Data correction method, electronic device, and storage medium

Abstract

This application provides a data correction method, electronic device, and storage medium, relating to the field of artificial intelligence technology. The method includes: acquiring first sensor data in multiple dimensions, determining multi-dimensional fused sensor data based on the data, calculating a first multi-dimensional difference parameter using the multi-dimensional fused sensor data, determining a comprehensive confidence level of the multi-dimensional fused sensor data by combining data quality and parameter estimation confidence, further determining a correction strategy for the sensor data based on the comprehensive confidence level, correcting the first sensor data according to the correction strategy and the first multi-dimensional difference parameter to obtain second sensor data, calculating a second multi-dimensional difference parameter, and determining the second sensor data as the target sensor data if the difference parameter is less than or equal to a preset threshold. This application aims to improve the problems of poor real-time performance and poor adaptive capability in existing solutions.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a data correction method, electronic device, and storage medium. Background Technology

[0002] With the continuous development of artificial intelligence technology, intelligent execution terminals are increasingly widely used in various industries. In the research and development and deployment of intelligent execution terminals, in order to improve the system's intelligence level and task execution capabilities, it is usually necessary to conduct extensive training and verification of its perception algorithms, decision-making models, and control strategies. Through simulation-to-real (sim-to-real) technology, the differences between the simulation environment and the real environment can be reduced or eliminated, enabling models, parameters, or strategies trained, optimized, or verified in the simulation environment to operate stably and effectively in the real environment.

[0003] However, existing sim-to-real schemes have poor real-time performance and poor adaptability during data correction. Summary of the Invention

[0004] The embodiments disclosed in this application provide a data correction method, electronic device, and storage medium, which can improve the problems of poor real-time performance and poor adaptability of existing sim-to-real solutions during data correction.

[0005] The embodiments of this application adopt the following technical solutions: Firstly, a data correction method is provided, comprising: acquiring first sensor data in multiple dimensions; determining multidimensional fused sensor data based on the first sensor data in multiple dimensions; determining a first multidimensional difference parameter based on the multidimensional fused sensor data, the first multidimensional difference parameter being used to characterize the degree of difference between sensor data in real and simulated environments; determining a comprehensive confidence level of the multidimensional fused sensor data based on the data quality and parameter estimation reliability of the multidimensional fused sensor data; determining a sensor data correction strategy based on the comprehensive confidence level of the multidimensional fused sensor data; correcting the first sensor data based on the sensor data correction strategy and the first multidimensional difference parameter to obtain second sensor data; determining a second multidimensional difference parameter based on the second sensor data in multiple dimensions; and determining the second sensor data as target sensor data if the second multidimensional difference parameter is less than or equal to a multidimensional difference parameter threshold.

[0006] The data correction method provided in this application, which integrates multidimensional sensor data acquisition and processing in real time, can calculate multidimensional difference parameters based on the first sensor data and dynamically evaluate the comprehensive confidence level. This allows the correction process to be executed immediately after data acquisition, without relying on offline training or batch processing, thereby improving the real-time performance of data correction. Therefore, it can achieve millisecond-level real-time response, adapting to rapid changes in dynamic environments. Furthermore, by introducing the evaluation of comprehensive confidence level and multidimensional difference parameters, this application can dynamically select appropriate correction strategies, flexibly responding to fluctuations in data quality and environmental changes, thus obtaining accurate correction results in different environments and scenarios, improving the adaptability of data correction.

[0007] In one possible implementation of the first aspect, the first sensor data includes at least one of the following: sensor data in a simulated environment or sensor data in a real environment.

[0008] The data calibration method provided in this application can effectively address the differences between various types of devices and environments by combining sensor data from both simulated and real-world environments. Simulated environment data can provide initial training and validation of the calibration algorithm, while real-world environment data, especially data transferred from other devices, can provide more diverse and complex application scenarios. By transferring and fusing data between devices, the same calibration strategy can be shared across multiple devices, further improving the universality and accuracy of the calibration.

[0009] In one possible implementation of the first aspect, determining multidimensional fused sensor data based on multiple first sensor data of different dimensions includes performing data preprocessing operations on the multiple first sensor data of different dimensions to obtain first preprocessed sensor data, performing a synchronization alignment operation on the first preprocessed sensor data to obtain second preprocessed sensor data, and performing a multidimensional fusion operation on the second preprocessed sensor data to obtain multidimensional fused sensor data.

[0010] In one possible implementation of the first aspect, determining a first multidimensional difference parameter based on multidimensional fused sensor data includes: determining the data deviation of the first sensor data in each dimension based on the calculation result of the difference between the first sensor data in each dimension and the calibration value; determining the deviation fusion weight of the first sensor data in each dimension based on the information reliability of the first sensor data in each dimension; and determining the first multidimensional difference parameter based on the multiplication and summation result of the data deviations of the first sensor data in multiple dimensions and the corresponding deviation fusion weights.

[0011] The data correction method provided in this application effectively improves the correction accuracy for multi-source data by weighting and fusing deviations from multiple dimensions based on the differences between multi-dimensional fused sensor data and calibration values, combined with the reliability of sensor data in each dimension. The multiplication and summation of the first sensor data deviation in each dimension and the fusion weights comprehensively reflects the differences between multi-dimensional sensor data and real-world environmental characteristics, further optimizing the data correction process. By introducing a deviation fusion weighting mechanism, the method in this application can adaptively adjust the weights of the first sensor data in each dimension, avoiding excessive influence of errors in a single dimension on the overall correction result.

