filter to calculate the grayscale gradient change rate, analyze the local grayscale change pattern, and adjust the acquisition parameters to sampling modes of different dimensions based on the texture direction characteristics of the grayscale map; The grayscale graph driven acquisition mode is enabled only when the element values ​​of the strategy matrix fluctuate dramatically in a short period of time.

2. A data acquisition system as claimed in claim 1, characterized in that: The vehicle status includes the battery SOC value, vehicle speed and driving mode; the road condition characteristics include road slope, bump index and curve curvature radius parameters.

3. A data acquisition system as claimed in claim 1, characterized in that: The vehicle state encoding vector is input into an adaptive projection matrix to generate a two-dimensional vector; wherein the weight of the first row of the projection matrix is ​​dynamically adjusted according to the SOC value.

4. A data acquisition system as claimed in claim 3, characterized in that: The construction of the adaptive projection matrix includes: calculating the influence degree between different state parameters according to each component of the vehicle state coding vector to form a state characteristic gradient matrix; each element of the state characteristic gradient matrix represents the relative change degree between two state parameters; according to the change trend of the state characteristic gradient matrix, extracting the direction with the most significant change as the main projection direction, and selecting the first two main directions to construct the projection matrix.

5. A data acquisition system as claimed in claim 4, characterized in that: The dynamic adjustment according to the SOC value includes: calculating the adjustment amplitude of the vehicle state projection vector when the SOC value changes, and adjusting the main direction of the projection matrix according to the change amplitude of the projection vector.

6. A data acquisition system as claimed in claim 1, characterized in that: The dynamic confirmation of the acquisition frequency includes: calculating the feature mean according to the compression strategy vector, if the feature mean is within the set threshold range, maintaining the current acquisition frequency; if the feature mean is lower than the minimum threshold, reducing the acquisition frequency; if the feature mean is higher than the maximum threshold, increasing the acquisition frequency.

7. A data acquisition method, based on a data acquisition system according to any one of claims 1 to 6, characterized in that: Also includes: Construct a three-dimensional state space based on the vehicle state and map the real-time vehicle state into the vehicle state code; Collect road condition characteristics and construct a road condition characteristic matrix; The vehicle state encoding vector is reduced to a two-dimensional vector, and a Kronecker product operation is performed on the road condition feature matrix to generate a strategy matrix; after performing singular value decomposition on the strategy matrix, the first n principal components are retained to form a compressed strategy vector, and the acquisition frequency is dynamically confirmed; The element values ​​of the strategy matrix are linearly mapped to generate a grayscale spectrum, and the adjustment range of the acquisition frequency is determined using the grayscale value change trend.

Images