Data acquisition method and device and electronic equipment

Abstract

The invention discloses a data collection method and device and electronic equipment, the data collection method can be applied to the technical field of automatic driving, and the method comprises the steps: obtaining original data related to a to-be-executed data collection task; extracting feature information from the original data based on a preset feature dimension; based on the feature information, the contribution degree of the original data in a preset evaluation dimension is quantified, and a trigger priority score of the data acquisition task is obtained; determining a target trigger level matched with the trigger priority score in preset trigger levels; different trigger levels in the preset trigger levels correspond to different resource allocation rules; and based on a resource allocation rule corresponding to the target trigger level, allocating system resources to the data acquisition task. According to the method, a complex driving environment is described through multi-dimensional feature fusion, collaborative optimization classification logic is provided, accurate resource allocation according to classification values is realized, and collaborative improvement of dimension, precision and resource efficiency of automatic driving data acquisition is achieved.

Description

Technical Field

[0001] This application relates to the field of autonomous driving technology, and in particular to data acquisition methods, devices and electronic devices. Background Technology

[0002] In the research and development and iteration of autonomous driving systems, the acquisition of high-quality data is a key foundation for supporting the continuous optimization of perception, decision-making, and control algorithms. Compared with traditional continuous data acquisition methods, triggered acquisition can capture key scenario data in a targeted manner based on preset conditions (such as high-value driving events such as emergency braking and abnormal lane changes). It has shown significant advantages in improving data quality and reducing storage and computing overhead, and has become the mainstream paradigm in the current autonomous driving data acquisition system.

[0003] However, existing technologies still have significant limitations in their methods for classifying trigger conditions. Most current solutions rely on data from a single sensor modality or employ simple combinations of logical rules, resulting in a relatively singular classification dimension that fails to comprehensively depict the multidimensional features of complex and dynamic traffic environments. Furthermore, feature extraction, condition evaluation, and trigger decision-making often lack systematic coordination, with each module operating independently, leading to insufficient overall classification accuracy and robustness. In addition, there is a lack of effective connection between classification results and subsequent resource scheduling strategies, failing to achieve a closed-loop linkage between classification and resource allocation. This can easily lead to wasted computational resources or the omission of key scenario data, hindering further improvements in the iterative efficiency of autonomous driving systems. Summary of the Invention

[0004] To address the problems of existing technologies, embodiments of this application provide a data acquisition method, apparatus, and electronic device. The technical solution is as follows: On the one hand, a data acquisition method is provided, the method comprising: Acquire raw data related to the data acquisition task to be performed; Based on a preset feature dimension, feature information is extracted from the original data; Based on the aforementioned feature information, the contribution of the raw data to the preset evaluation dimension is quantified to obtain the trigger priority score of the data collection task. A target trigger level that matches the trigger priority score is determined from the preset trigger levels; different trigger levels in the preset trigger levels correspond to different resource allocation rules; Based on the resource allocation rules corresponding to the target trigger level, system resources are allocated to the data acquisition task.

[0005] On the other hand, a data acquisition device is provided, the device comprising: The data acquisition module is used to acquire raw data related to the data collection task to be performed; The feature extraction module is used to extract feature information from the original data based on a preset feature dimension; The priority scoring module is used to quantify the contribution of the raw data in a preset evaluation dimension based on the feature information, and obtain the trigger priority score of the data collection task. The level determination module is used to determine a target trigger level that matches the trigger priority score from a preset trigger level; different trigger levels in the preset trigger levels correspond to different resource allocation rules; The resource allocation module is used to allocate system resources to the data acquisition task based on the resource allocation rules corresponding to the target trigger level.

[0006] In one exemplary embodiment, different trigger levels in the preset trigger levels correspond to different scoring threshold ranges, and the scoring threshold ranges are divided based on a first scoring threshold and a second scoring threshold, wherein the first scoring threshold is greater than the second scoring threshold; the level determination module includes: The second level determination module is used to determine the target trigger level as the second trigger level if the trigger priority score is less than the first score threshold and the trigger priority score is greater than or equal to the second score threshold. Accordingly, the resource allocation module includes: The resource monitoring module is used to monitor the real-time resource load of the system when the target trigger level is the second trigger level; The dynamic allocation module is used to allocate available system resources to execute the data acquisition task based on the real-time resource load.

[0007] In one exemplary embodiment, the level determination module further includes: The first level determination module is used to determine the target trigger level as the first trigger level if the trigger priority score is greater than or equal to the first score threshold. Accordingly, the resource allocation module includes: The priority allocation module is used to prioritize the allocation of system resources to execute the data acquisition task when the target trigger level is the first trigger level.

[0008] In one exemplary embodiment, the level determination module further includes: The third-level determination module is used to determine the target trigger level as the third trigger level if the trigger priority score is less than the second score threshold. Accordingly, the resource allocation module includes: The idle resource allocation module is used to allocate idle system resources to perform the data acquisition task when the target trigger level is the third trigger level and the system resources are detected to be idle.

[0009] In one exemplary embodiment, the apparatus further includes a threshold optimization module for optimizing the scoring threshold, the threshold optimization module comprising: The result acquisition module is used to acquire the resource allocation result based on the target trigger level; The threshold adjustment module is used to optimize at least one of the first scoring threshold and the second scoring threshold based on the matching degree between the resource allocation rule corresponding to the target trigger level and the resource allocation result.