[0012] In one possible implementation of the first aspect, a sensor data correction strategy is determined based on the overall confidence level of the multidimensional fused sensor data, including: determining a first sensor data correction strategy when the overall confidence level is greater than a first overall confidence level threshold; determining a second sensor data correction strategy when the overall confidence level is greater than a second overall confidence level threshold and less than or equal to the first overall confidence level threshold; and determining a third sensor data correction strategy when the overall confidence level is less than or equal to the second overall confidence level threshold, wherein the first overall confidence level threshold is greater than the second overall confidence level threshold.

[0013] The data correction method provided in this application allows for the selection of corresponding data correction strategies based on different comprehensive confidence levels. This ensures that the most suitable correction method is chosen under different data quality conditions, thereby guaranteeing the reliability and convergence efficiency of sensor data correction.

[0014] In one possible implementation of the first aspect, a sensor data correction strategy for correcting first sensor data to obtain second sensor data includes: when the sensor data correction strategy is the first sensor data correction strategy, determining the correction frequency and correction order of the physical parameters in the first sensor data based on the matching relationship between the physical parameters in the first sensor data and preset data information; correcting the physical parameters in the first sensor data using a first data adjustment step size based on the correction frequency and correction order to obtain second sensor data; the data information includes physical parameters, the physical parameters and their corresponding correction frequencies, and the correction order corresponding to the physical parameters; the physical parameters and their corresponding correction frequencies and correction order are determined based on the degree of influence and identification difficulty of the physical parameters.

[0015] In one possible implementation of the first aspect, correcting first sensor data according to a sensor data correction strategy and a first multidimensional difference parameter to obtain second sensor data includes: when the sensor data correction strategy is a second sensor data correction strategy, determining the correction frequency and correction order of the physical parameters in the first sensor data according to the matching relationship between the physical parameters in the first sensor data and preset data information, and correcting the physical parameters in the first sensor data using a second data adjustment step size according to the correction frequency and correction order to obtain second sensor data; wherein the first data adjustment step size is greater than the second data adjustment step size.

[0016] In one possible implementation of the first aspect, the first sensor data is corrected according to the sensor data correction strategy and the first multidimensional difference parameter to obtain the second sensor data, including: when the sensor data correction strategy is a third sensor data correction strategy, the physical parameters in the first sensor data are corrected with the optimization objectives of maximizing information gain and minimizing risk to obtain the third sensor data, and the step of determining multidimensional fused sensor data based on the third sensor data is continued according to the third sensor data.

[0017] In one possible implementation of the first aspect, the method further includes: if the second multidimensional difference parameter is greater than the multidimensional difference parameter threshold, continuing to perform the step of determining the overall confidence level of the multidimensional fused sensor data.

[0018] Secondly, a data correction device is provided, the device comprising: The acquisition module is used to acquire first sensor data in multiple dimensions and determine multi-dimensional fused sensor data based on the first sensor data in multiple dimensions. The multidimensional difference parameter determination module is used to determine the first multidimensional difference parameter based on multidimensional fused sensor data. The comprehensive confidence level determination module is used to determine the comprehensive confidence level of the multi-dimensional fused sensor data based on the data quality of the multi-dimensional fused sensor data and the reliability of the parameter estimation of the multi-dimensional fused sensor data. The calibration strategy determination module is used to determine the sensor data calibration strategy based on the comprehensive confidence level of the multi-dimensional fused sensor data. The calibration module is used to calibrate the first sensor data in multiple dimensions according to the sensor data calibration strategy and the first multidimensional difference parameter, so as to obtain the second sensor data in multiple dimensions. The judgment module is used to determine the second multidimensional difference parameter based on the second sensor data in multiple dimensions. If the second multidimensional difference parameter is less than or equal to the multidimensional difference parameter threshold, the second sensor data is determined as the target sensor data. The multidimensional difference parameter is used to characterize the degree of difference between the sensor data in the real environment and the simulation environment.

[0019] Thirdly, an electronic device is provided, comprising: a memory and at least one processor. The memory is communicatively connected to the processor. The memory is used to store computer program code, which includes computer instructions. When the processor executes the computer instructions, it causes the electronic device to perform a method as described in the first aspect and any possible implementation thereof.

[0020] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer instructions. When these computer instructions are executed by a processor, they are used to implement the method as described in the first aspect and any possible implementation thereof.

[0021] Fifthly, embodiments of this application provide a computer program product that, when running on a computer / executed by the computer's processor, implements the method described in the first aspect and any possible design thereof. The computer may be the data correction device described in the second aspect and any possible implementation thereof.

[0022] Understandably, the technical effects of the second to fifth aspects refer to the technical effects of the first aspect and any of its embodiments, and will not be repeated here. Attached Figure Description

[0023] Figure 1 A flowchart illustrating the steps of a data correction method provided in this application embodiment; Figure 2 A second flowchart illustrating the steps of a data correction method provided in this application embodiment; Figure 3 A flowchart of the steps of a data correction method provided in this application embodiment is shown in the third part. Figure 4 A flowchart of the steps of a data correction method provided in this application embodiment is shown in Figure 4. Figure 5 A block diagram of a data correction device provided in an embodiment of this application; Figure 6 This is a schematic diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0024] The technical solutions in some embodiments of this application will be clearly and completely described below with reference to the figures. Obviously, the embodiments described in the specification are only some embodiments of this application, and not all embodiments. Based on the embodiments provided in this application, all other embodiments obtained by those skilled in the art are within the scope of protection of this application.

[0025] Unless the context otherwise requires, throughout the specification and claims, the term "comprising" is interpreted as open-ended and encompassing, meaning "including, but not limited to." In the description of the specification, terms such as "one embodiment," "some embodiments," "exemplary embodiment," "exemplary," or "some examples," etc., are intended to indicate that a particular parameter, structure, material, or characteristic associated with that embodiment or example is included in at least one embodiment or example of this application. The illustrative representations of the above terms do not necessarily refer to the same embodiment or example.