[0010] In one exemplary implementation, the priority scoring module includes: The weight acquisition module is used to acquire the weight coefficients corresponding to each evaluation dimension in the preset evaluation dimensions; The contribution quantification module is used to comprehensively quantify the contribution of the original data to each evaluation dimension based on the score corresponding to the feature information under each evaluation dimension and the weight coefficient corresponding to each evaluation dimension. The scoring and integration module is used to integrate the contribution levels corresponding to each of the aforementioned feature dimensions to obtain the trigger priority score for the data collection task.

[0011] In one exemplary embodiment, the preset evaluation dimension includes a data utility dimension; the device further includes a weight adjustment module for adaptively adjusting the weight coefficients corresponding to the data utility dimension, the weight adjustment module comprising: The feedback acquisition module is used to acquire feedback on the utility of the data obtained after performing the data acquisition task in practical applications. The utility weight adjustment module is used to adjust the weight coefficients corresponding to the data utility dimension based on the utility feedback.

[0012] In one exemplary embodiment, the raw data indicates the scene type, and the preset evaluation dimension includes a scene scarcity dimension; the device further includes a scene enhancement module for scarce scene identification and enhancement, the scene enhancement module comprising: The distribution acquisition module is used to acquire the distribution of target trigger levels corresponding to each scenario type in historical data collection tasks; The scarcity identification module is used to identify scarcity scenario types with a collection frequency lower than a preset frequency threshold based on the distribution. The scoring enhancement module is used to improve the scoring of feature information associated with the scarce scene type in the scarce dimension of the scene.

[0013] On the other hand, an electronic device is provided, including a processor and a memory, wherein the memory stores at least one instruction or at least one program, the at least one instruction or the at least one program being loaded and executed by the processor to implement the data acquisition method of any of the above aspects.

[0014] On the other hand, a computer-readable storage medium is provided, wherein at least one instruction or at least one program is stored therein, the at least one instruction or the at least one program being loaded and executed by a processor to implement the data acquisition method as described above.

[0015] On the other hand, a computer program product or computer program is provided, which includes computer instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the electronic device to perform the data acquisition method described above.

[0016] This application's embodiments extract feature information from raw data by pre-setting feature dimensions, breaking the limitations of traditional single-dimensional classification. It comprehensively depicts complex driving environments through multi-dimensional feature fusion, fundamentally solving the problem of single-dimensional classification. Through a coherent technical chain of quantifying contribution levels, generating priority scores, and matching trigger levels, it organically integrates the previously fragmented feature extraction, condition evaluation, and trigger decision-making processes, forming a collaboratively optimized classification logic. This effectively overcomes the accuracy deficiencies caused by fragmented classification steps. By establishing a mechanism that pre-binds trigger levels with resource allocation rules, it achieves dynamic coupling between classification results and resource allocation, enabling the system to accurately allocate computing resources based on classification value. This completely resolves the contradictions of resource waste and data omissions caused by poor resource adaptability. This achieves a synergistic improvement in dimensional coverage, classification accuracy, and resource efficiency in autonomous driving data acquisition. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart illustrating a data acquisition method provided in an embodiment of this application; Figure 2 This is a flowchart illustrating another data acquisition method provided in an embodiment of this application; Figure 3 This is a structural block diagram of a data acquisition device provided in an embodiment of this application; Figure 4 This is a hardware structure block diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0019] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0020] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or server that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0021] It is understood that in the specific embodiments of this application, data such as user information are involved. When the above embodiments of this application are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0022] In existing technologies, trigger-based data acquisition schemes often rely on single sensor data or simple rule combinations, such as using vehicle speed, acceleration, or a single event (e.g., emergency braking) as trigger conditions. The limitations of these methods are becoming increasingly apparent in complex and dynamic autonomous driving scenarios: First, the single classification dimension makes it difficult to comprehensively capture multimodal features in dynamic traffic environments, resulting in insufficient ability to identify edge scenarios or long-tail events; second, feature extraction, condition evaluation, and trigger decision-making are usually designed independently, lacking cross-module collaborative optimization, which easily leads to classification bias or missed triggers; furthermore, classification results are often disconnected from system resource scheduling, failing to dynamically allocate computing, storage, and communication resources based on data value, thus causing resource waste or loss of high-value data. These shortcomings collectively constrain the efficiency and quality of autonomous driving data acquisition, affecting the system iteration speed.

[0023] Therefore, this application provides a data acquisition method. Please refer to [link / reference]. Figure 1 The diagram illustrates a data acquisition method according to an embodiment of this application. It should be noted that while this specification provides method steps as shown in the embodiments or flowcharts, more or fewer steps may be included based on conventional or non-inventive methods. The order of steps listed in the embodiments is merely one possible execution order among many and does not represent the only possible execution order. In actual system or product execution, the methods can be executed sequentially according to the embodiments or accompanying drawings, or in parallel (e.g., in a parallel processor or multi-threaded processing environment). This method acquires raw data related to the data acquisition task through multi-source feature synchronization technology, ensuring the temporal consistency of multi-sensor data. Based on preset feature dimensions, it extracts dynamic features, static features, algorithm features, functional features, and system features from the raw data. Through multi-dimensional feature fusion, it constructs a unified multi-dimensional feature vector and performs normalization processing. Based on the feature information, it quantifies the contribution of the raw data to preset evaluation dimensions (such as safety relevance, scene scarcity, and data utility value) and calculates the trigger priority score through a weighted scoring model. It determines the target trigger level that matches the score from preset trigger levels (such as mandatory trigger, recommended trigger, and optional trigger). Based on the resource allocation rules corresponding to the target trigger level, it realizes the linkage between classification decision and acquisition resources, dynamically allocates system resources for the data acquisition task, and ensures that high-priority tasks are executed first, while low-priority tasks are processed when resources are idle, thereby improving the dimensional coverage, classification accuracy, and resource efficiency of data acquisition.

[0024] For details, please refer to Figure 2 The diagram shown is a flowchart illustrating another data acquisition method provided in an embodiment of this application. Figure 2 As shown, the method may include: S201, Obtain raw data related to the data acquisition task to be executed.

[0025] Specifically, to ensure the temporal consistency and system synergy of the acquired data, multi-sensor synchronous acquisition technology is required to integrate and collect feature data. Specifically, high-precision clock synchronization mechanisms such as PTP (Precision Time Protocol) can be used to strictly align the timestamps of sensors distributed across different nodes during data acquisition, thereby effectively improving the consistency and fusionability of multi-source data in the temporal dimension and providing a reliable foundation for subsequent data processing and analysis.

[0026] S203 extracts feature information from the original data based on preset feature dimensions.

[0027] Among them, the preset feature dimensions refer to the pre-defined multi-angle data classification system used to characterize driving scenarios, such as dynamic, static, and algorithmic feature dimensions.

[0028] Specifically, to comprehensively characterize the system's operational status and environmental interaction characteristics, a feature dataset covering multi-dimensional information needs to be constructed. Specifically, the feature extraction process includes the following aspects: dynamic feature extraction focuses on real-time changing indicators such as the number of traffic participants and predicted movement trajectories; static feature extraction covers stable environmental attributes such as road type, traffic signs, and weather conditions; algorithmic feature extraction focuses on internal algorithmic behaviors such as changes in perception results and the completeness of planned trajectories; functional feature extraction records the execution status of functions such as FCW (Forward Collision Warning) / AEB (Advanced Emergency Braking) trigger status, lane-changing results, and user-initiated takeover; system feature acquisition monitors resource and health parameters including CPU (Central Processing Unit) / GPU (Graphics Processing Unit) utilization, remaining storage space, and system fault status. Through the collaborative extraction of these multi-dimensional features, structured data support can be provided for subsequent analysis and decision-making.

[0029] S205, based on feature information, quantifies the contribution of the raw data to the preset evaluation dimensions to obtain the trigger priority score of the data collection task.

[0030] Among them, the degree of contribution represents a quantitative indicator of the value contribution of the original data to each evaluation dimension (such as security and scarcity).

[0031] Specifically, multidimensional indicators are transformed into comparable priority scores. In practice, multidimensional feature fusion is performed, that is, feature values ​​from different evaluation dimensions are constructed into a unified multidimensional feature vector, which is then normalized to eliminate the influence of dimensions, and input into a predefined classification matrix for initial mapping; according to preset weight coefficients, the scores of each dimension are weighted and combined to calculate the trigger priority score of each data collection task, thereby providing accurate data basis for task scheduling.

[0032] Specifically, the weight coefficients corresponding to each evaluation dimension in the preset evaluation dimensions are obtained; for the feature information corresponding to each feature dimension in the preset feature dimensions, the contribution of the original data to each evaluation dimension is comprehensively quantified based on the score corresponding to the feature information under each evaluation dimension and the weight coefficient corresponding to each evaluation dimension; and the trigger priority score of the data collection task is obtained by comprehensively considering the contribution of each feature dimension.

[0033] The weighting coefficients reflect the relative importance of each evaluation dimension in the decision-making process.

[0034] Among them, comprehensive quantification refers to the calculation process of integrating multi-dimensional scores through mathematical models.

[0035] Specifically, a classification matrix can be constructed based on the extracted multidimensional features, and a weighted scoring model can be used to quantify the trigger priority. For example, refer to formula (1): Priority = α·Safety + β·Rarity + γ·Utility (1) Here, Safety represents safety relevance, Rarity reflects the scarcity of the scenario, Utility represents data utility value, and α, β, and γ are the weight coefficients for the corresponding dimensions. This model transforms subjective evaluation dimensions into calculable numerical indicators, thereby systematically outputting the trigger priority score of data collection tasks.

[0036] In practice, different weighting coefficient combinations can be set for different driving modes (such as highways and urban roads) to achieve scene adaptive evaluation.

[0037] As can be seen from the above technical solutions of the embodiments of this application, by introducing weight coefficients for multiple evaluation dimensions and weighting and synthesizing the contribution of each feature dimension, the embodiments of this application achieve refined and interpretable calculation of trigger priority scoring, clarify the relative importance of different evaluation dimensions (such as security, scarcity, and utility) in decision-making, and make the priority scoring results more accurately reflect the comprehensive value of the data collection task. This enhances the ability to handle complex, multi-objective optimization problems and provides a technical foundation for the system to flexibly adjust collection strategies under different business orientations (such as a greater emphasis on security or a greater emphasis on data diversity).