[0026] The use of “configured as” in this article implies an open and inclusive language that does not exclude the applicability to or configuration of devices to perform additional tasks or steps.

[0027] With the continuous development of artificial intelligence technology, intelligent execution terminals are increasingly widely used in various industries. An intelligent execution terminal can refer to a terminal device that integrates sensing, computing, decision-making, and execution capabilities, enabling it to acquire and analyze external environmental information under the control of preset rules or intelligent algorithms, and autonomously or semi-autonomously complete corresponding operational tasks accordingly. Intelligent execution terminals can be various industrial robots, service robots, or intelligent equipment, automated devices, or embedded execution devices with automatic control and execution capabilities.

[0028] In the research and development and deployment of intelligent execution terminals, in order to improve the system's intelligence level and task execution capabilities, it is usually necessary to conduct extensive training and verification of its perception algorithms, decision-making models, and control strategies. However, directly training and debugging intelligent execution terminals in real-world environments often faces problems such as high cost, long cycle, significant safety risks, and difficulty in reproducing experimental conditions. For example, in industrial production or complex operation scenarios, frequent real-world testing can not only disrupt normal production order but may also lead to equipment damage or safety accidents due to immature strategies. Therefore, using simulation environments to model, train, and verify intelligent execution terminals has gradually become a widely adopted technical approach in the industry.

[0029] Simulation environments typically provide a low-cost, highly controllable, and highly repeatable way to model physical processes, interactions, and sensor feedback in real-world environments. They allow for the rapid construction of diverse scenarios and large-scale training and testing of intelligent actuators, significantly improving algorithm development efficiency. However, because simulation models inevitably differ from real-world models in terms of physical parameters, sensor characteristics, and environmental complexity, directly deploying models or control strategies trained in simulation environments to real-world environments often leads to performance degradation or even failure.

[0030] In one feasible implementation, sim-to-real technology can reduce or eliminate the differences between the simulation environment and the real environment, enabling models, parameters, or policies trained, optimized, or validated in the simulation environment to run stably and effectively in the real environment. The core objective of sim-to-real is to improve the transferability and reliability of simulation results in real-world scenarios, thereby fully leveraging the advantages of simulation training in cost control, efficiency improvement, and security assurance.

[0031] While existing simulation-to-real techniques have alleviated the discrepancy between simulation and reality to some extent, significant shortcomings remain. First, most existing solutions focus on single-modal or limited sensor data, failing to fully utilize the complementary information between multi-source heterogeneous sensors, resulting in an incomplete characterization of environmental differences. Second, existing methods generally lack quantitative evaluation mechanisms for sensor data quality and parameter estimation reliability, easily introducing erroneous correction results and affecting system stability when data noise is high or observational information is insufficient. Third, many simulation-to-real methods rely on offline modeling or static parameter correction, making it difficult to dynamically adjust according to environmental changes during operation, leading to insufficient adaptability under long-term operation or complex conditions. Finally, existing technologies are largely based on passive observation; when environmental stimuli are insufficient, it is difficult to effectively obtain information discriminative for parameter correction, thus limiting further improvements in simulation-to-real performance.

[0032] To address the problems in related technologies, this application provides a data correction method. By fusing sensor data from multiple dimensions and combining difference parameters with comprehensive confidence levels, a sensor data correction strategy is adaptively determined, thereby achieving refined correction of sensor data. The data correction method provided in this application can be applied to application scenarios requiring data migration or model generalization between simulation and real environments, such as industrial robots, service robots, and autonomous driving. It is understood that the executing entity of this method can be an intelligent execution terminal such as a robot. This intelligent execution terminal integrates sensors, a processor, and a memory, and executes program instructions stored in the memory through the processor to implement the data correction method; alternatively, the executing entity can also be an edge computing device, a cloud server, or a collaborative system composed of these connected to the intelligent execution terminal.

[0033] Figure 1 This is a flowchart illustrating a data correction method provided in an embodiment of this application. For example, please refer to... Figure 1 As shown, the data correction method may include S101 to S105: S101: Acquire first sensor data in multiple dimensions, and determine multi-dimensional fused sensor data based on the first sensor data in multiple dimensions.

[0034] First-sensor data of different dimensions refers to sensor data from different modalities, that is, data collected by sensors with different sensing mechanisms and information expression methods, used to jointly perceive target objects, environmental states, or execution processes from different physical modalities. For example, first-sensor data of different dimensions can include data collected by visual sensors, force sensors, and inertial modal sensors. The visual sensor can be an RGB-D camera, which can simultaneously acquire color and depth information, thereby providing visual positioning, structural, or pose information of the target object in space, providing basic data support for subsequent spatial modeling and position estimation. The force sensor can be a six-dimensional force sensor, which can sense the contact force and corresponding torque changes between the target object and the actuator, used to reflect the force state, contact stability, or abnormal situations during the interaction process. The inertial modal sensor can be an IMU sensor, which can monitor the acceleration, angular velocity, or attitude changes of the actuator during motion, used to characterize changes in motion state and dynamic behavior characteristics.

[0035] In one feasible implementation, the first sensor data may include at least one of the following: sensor data in a simulated environment or sensor data in a real environment.

[0036] Sensor data in a simulation environment refers to sensor data generated in a virtual simulation environment based on a preset environmental model, physical rules, or motion constraints. This simulation data is used to mimic the sensing results of sensors in real-world environments under different scenarios, operating conditions, or execution scenarios. Its data format, structure, and semantics are consistent with or correspond to sensor data in real-world environments, thus enabling it to characterize the perceived information of target objects, environmental states, or execution processes. By introducing sensor data from a simulation environment, a wider range of sensor data with greater scenario diversity can be obtained without the need for actual equipment deployment or repeated testing.