[0038] Furthermore, the preset evaluation dimensions include a data utility dimension. This involves obtaining utility feedback on the data acquired after performing the data collection task in practical applications; and adjusting the weight coefficients corresponding to the data utility dimension based on this utility feedback.

[0039] Among them, utility feedback refers to the actual value return generated by the collected data in downstream applications (such as model training).

[0040] Specifically, by establishing a classification feedback loop, the system continuously analyzes the collection utility of different data classification results in practical applications, and dynamically adjusts the weight coefficients corresponding to each evaluation dimension accordingly. Specifically, utility feedback includes the actual value indicators of the collected data in downstream applications (such as perception model training and decision algorithm optimization), such as the improvement in model accuracy, the improvement in algorithm recall, or user-defined utility scores. The system periodically (e.g., weekly or monthly) collects historical utility feedback data for each scenario type and calculates a moving average or weighted average. If the utility feedback for a certain scenario type is consistently higher than the historical baseline (e.g., exceeding the baseline by more than 10%), the weight coefficient of the data utility dimension is automatically increased (e.g., increasing γ from 0.2 to 0.25); conversely, if the utility feedback significantly decreases, the weight coefficient is appropriately reduced. The adjustment of weight coefficients can be achieved through exponential smoothing to ensure a smooth transition and avoid drastic fluctuations, ensuring that the system can continuously prioritize the collection of data with high value for autonomous driving iteration, forming a closed-loop optimization from data application to feedback collection decisions.

[0041] In practice, a dynamic weight adjustment strategy can be introduced to enable the weight coefficients to be updated adaptively according to environmental changes and scenario requirements, thereby achieving accurate response and continuous optimization of task classification and priority calculation to actual scenarios.

[0042] For example, statistics show that collecting data on rainy nights significantly improves the perception model. The system automatically increases the weight coefficient of such scenarios in the data utility dimension by 20%, and adjusts the weight coefficients of other dimensions accordingly.

[0043] As can be seen from the above technical solutions of the embodiments of this application, the embodiments of this application add a weight coefficient adjustment mechanism based on actual utility feedback for the data utility dimension. By feeding back the back-end value of data collection (i.e., the application effect of collected data in actual algorithm training or model optimization) to the front-end trigger decision-making stage, a value closed loop of data application feeding back into data collection is formed. Through this feedback-driven mechanism, the system can automatically identify and strengthen the collection tendency of highly useful data characteristics, making data collection activities more focused on scenarios that truly drive the iteration of autonomous driving systems, thereby continuously improving the quality and relevance of the collected datasets and optimizing the long-term return on investment of data collection.

[0044] Furthermore, the raw data indicates the scene type, and the preset evaluation dimensions include the scene scarcity dimension. The distribution of target trigger levels corresponding to each scene type in historical data collection tasks is obtained; based on the distribution, scarce scene types with collection frequencies below a preset frequency threshold are identified; and the scoring of feature information associated with scarce scene types under the scene scarcity dimension is improved.

[0045] In practice, scarce scenario types usually refer to scenario categories that appear less frequently than a preset threshold in historical data collection but have significant value. Data blind spots are filled through proactive discovery and focused collection.

[0046] Specifically, a scene coverage loop is established to continuously analyze the distribution of different scene classification results in historical data collection tasks. This automatically identifies scarce scene types with insufficient coverage and dynamically supplements corresponding classification rules and collection strategies accordingly. This achieves intelligent discovery and autonomous supplementation of scene coverage gaps, ensuring the comprehensiveness and balance of data collection. Specifically, based on the distribution of historical data collection tasks, the system identifies scarce scene types with a collection frequency below a preset frequency threshold (e.g., less than 0.5% in historical data) (such as "emergency vehicle avoidance" or "pedestrian detection in extreme weather"). For these scarce scenes, the system automatically applies scoring enhancement strategies, such as multiplying the baseline score of the feature information associated with the scene in the scene scarcity dimension by an enhancement factor (e.g., 1.5 to 2.0), or directly setting a higher fixed score (e.g., above 0.8). The enhancement factor can be adjusted dynamically according to the degree of scarcity; the higher the scarcity (e.g., frequency below 0.1%), the larger the enhancement factor. At the same time, the system records the collection history of these scenarios to ensure continuous monitoring of their distribution in subsequent evaluations, and optimizes and enhances parameters through feedback mechanisms, thereby proactively filling data blind spots, improving the diversity of the dataset and the system's robustness to edge scenarios.

[0047] For example, the analysis found that the "emergency vehicle avoidance" scenario accounted for less than 0.1% of the historical data. The baseline score of this type of scenario in the scarcity dimension was automatically increased from 0.3 to 0.8, and the weight coefficients of other dimensions were adjusted accordingly.

[0048] As can be seen from the above technical solutions of the embodiments of this application, the embodiments of this application identify scarce scene types with low collection frequency by analyzing the distribution of historical data, and increase the weight of such scene-related features in the scoring. This ensures that the data collection task not only focuses on the current high-value, high-frequency scenes, but also consciously covers those edge scenes that have a low probability of occurrence but are crucial. This effectively promotes the diversity and balance of the dataset, which is of great significance for improving the robustness and safety of autonomous driving systems when facing rare but dangerous scenes.