[0037] Real-world sensor data refers to sensor data collected by actual equipment in a real operating environment. For example, real-world sensor data can refer to remote control data, which is sensor data collected synchronously by a human operator performing an operation in a real environment. Remote control data can accurately reflect actual environmental conditions, equipment response characteristics, and operational behavior features, and the sensor information it contains has high authenticity and reference value.

[0038] By using sensor data from both the simulated and real environments as the primary sensor data source, we can ensure data diversity while introducing perceptual features from the real environment. This allows subsequent processing based on the primary sensor data to possess both the scalability of the simulated data and the reliability of the real data, thereby improving the overall adaptability and generalization of data usage.

[0039] Determining multidimensional fused sensor data based on first-sensor data from multiple dimensions refers to the comprehensive processing of first-sensor data from different modal sensors, while maintaining the validity of information from each modality, to form a fused data representation that can uniformly characterize the target object, environmental state, or execution process. Multidimensional fused sensor data can simultaneously contain key information reflected by different modal sensors, ensuring the correlation and consistency of first-sensor data from different modalities within the same data space, thus avoiding the limitations of single-modal sensor data in terms of perception dimensions. By determining multidimensional fused sensor data based on first-sensor data from multiple dimensions, complementary relationships can be established between visual perception information, force perception information, and motion state perception information. This allows multidimensional fused sensor data to more comprehensively reflect the spatial state, interaction state, and motion state of the target object, providing a unified data foundation for subsequent state analysis, data correction, or control decisions.

[0040] Moreover, the specific implementation method can be as follows: Figure 2 As shown, it may include S1011 to S1013.

[0041] S1011: Perform data preprocessing operations on the first sensor data of multiple different dimensions to obtain first preprocessed sensor data.

[0042] As an example, for sensor data in the visual dimension, the locations of key feature points can be extracted, and the calculation process is shown in formula (1).

[0043] (1) For sensor data in the force dimension, contact force information can be obtained, and the calculation process is shown in formula (2).

[0044] (2) For sensor data in the inertial dimension, its motion state can be obtained, and the calculation process is shown in formula (3).

[0045] (3) In the formula, For linear acceleration, ω is the angular velocity.

[0046] S1012: Perform a synchronization alignment operation on the first preprocessed sensor data to obtain the second preprocessed sensor data.

[0047] Since sensors of different dimensions may differ in sampling frequency, triggering mechanism, and communication latency, a synchronization alignment operation needs to be performed on the first preprocessed sensor data to ensure consistency of multidimensional data in the time dimension. Through this synchronization alignment operation, preprocessed data from different dimensions can be mapped to a unified time reference, ensuring that sensor data from each dimension have a corresponding relationship within the same time window. This yields the second preprocessed sensor data, providing the foundation for subsequent multidimensional fusion.

[0048] S1013: Perform a multi-dimensional fusion operation on the second preprocessed sensor data to obtain multi-dimensional fused sensor data.

[0049] After obtaining the time-aligned second preprocessed sensor data, a multidimensional fusion operation is performed to comprehensively utilize the complementary information contained in the sensor data from different dimensions. By comprehensively characterizing the correlation, consistency, and complementarity of the sensor data from each dimension, multidimensional fused sensor data that can fully reflect the system state is formed.

[0050] S102: Determine the first multidimensional difference parameter based on the multidimensional fusion sensor data.

[0051] After obtaining multidimensional fused sensor data, the sensor data from the simulation environment and the real environment can be compared and analyzed based on the multidimensional fused sensor data to determine the first multidimensional difference parameter that reflects the difference characteristics between the two.

[0052] The first multidimensional difference parameter characterizes the degree of difference between sensor data in real and simulated environments across multiple sensing modalities, data features, and state representations. This degree of difference can be reflected as an overall trend or as changes in local features, thus enabling a multidimensional characterization of the perceptual deviation between the simulated and real environments. By determining the first multidimensional difference parameter based on multidimensional fused sensor data, the differences between sensor data in simulated and real environments can be described in a parameterized form. This unifies the expression of difference information that was originally scattered across different modalities or data formats, providing a clear basis for subsequent data correction, model adjustment, or environment migration.

[0053] The specific implementation method can be as follows: Figure 3 As shown, it may include S1021 to S1023.

[0054] S1021: Based on the calculation results of the difference between the first sensor data and the calibration value for each dimension, determine the data deviation of the first sensor data for each dimension.

[0055] Calibration values ​​are reference values ​​determined based on sensor data in real-world environments. They characterize the baseline sensing state of a sensor in a real-world operating environment. Calibration values ​​reflect the actual response characteristics of the sensor in a real-world environment and serve as a reference benchmark for comparing sensor data or fused data in a simulated environment.

[0056] After obtaining the first sensor data for each dimension, the first sensor data is compared with the calibration value for the corresponding dimension to determine the sensor data deviation, which characterizes the sensor data for that dimension relative to a real-world baseline. Sensor data deviation describes the degree of deviation between the sensor data in the simulation environment or the current sensor data and a real-world sensor baseline.

[0057] As an example, the sensor data deviation corresponding to the visual dimension can be shown in Equation 4.

[0058] (4) In the formula, This could refer to the sensor data deviation corresponding to the visual dimension. It can be the calibration data corresponding to the visual dimension. It could refer to the first sensor data corresponding to the visual dimension.

[0059] The sensor data deviation corresponding to the force dimension can be shown in Formula 5.

[0060] (5) In the formula, This could refer to the deviation in sensor data corresponding to the force dimension. It can be calibration data corresponding to the force dimension. It could refer to the data from the first sensor corresponding to the force perception dimension.

[0061] The sensor data deviation corresponding to the inertial dimension can be shown in Equation 6.

[0062] (6) In the formula, This could refer to the sensor data deviation corresponding to the inertial dimension. It can be the calibration data corresponding to the inertial dimension. It could refer to the first sensor data corresponding to the inertial dimension.

[0063] S1022: Determine the bias fusion weight of the first sensor data for each dimension based on the information reliability of the first sensor data for each dimension.

[0064] Information reliability characterizes the credibility of the first sensor data for a corresponding dimension under the current scenario or operating condition. It reflects the impact of environmental interference, acquisition stability, or sensing effectiveness on the sensor data for that dimension. Sensor data with high information reliability indicates a high consistency between the perceived information reflected in that dimension and the actual state, while sensor data with low information reliability indicates a relatively high uncertainty in the sensing results for that dimension. Based on the information reliability corresponding to the first sensor data for each dimension, corresponding deviation fusion weights can be assigned to the data deviations of the first sensor data for each dimension. Deviation fusion weights characterize the degree of influence of the deviations of sensor data in each dimension on the overall difference characterization, enabling the deviations of different dimensions to participate in the subsequent comprehensive difference determination process according to their reliability.

[0065] As an example, the calculation process for the bias fusion weight can be shown in Equation 7.

[0066] (7) In the formula, This could refer to the bias fusion weights of the first sensor data in the i-th dimension. It can refer to the reliability assessment result of the information of the first sensor data in the i-th dimension. It can be a smoothing factor, for example, It can be 0.8.

[0067] S1023: Determine the first multidimensional difference parameter based on the data deviation of the first sensor data in multiple dimensions and the multiplication and addition result of the corresponding deviation fusion weights.

[0068] After obtaining the sensor data deviation and corresponding deviation fusion weight for each dimension, the sensor data deviation and its corresponding deviation fusion weight for each dimension can be integrated to form the first multidimensional difference parameter used to characterize the overall difference between the simulation environment and the real environment.

[0069] As an example, the calculation process of the first multidimensional difference parameter can be shown in Equation 8.

[0070] (8) In the formula, It could refer to the first multidimensional difference parameter.

[0071] S103: Determine the overall confidence level of the multidimensional fusion sensor data based on the data quality of the multidimensional fusion sensor data and the reliability of the parameter estimation of the multidimensional fusion sensor data.

[0072] Data quality refers to the effectiveness and usability of sensor data during acquisition and use. It characterizes whether sensor data is complete, stable, and accurately reflects the corresponding physical or environmental state. High-quality sensor data typically exhibits fewer abnormal fluctuations, lower noise interference, or better continuity, while low-quality sensor data may contain missing, distorted, or unstable data. Parameter estimation reliability refers to the degree to which parameters obtained from sensor data reflect the true state. It characterizes the consistency between parameter estimation results and the actual environmental state. Parameter estimation reliability reflects whether a parameter has good stability, repeatability, or reference value, thus reflecting the reliability of the parameter in subsequent analysis or decision-making. Overall confidence level is an indicator that comprehensively characterizes the overall reliability of sensor data or parameters obtained from sensor data. Overall confidence level reflects the combined impact of factors such as data quality and parameter estimation reliability on the reliability of sensor data, and is used to uniformly assess the reliability of sensor data from different dimensions or different sources.

[0073] The data quality of the first sensor data in each dimension can be evaluated according to preset evaluation indicators. For example, for the visual dimension's first sensor data, its data quality can be evaluated based on the relative relationship between noise level and effective signal strength; when the noise level is higher relative to the signal response, the corresponding data quality score decreases. For the visual dimension's first sensor data, its data quality can be evaluated based on the deviation between the measured torque value and the expected torque value. For the inertial dimension's first sensor data... After obtaining the data quality assessment results of the first sensor data in each dimension, the data quality assessment results of the first sensor data in each dimension can be adaptively weighted and fused to obtain the comprehensive data quality assessment result. The specific calculation formula is shown in Formula 9.

[0074] (9) In the formula, This can be used to assess the overall data quality. The fusion weights of the first sensor data in each dimension can be assigned. It can provide data quality assessment results for the first sensor data in various dimensions.

[0075] The calculation formulas for the reliability of parameter estimation of sensor data in various dimensions can be shown in Formulas 10 to 14.

[0076] Fisher's information matrix theory is used to quantify the identifiability and accuracy of parameter estimation. The Fisher information matrix reflects the sensitivity of observed data to parameter changes, and its value is directly related to the confidence level of the parameter estimation.

[0077] Define the observation model function The output predictions for the physics simulator include state variables such as position, velocity, force, and torque. (10) In the formula, The vector of physical parameters to be estimated.

[0078] For any parameter The Fisher information is calculated as follows: (11) In the formula, To observe the noise covariance matrix, it is constructed using the noise characteristics of a multimodal sensor: (12) In the formula, , , These represent the dimensions of visual, force, and inertial observations, respectively.

[0079] The complete Fisher information matrix is ​​as follows: (13) In the formula, It is a Jacobian matrix.

[0080] To avoid matrix singularity issues, normalized confidence scores are used for calculation. (14) In the formula, To estimate the reliability of the comprehensive parameters.

[0081] After obtaining the comprehensive data quality assessment results and the confidence level of the comprehensive parameter estimation, the comprehensive confidence level of the multidimensional fused sensor data can be determined, and the calculation process can be shown in Formula 15.

[0082] (15) In the formula, For overall confidence level.

[0083] S104: Determine the sensor data correction strategy based on the overall confidence level of the multi-dimensional fused sensor data.

[0084] Determining the sensor data correction strategy based on the comprehensive confidence level of multidimensional fusion sensor data can enable differentiated correction under different data reliability levels, making the sensor data correction process adaptive and thus improving the effectiveness and stability of the correction results in real-world environments.