[0049] S207, determine the target trigger level that matches the trigger priority score in the preset trigger levels; different trigger levels in the preset trigger levels correspond to different resource allocation rules.

[0050] For example, based on preset thresholds, the triggering conditions are divided into three categories: "forced triggering", "recommended triggering" and "optional triggering". Each category corresponds to different resource allocation intensity and execution urgency, thereby achieving refined and differentiated resource management at the task scheduling level.

[0051] S209 allocates system resources for data acquisition tasks based on resource allocation rules corresponding to the target trigger level.

[0052] Specifically, based on the target trigger level, the data collection resources of the system are dynamically allocated. For example, in practice, more than 70% of the core resources can be allocated to "forced trigger" tasks, thereby ensuring the execution guarantee of high-priority collection tasks and the overall optimization of system resource utilization efficiency at the resource level.

[0053] Specifically, different trigger levels in the preset trigger levels correspond to different scoring threshold ranges. These ranges are divided based on a first scoring threshold and a second scoring threshold, with the first threshold being greater than the second. If the trigger priority score is greater than or equal to the first scoring threshold, the target trigger level is determined to be the first trigger level. When the target trigger level is the first trigger level, system resources are prioritized for executing the data acquisition task. If the trigger priority score is less than the first scoring threshold but greater than or equal to the second scoring threshold, the target trigger level is determined to be the second trigger level. When the target trigger level is the second trigger level, the real-time resource load of the system is monitored. Based on the real-time resource load, available system resources are allocated to execute the data acquisition task. If the trigger priority score is less than the second scoring threshold, the target trigger level is determined to be the third trigger level. When the target trigger level is the third trigger level, if system resources are detected to be idle, the idle system resources are allocated to execute the data acquisition task.

[0054] Among them, real-time resource load reflects the dynamic occupancy status of the system's computing and storage resources.

[0055] For example, the trigger levels are divided into three categories based on the trigger priority score: forced trigger (trigger priority score ≥ 0.8), where data acquisition tasks at this trigger level will be immediately allocated the best system resources to perform acquisition; recommended trigger (0.5 ≤ trigger priority score < 0.8), where resources are allocated as needed to perform acquisition if system resources allow; and optional trigger (trigger priority score < 0.5), where acquisition is only performed when the system detects that resources are idle.

[0056] In practice, multiple scoring thresholds can be used to more finely divide the priority scoring range, thereby constructing a multi-level triggering system to adapt to the needs of detailed management of task priorities and resource matching in complex scenarios.

[0057] As can be seen from the above technical solutions of the embodiments of this application, the embodiments of this application, through a refined hierarchical classification system and a highly adaptable resource allocation strategy designed for each level, realize the dual consideration and dynamic coupling of task priority and system resource status. For high-priority tasks (such as mandatory triggering tasks), the system unconditionally prioritizes resource protection, thereby significantly reducing the decision delay of task scheduling and ensuring that critical acquisition tasks can be responded to and executed in a timely manner. For medium-priority tasks (such as recommended triggering tasks), real-time resource load monitoring is introduced to realize on-demand and elastic allocation of resources, balancing task value and system load. For low-priority tasks (such as optional triggering tasks), idle system resources are used for execution, maximizing resource utilization. While ensuring the critical functions of the system, the overall resource utilization efficiency and system economy are greatly improved.

[0058] Furthermore, obtain the resource allocation result based on the target trigger level; based on the matching degree between the resource allocation rule corresponding to the target trigger level and the resource allocation result, optimize at least one of the first scoring threshold and the second scoring threshold.

[0059] Among them, the matching degree characterizes the degree to which the resource allocation result conforms to the expected allocation rule.

[0060] Among them, optimization refers to the iterative adjustment process of threshold parameters based on feedback data.

[0061] Specifically, a resource adaptation loop is established to monitor the matching degree between classification decisions and resource allocation results in real time. Based on this, the scoring threshold is dynamically evaluated and optimized to continuously improve the accuracy of trigger level classification and resource scheduling efficiency, enabling the system to self-optimize and adapt to different scenarios in long-term operation. Specifically, the matching degree can be quantified by calculating the conformity rate between the resource allocation results and the expected allocation rules. For example, the consistency ratio between the actual resource allocation and the resource allocation required by the target trigger level can be statistically analyzed (e.g., whether high-priority tasks actually receive priority resource allocation in actual execution). If the matching degree is lower than a preset threshold (e.g., 90%), the system will automatically initiate a threshold optimization algorithm, such as using gradient descent or heuristic search, to adjust at least one of the first and second scoring thresholds. For example, when the system detects that "forced trigger" tasks are frequently delayed due to insufficient resources, it may appropriately increase the first scoring threshold to narrow the scope of forced triggers and ensure that resources are concentrated on the highest-value tasks; conversely, if "recommended trigger" tasks are over-allocated resources, the second scoring threshold may be lowered to expand the scope of recommended triggers. This iterative optimization mechanism based on real-time feedback enables the scoring threshold to adapt to changes in system operating status and scenario, thereby improving classification accuracy and resource efficiency.

[0062] As can be seen from the above technical solutions of the embodiments of this application, the embodiments of this application dynamically adjust the scoring threshold by obtaining resource allocation results and performing matching degree analysis with preset rules, thus forming a self-evolving and continuously optimized closed-loop control system. This enables the entire classification and resource scheduling system to have adaptive and learning capabilities, and can continuously calibrate the correspondence between the scoring threshold and the trigger level based on actual operational feedback, thereby overcoming classification bias caused by scene changes, data distribution shifts, or improper initial threshold settings. This significantly improves the robustness and scene adaptability of the system in long-term deployment, ensuring that the data acquisition strategy can maintain optimal performance with technological iterations and environmental evolution.

[0063] To facilitate a full understanding of the solution proposed in this application, the data acquisition triggering process of this method will be illustrated below with an example of a typical autonomous driving scenario.

[0064] Suppose an autonomous vehicle encounters the following scenario while driving on an urban road: a pedestrian suddenly crosses the road ahead (dynamic feature); the road type is an urban road, the weather is clear and visibility is good (static feature); the perception system detects the pedestrian stably, but trajectory prediction is uncertain (algorithm feature); the AEB system is ready but has not yet been triggered (functional feature); the system's computational resource load is moderate (CPU utilization 60%), and storage space is sufficient (system feature). Based on the above multi-dimensional features, the classification matrix shown in Table 1 is constructed and quantitatively scored.

[0065] Table 1: Feature Dimension Scoring Matrix

[0066] With weighting coefficients set at α=0.5 (safety), β=0.3 (scarcity), and γ=0.2 (utility), the trigger priority scores for each dimension are calculated sequentially: Dynamic characteristics: 0.5×0.9 + 0.3×0.7 + 0.2×0.8 = 0.45 + 0.21 + 0.16 = 0.82 Static characteristic: 0.5×0.5 + 0.3×0.3 + 0.2×0.6 = 0.25+0.09+0.12 = 0.46 Algorithm characteristics: 0.5×0.7 + 0.3×0.5 + 0.2×0.7 = 0.35 + 0.15 + 0.14 = 0.64 Functional characteristics: 0.5×0.8 + 0.3×0.6 + 0.2×0.9 = 0.40+0.18+0.18 = 0.76 System characteristics: 0.5×0.4 + 0.3×0.4 + 0.2×0.5 = 0.20 + 0.12 + 0.10 = 0.42 The overall trigger priority score is obtained by combining the scores from various dimensions: Priority = (0.82+0.46+0.64+0.76+0.42) / 5 = 3.1 / 5 = 0.62 Based on the preset scoring threshold, this score belongs to the "recommended trigger" level (0.5 ≤ priority < 0.8). The system will allocate available resources as needed to execute this data collection task based on the real-time resource load.

[0067] As can be seen from the above technical solutions of the embodiments of this application, the embodiments of this application enable data collection tasks to be accurately classified according to their comprehensive contribution in the preset evaluation dimensions through multi-dimensional feature fusion and dynamic priority evaluation. Based on the classification results, differentiated resource allocation rules are automatically matched, which effectively reduces the overall delay of task classification and decision-making, realizes the direct mapping between trigger level and resource allocation strategy, and enables the autonomous driving system to take into account both data value and resource efficiency in complex scenarios. It realizes the fundamental transformation of trigger condition judgment from rule-driven to value-driven, and significantly improves the intelligence level of data collection and the overall system efficiency.

[0068] Corresponding to the data acquisition methods provided in the above embodiments, this application also provides a data acquisition device. Since the data acquisition device provided in this application corresponds to the data acquisition methods provided in the above embodiments, the implementation methods of the aforementioned data acquisition methods are also applicable to the data acquisition device provided in this embodiment, and will not be described in detail in this embodiment.

[0069] Please see Figure 3 The diagram shown is a structural schematic of a data acquisition device provided in an embodiment of this application. This device has the function of implementing the data acquisition method described in the above-described method embodiments. This function can be implemented by hardware or by hardware executing corresponding software. Figure 3 As shown, the device may include: Data acquisition module 310 is used to acquire raw data related to the data acquisition task to be performed; The feature extraction module 320 is used to extract feature information from the original data based on a preset feature dimension; The priority scoring module 330 is used to quantify the contribution of the raw data to the preset evaluation dimension based on feature information, and obtain the trigger priority score of the data collection task. The level determination module 340 is used to determine the target trigger level that matches the trigger priority score from the preset trigger levels; different trigger levels in the preset trigger levels correspond to different resource allocation rules; The resource allocation module 350 is used to allocate system resources to data acquisition tasks based on resource allocation rules corresponding to the target trigger level.

[0070] In one exemplary embodiment, different trigger levels in the preset trigger levels correspond to different scoring threshold ranges, and the scoring threshold ranges are divided based on a first scoring threshold and a second scoring threshold, wherein the first scoring threshold is greater than the second scoring threshold; the level determination module includes: The second-level determination module is used to determine the target trigger level as the second trigger level if the trigger priority score is less than the first score threshold and the trigger priority score is greater than or equal to the second score threshold. Correspondingly, the resource allocation module includes: The resource monitoring module is used to monitor the real-time resource load of the system when the target trigger level is the second trigger level. The dynamic allocation module is used to allocate available system resources to perform data acquisition tasks based on real-time resource load.