[0085] In one feasible implementation method, the specific implementation can be as follows: Figure 4 As shown, it may include S1041 to S1043.

[0086] S1041: If the overall confidence level is greater than the first overall confidence threshold, determine the sensor data correction as the first sensor data correction strategy; S1042: When the overall confidence level is greater than the second overall confidence level threshold and less than or equal to the first overall confidence level threshold, determine that the sensor data correction is the second sensor data correction strategy; S1043: If the overall confidence level is less than or equal to the second overall confidence level threshold, determine the sensor data correction as the third sensor data correction strategy.

[0087] The first sensor data correction strategy is used to adapt multidimensional fusion sensor data with high overall confidence. In this case, the multidimensional fusion sensor data has high consistency with the real environment, requiring only relatively conservative or light correction to maintain the validity of the original perceived information. The second sensor data correction strategy is used to adapt multidimensional fusion sensor data with medium overall confidence. In this case, by adopting a relatively balanced correction method, the differences between the simulated and real environments can be specifically corrected while preserving the original perceived characteristics, thereby improving the applicability of the sensor data. The third sensor data correction strategy is used to adapt multidimensional fusion sensor data with low overall confidence. In this case, the deviation between the multidimensional fusion sensor data and the real environment is relatively large. By adopting a more robust correction strategy, the impact of unreliable data on subsequent analysis or decision-making can be reduced.

[0088] As an example, the first and second comprehensive confidence thresholds can be pre-set confidence boundary parameters used to classify the comprehensive confidence of multi-dimensional fused sensor data. For instance, the first comprehensive confidence threshold could be 0.8, and the second comprehensive confidence threshold could be 0.4.

[0089] S105: Correct the first sensor data according to the sensor data correction strategy and the first multidimensional difference parameter to obtain the second sensor data.

[0090] After determining the sensor data correction strategy corresponding to the current sensor data, the corresponding correction strategy can be executed. The specific implementation method can include any one of S1051 to S1052, S1053 to S1054 or S1055 to S1056.

[0091] S1051: When the sensor data correction strategy is the first sensor data correction strategy, the correction frequency and correction order of the physical parameters in the first sensor data are determined according to the matching relationship between the physical parameters in the first sensor data and the preset data information. S1052: Based on the calibration frequency and calibration sequence, the physical parameters in the first sensor data are corrected using the first data adjustment step size to obtain the second sensor data.

[0092] When the sensor data correction strategy is the first sensor data correction strategy, it indicates that the multi-dimensional fused sensor data corresponding to the first sensor data has a high overall confidence level and a high degree of consistency with the sensor data in the real environment. Therefore, in this case, a hierarchical and progressive correction strategy can be adopted to quickly and stably correct the physical parameters in the sensor data, so as to avoid the instability risk caused by over-adjustment while ensuring correction efficiency.

[0093] The hierarchical, progressive correction strategy determines the correction frequency and order for each physical parameter based on the degree of influence of each physical parameter on the overall sensing result of the sensor data and the difficulty of identifying the corresponding parameter. The degree of influence characterizes the magnitude of the impact of changes in physical parameters on the sensing result or system state, while the difficulty of identification characterizes how easily a physical parameter can be accurately estimated or corrected under the current data conditions.

[0094] Specifically, physical parameters can be stratified according to the principle of decreasing influence and increasing identification difficulty, into multiple levels such as coarse-grained, fine-grained, and micro-grained parameters. Coarse-grained parameters have a greater overall impact on the sensing results and are relatively easy to identify, thus having the highest correction priority and the highest corresponding correction frequency. Fine-grained parameters have a lower correction priority and a moderate correction frequency. Micro-grained parameters have a relatively smaller overall impact and are more difficult to identify, thus having the lowest correction priority and the lowest corresponding correction frequency. The level information, corresponding correction frequency, and correction order of each physical parameter can be stored in preset data information. During the correction operation, the physical parameters in the first sensor data can be used as the matching basis. The corresponding correction order and correction frequency are matched in the data information, and the corresponding physical parameters are gradually corrected using the first data adjustment step size, thereby calibrating the physical parameters in the first sensor data and obtaining the corrected second sensor data.

[0095] As an example, the specific optimization process for coarse-grained parameters, fine-grained parameters, and particle-grained parameters can be shown in Equations 16 to 21.

[0096] First-layer coarse-grained parameter optimization: (16) (17) In the formula, This is the first layer of coarse-grained parameter vectors, which mainly contains global parameters that have the greatest impact on system dynamics (such as robot link mass and gravitational acceleration). This is a loss function constructed only for coarse-grained parameters. This represents the optimal coarse-grained parameter obtained after the first layer of optimization. In subsequent fine-grained optimization processes, this parameter is locked as a constant and is not updated.

[0097] Second-layer fine-grained parameter optimization: (18) (19) In the formula, This is the second-layer loss function. In this step, the first multidimensional difference parameter... The calculation depends on the fine-grained parameters to be optimized. (such as joint damping, friction coefficient) and the optimal coarse-grained parameters fixed in the previous layer. .

[0098] Third-layer micro-parameter optimization: (20) (twenty one) In the formula, This is the third layer fine-tuning loss function. At this point, the coarse-grained parameters... and fine-grained parameters All remain constant, except for microscopic parameters. Fine-tuning and optimization are performed on factors such as air resistance coefficient and minor sensor deviations.

[0099] S1053: When the sensor data correction strategy is the second sensor data correction strategy, the correction frequency and correction order of the physical parameters in the first sensor data are determined according to the matching relationship between the physical parameters in the first sensor data and the preset data information. S1054: Based on the calibration frequency and calibration order, the physical parameters in the first sensor data are corrected using the second data adjustment step size to obtain the second sensor data; the first data adjustment step size is greater than the second data adjustment step size.