[0071] In one exemplary embodiment, the level determination module further includes: The first-level determination module is used to determine the target trigger level as the first trigger level if the trigger priority score is greater than or equal to the first score threshold. Correspondingly, the resource allocation module includes: The priority allocation module is used to prioritize the allocation of system resources to execute data acquisition tasks when the target trigger level is the first trigger level.

[0072] In one exemplary embodiment, the level determination module further includes: The third-level determination module is used to determine the target trigger level as the third trigger level if the trigger priority score is less than the second score threshold. Correspondingly, the resource allocation module includes: The idle resource allocation module is used to allocate idle system resources to perform data acquisition tasks when the target trigger level is the third trigger level and the system resources are detected to be idle.

[0073] In one exemplary embodiment, the apparatus further includes a threshold optimization module for optimizing the scoring threshold, the threshold optimization module comprising: The result acquisition module is used to obtain resource allocation results based on the target trigger level; The threshold adjustment module is used to optimize at least one of the first scoring threshold and the second scoring threshold based on the matching degree between the resource allocation rules and resource allocation results corresponding to the target trigger level.

[0074] In one exemplary implementation, the priority scoring module includes: The weight acquisition module is used to obtain the weight coefficients corresponding to each of the preset evaluation dimensions. The contribution quantification module is used to comprehensively quantify the contribution of the original data to each evaluation dimension based on the score corresponding to the feature information of each feature dimension in the preset feature dimensions and the weight coefficient corresponding to each evaluation dimension. The scoring module is used to combine the contribution levels of each feature dimension to obtain the trigger priority score of the data collection task.

[0075] In one exemplary embodiment, the preset evaluation dimension includes a data utility dimension; the apparatus further includes a weight adjustment module for adaptively adjusting the weight coefficients corresponding to the data utility dimension, the weight adjustment module including: The feedback acquisition module is used to obtain feedback on the utility of the data obtained after performing the data acquisition task in practical applications; The utility weight adjustment module is used to adjust the weight coefficients corresponding to the data utility dimension based on utility feedback.

[0076] In one exemplary embodiment, the raw data indicates the scene type, and the preset evaluation dimensions include a scene scarcity dimension; the device further includes a scene enhancement module for scarce scene identification and enhancement, the scene enhancement module including: The distribution acquisition module is used to acquire the distribution of target trigger levels corresponding to each scenario type in historical data collection tasks; The scarcity identification module is used to identify scarcity scenario types with a collection frequency lower than a preset frequency threshold based on the distribution. The scoring enhancement module is used to improve the scoring of feature information associated with scarce scene types in the dimension of scene scarcity.

[0077] It should be noted that the apparatus provided in the above embodiments is only illustrated by the division of the above functional modules when implementing its functions. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the apparatus and method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process can be found in the method embodiments, which will not be repeated here.

[0078] This application provides an electronic device including a processor and a memory. The memory stores at least one instruction or at least one program segment, which is loaded and executed by the processor to implement any of the data acquisition methods provided in the above method embodiments.

[0079] Memory is used to store software programs and modules. The processor executes these stored software programs and modules to perform various functional applications and data processing. Memory can primarily consist of a program storage area and a data storage area. The program storage area stores the operating system, application programs required for functionality, etc.; the data storage area stores data created based on device usage, etc. Furthermore, memory can include high-speed random access memory (RAM) and non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device. Accordingly, memory can also include a memory controller to provide the processor with access to the memory.

[0080] The method embodiments provided in this application can be executed in a computer terminal, server or similar computing device, that is, the above-mentioned electronic device may include a computer terminal, server or similar computing device. Figure 4 This is a hardware structure block diagram of a computer device for running a data acquisition method provided in an embodiment of the present invention, such as... Figure 4As shown, the internal structure of this computer device may include, but is not limited to, a processor, a network interface, and a memory. The processor, network interface, and memory within the computer device can be connected via a bus or other means, as illustrated in the embodiments of this specification. Figure 4 Taking the example of a connection between China and Israel via a bus.

[0081] The processor (or CPU, Central Processing Unit) is the computing and control core of the computer device. The network interface may optionally include a standard wired interface or a wireless interface (such as Wi-Fi, mobile communication interface, etc.). Memory is the storage device in the computer device used to store programs and data. It is understood that the memory here can be a high-speed RAM storage device, or a non-volatile storage device, such as at least one disk storage device; optionally, it can also be at least one storage device located remotely from the aforementioned processor. The memory provides storage space, which stores the operating system of the electronic device, including but not limited to: Windows (an operating system), Linux (an operating system), Android (a mobile operating system), iOS (a mobile operating system), etc., which are not limited in this invention; and the storage space also stores one or more instructions suitable for being loaded and executed by the processor, which can be one or more computer programs (including program code). In the embodiments of this specification, the processor loads and executes one or more instructions stored in the memory to implement the data acquisition method provided in the above method embodiments.

[0082] Embodiments of this application also provide a computer-readable storage medium that can be disposed in an electronic device to store at least one instruction or at least one program related to implementing a data acquisition method. The at least one instruction or the at least one program is loaded and executed by the processor to implement any of the data acquisition methods provided in the above-described method embodiments.