[0100] When the sensor data calibration strategy is the second sensor data calibration strategy, it indicates that the overall confidence level of the multi-dimensional fused sensor data is at a moderate level, and there is still a certain degree of uncertainty and fluctuation in difference between the first sensor data and the sensor data in the real environment. In this case, if a large data adjustment step size is directly used for calibration, it is easy to cause frequent fluctuations in physical parameters during multiple calibration processes, leading to unstable calibration results. Therefore, in the second sensor data calibration strategy, while using a second data adjustment step size smaller than the first data adjustment step size to calibrate the physical parameters, regularization constraints can also be introduced to restrict the adjustment process of physical parameters, so as to suppress the oscillating changes of physical parameters during the calibration process. Regularization constraints are used to constrain the adjustment amplitude, adjustment trend, or continuity of change of physical parameters, so that the changes of physical parameters within adjacent calibration periods remain smooth and consistent.

[0101] It should be noted that the second sensor data correction strategy is the same as the layered and progressive correction strategy used in the first sensor data correction strategy, so it will not be described again.

[0102] S1055: When the sensor data correction strategy is the third sensor data correction strategy, the physical parameters in the first sensor data are corrected with the optimization objectives of maximizing information gain and minimizing risk to obtain the third sensor data.

[0103] S1056: Continue executing the step of determining multi-dimensional fused sensor data based on the third sensor data from multiple dimensions.

[0104] When the sensor data calibration strategy is the third sensor data calibration strategy, it indicates that the overall confidence level of the current multi-dimensional fused sensor data is at a low level, and the uncertainty contained in the first sensor data is relatively large. Directly applying conventional calibration to the physical parameters may cause the parameter estimates to deviate from the true values, thereby amplifying system errors or introducing potential risks. Therefore, in this case, the update process of the physical parameters can be restricted or temporarily frozen. The primary goal is no longer to quickly approximate the calibration value, but to transform the calibration process into an optimization problem guided by information acquisition and risk control. Here, information gain is used to characterize the ability to improve the identifiability of physical parameters and reduce uncertainty through the current calibration or data acquisition behavior; risk is used to characterize the safety risks, stability risks, or performance degradation risks that may be brought about by adjusting the physical parameters under the current environmental state and system constraints.

[0105] As an example, the specific implementation process for obtaining third sensor data can be as follows: First, several candidate operation actions can be generated. Candidate operation actions refer to a set of exploratory operation actions that the system can currently execute, under the premise of satisfying system safety constraints, environmental constraints, and execution capability constraints. Candidate operation actions may include, but are not limited to: changing the sensor's observation posture, adjusting the relative position between the sensor and the environment, switching the observation angle or observation range, executing a preset exploration path or sensing action, etc. Introducing a set of candidate operation actions provides a basis for subsequently selecting data acquisition methods with higher information value. Then, the information gain corresponding to each candidate operation action can be evaluated. Information gain characterizes the improvement effect on the degree of uncertainty reduction of physical parameters in the first sensor data after executing the corresponding candidate operation action and acquiring new sensor observation data. In other words, information gain can reflect the potential contribution of different candidate operation actions in improving the identifiability of physical parameters and reducing parameter estimation ambiguity. Finally, the target action is determined based on the information gain and risk value. More valuable data is obtained based on the target action, generating third sensor data.

[0106] The information gain for each candidate operation can be calculated using formulas 22 to 27.

[0107] (twenty two) (twenty three) (twenty four) In the formula, For parameter dimensions, The inverse of the Fisher information matrix represents the parametric covariance.

[0108] Conditional entropy Calculation via Bayesian update: (25) In the formula, To perform the action The Fisher information increment generated from the observations: (26) This is the Jacobian matrix related to the action.

[0109] To meet real-time requirements, a simplified information gain calculation is used: (27) After obtaining the information gain of each candidate operation, the comprehensive evaluation score of each candidate operation can be determined based on the information gain and risk value of each candidate operation. Then, the target operation can be determined based on the comprehensive evaluation score. The calculation formula and related constraint formulas of the comprehensive evaluation score can be shown in formulas 28 to 40.

[0110] (28) In the formula, To comprehensively evaluate the score, This represents the risk value of the action. This is the risk weighting coefficient.

[0111] (29) In the formula, To constrain the range of motion, This represents the maximum range of motion.

[0112] Maximum range of motion Based on robot joint limitations: (30) (31) (32) Collision safety constraints: (33) In the formula, To perform the action The minimum collision distance (the minimum Euclidean distance between the robot body and environmental obstacles). This is the safe distance threshold.

[0113] Joint constraint: (34) In the formula, , Due to joint angle limitations, Changes in joint angle caused by movement.

[0114] Dynamic feasibility constraints: (35) In the formula, The joint torque required to perform the action, This represents the upper limit of the joint torque.

[0115] Risk measurement function: (36) In the formula, each risk component is defined as: Collision risk: (37) Stability risks: (38) In the formula, This is a preset safe reset posture for the robot or a static hold posture at the current moment, used to prevent the robot's center of gravity from becoming unstable due to exploratory movements.

[0116] Energy consumption risk: (39) In the formula, This is a positive definite weighting matrix used to adjust the weights of different action components on energy consumption costs. The weighting coefficients are set as follows: , , Constrained optimization is solved using sequential quadratic programming (SQP): In the formula, Let inequality constraint vector be . This is the equality constraint vector.

[0117] S1056: Based on the third sensor data, continue to execute the step of determining multi-dimensional fused sensor data based on the third sensor data from multiple dimensions.

[0118] After obtaining the third sensor data, multidimensional fusion processing can be re-executed based on this data to update the overall perception of the current environmental and system states. By fusing the third sensor data, the consistency between data from different modal sensors and changes in overall confidence can be reassessed.