[0083] Optionally, in this embodiment, the storage medium may include, but is not limited to, various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0084] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, specific embodiments have been described above. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps described in the claims can be performed in a different order than that shown in the embodiments and still achieve the desired result. Additionally, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0085] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the apparatus embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0086] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware or by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk.

[0087] The above are merely preferred embodiments of this application and are not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A data acquisition method, characterized in that, The method includes: Acquire raw data related to the data acquisition task to be performed; Based on a preset feature dimension, feature information is extracted from the original data; Based on the aforementioned feature information, the contribution of the raw data to the preset evaluation dimension is quantified to obtain the trigger priority score of the data collection task. A target trigger level that matches the trigger priority score is determined from the preset trigger levels; different trigger levels in the preset trigger levels correspond to different resource allocation rules; Based on the resource allocation rules corresponding to the target trigger level, system resources are allocated to the data acquisition task.

2. The data acquisition method according to claim 1, characterized in that, Different trigger levels in the preset trigger levels correspond to different scoring threshold ranges. The scoring threshold ranges are divided based on a first scoring threshold and a second scoring threshold, where the first scoring threshold is greater than the second scoring threshold. The step of determining a target trigger level that matches the trigger priority score from a preset trigger level includes: If the trigger priority score is less than the first score threshold and the trigger priority score is greater than or equal to the second score threshold, then the target trigger level is determined to be the second trigger level. Accordingly, allocating system resources for the data acquisition task based on the resource allocation rules corresponding to the target trigger level includes: When the target trigger level is the second trigger level, monitor the real-time resource load of the monitoring system; Based on the real-time resource load, available system resources are allocated to execute the data acquisition task.

3. The data acquisition method according to claim 2, characterized in that, The step of determining a target trigger level that matches the trigger priority score from the preset trigger levels further includes: If the trigger priority score is greater than or equal to the first score threshold, then the target trigger level is determined to be the first trigger level; Accordingly, allocating system resources for the data acquisition task based on the resource allocation rules corresponding to the target trigger level includes: When the target trigger level is the first trigger level, system resources are allocated first to execute the data acquisition task.

4. The data acquisition method according to claim 2, characterized in that, The step of determining a target trigger level that matches the trigger priority score from the preset trigger levels further includes: If the trigger priority score is less than the second score threshold, then the target trigger level is determined to be the third trigger level; Accordingly, allocating system resources for the data acquisition task based on the resource allocation rules corresponding to the target trigger level includes: When the target trigger level is the third trigger level, and system resources are detected to be in an idle state, the idle system resources are allocated to execute the data acquisition task.

5. The data acquisition method according to any one of claims 2 to 4, characterized in that, The method further includes: Obtain the resource allocation result based on the target trigger level; Based on the matching degree between the resource allocation rule corresponding to the target trigger level and the resource allocation result, optimize at least one of the first scoring threshold and the second scoring threshold.

6. The data acquisition method according to any one of claims 1 to 4, characterized in that, The step of quantifying the contribution of the raw data in a preset evaluation dimension based on the feature information to obtain the trigger priority score of the data collection task includes: Obtain the weight coefficients corresponding to each evaluation dimension in the preset evaluation dimensions; For the feature information corresponding to each feature dimension in the preset feature dimensions, based on the score corresponding to the feature information under each evaluation dimension and the weight coefficient corresponding to each evaluation dimension, the contribution of the original data to each evaluation dimension is comprehensively quantified. By combining the contribution levels of each of the aforementioned feature dimensions, the trigger priority score of the data acquisition task is obtained.

7. The data acquisition method according to claim 6, characterized in that, The preset evaluation dimensions include a data utility dimension; the method further includes: Obtain feedback on the utility of the data acquired after performing the data acquisition task in practical applications; Based on the utility feedback, the weight coefficients corresponding to the data utility dimension are adjusted.

8. The data acquisition method according to claim 6, characterized in that, The raw data indicates the scene type, and the preset evaluation dimensions include scene scarcity dimensions; the method further includes: In historical data collection tasks, the distribution of target trigger levels for each scenario type is obtained; Based on the distribution, identify scarce scene types with a collection frequency lower than a preset frequency threshold; Improve the score of feature information associated with the scarce scene type in the scarce dimension of the scene.

9. A data acquisition device, characterized in that, The device includes: The data acquisition module is used to acquire raw data related to the data collection task to be performed; The feature extraction module is used to extract feature information from the original data based on a preset feature dimension; The priority scoring module is used to quantify the contribution of the raw data in a preset evaluation dimension based on the feature information, and obtain the trigger priority score of the data collection task. The level determination module is used to determine a target trigger level that matches the trigger priority score from a preset trigger level; different trigger levels in the preset trigger levels correspond to different resource allocation rules; The resource allocation module is used to allocate system resources to the data acquisition task based on the resource allocation rules corresponding to the target trigger level.

10. An electronic device, characterized in that, It includes a processor and a memory, wherein the memory stores at least one instruction or at least one program, the at least one instruction or the at least one program being loaded and executed by the processor to implement the data acquisition method as described in any one of claims 1 to 8.

Images