[0119] When the sensor data correction strategy is the third sensor data correction strategy, an active exploration mechanism can be triggered to obtain high-quality sensor data while pausing or weakening the physical parameter update process. The active exploration mechanism guides the system to obtain sensor data with greater informational value regarding physical parameters by adjusting the observation attitude, performing specific exploration actions, or changing the sensing conditions, while meeting safety constraints and risk limits.

[0120] S106: Based on the second sensor data from multiple dimensions, determine the second multidimensional difference parameter. If the second multidimensional difference parameter is less than or equal to the multidimensional difference parameter threshold, determine the second sensor data as the target sensor data.

[0121] After obtaining the second sensor data, a second multidimensional difference parameter can be constructed based on the differences between the second sensor data and the first sensor data before calibration. This second multidimensional difference parameter can then be compared with a preset threshold. The preset threshold represents the minimum requirements for sensor data consistency and stability in the current application scenario. When the second multidimensional difference parameter is less than or equal to the preset threshold, it indicates that the second sensor data has achieved the expected calibration target in multiple dimensions, the physical parameter adjustment results tend to be stable, and the calibration process has not introduced significant abnormal fluctuations.

[0122] In the above scenario, the second sensor data can be identified as the target sensor data. The target sensor data can serve as input data for subsequent multi-dimensional fusion processing, task decision-making, or control execution. It can also be used as reference data in subsequent data correction processes to reduce unnecessary repetitive correction operations.

[0123] In one feasible implementation, the method further includes: if the second multidimensional difference parameter is greater than the multidimensional difference parameter threshold, continuing to perform the step of determining the overall confidence level of the sensor data.

[0124] When the second multidimensional difference parameter exceeds the multidimensional difference parameter threshold, it indicates that the current sensor data calibration has not yet achieved the expected stability or accuracy. At this point, the step of determining the overall confidence level of the sensor data can be performed to further evaluate the consistency and reliability of the current sensor data across multiple dimensions. By calculating the overall confidence level, the parameters and steps in the calibration process can be dynamically adjusted to ensure that the calibration effect meets the target requirements.

[0125] The data correction method provided in this application overcomes the limitations of existing solutions that focus on a single modality or a small amount of sensor data by acquiring multi-dimensional sensor data and using a multi-dimensional fusion mechanism. By simultaneously acquiring first sensor data from different dimensions and forming multi-dimensional fused sensor data based on preprocessing and synchronous alignment, it can fully explore the complementary information between multi-source heterogeneous sensors. This allows the characterization of differences between the real and simulated environments to no longer rely on a single observation dimension, but rather to comprehensively represent them from multiple perception levels, thereby improving the comprehensiveness and accuracy of environmental difference modeling. Furthermore, based on the multi-dimensional fused sensor data, this application uses a joint evaluation mechanism of data quality and parameter estimation reliability to determine the comprehensive confidence level. By quantitatively evaluating the quality of sensor data and the reliability of parameter estimation, the impact of low-confidence data on the correction results can be reduced in cases of high data noise, insufficient observation information, or partial sensor anomalies. This avoids instability problems caused by blind or over-correction, improving the reliability and security of the data correction process from a mechanism perspective. In addition, this application significantly enhances the system's adaptive capability during operation through a dynamic correction strategy selection mechanism driven by comprehensive confidence level. Unlike traditional sim-to-real methods that rely on offline modeling or static parameter correction, this application dynamically adjusts the correction strategy, frequency, and magnitude of sensor data based on changes in multidimensional difference parameters and overall confidence level during system operation. This allows the correction process to continuously evolve with environmental changes, effectively adapting to environmental uncertainties under long-term operation or complex conditions. When multidimensional difference parameters remain large and overall confidence level is low, this application utilizes a correction mechanism aimed at information gain and risk control. By triggering active exploration or information enhancement processes, it compensates for the information deficiency caused by insufficient environmental stimuli in passive observation. This approach proactively acquires data with greater discriminative power for parameter correction, further improving the alignment between simulation and real-world environments and overcoming the performance bottleneck of traditional passive sim-to-real methods in complex environments.

[0126] Figure 5 This is a block diagram of a data correction device provided in an embodiment of this application. Figure 5 As shown, the data correction device 500 includes: The acquisition module 501 is used to acquire first sensor data in multiple dimensions and determine multi-dimensional fused sensor data based on the first sensor data in multiple dimensions. The multidimensional difference parameter determination module 502 is used to determine the first multidimensional difference parameter based on the multidimensional fused sensor data; The comprehensive confidence determination module 503 is used to determine the comprehensive confidence of the multi-dimensional fusion sensor data based on the data quality of the multi-dimensional fusion sensor data and the confidence of the parameter estimation of the multi-dimensional fusion sensor data. The calibration strategy determination module 504 is used to determine the sensor data calibration strategy based on the comprehensive confidence level of the multi-dimensional fused sensor data. The calibration module 505 is used to calibrate the first sensor data in multiple dimensions according to the sensor data calibration strategy and the first multidimensional difference parameter, so as to obtain the second sensor data in multiple dimensions. The judgment module 506 is used to determine a second multidimensional difference parameter based on the second sensor data in multiple dimensions. If the second multidimensional difference parameter is less than or equal to the multidimensional difference parameter threshold, the second sensor data is determined as the target sensor data. The multidimensional difference parameter is used to characterize the degree of difference between the sensor data in the real environment and the simulation environment.

[0127] In other embodiments, an electronic device is provided, which may be the data correction device described above, for performing the method steps executed by the terminal device or cloud device in the above method flow. The internal structure diagram of this electronic device may be as follows: Figure 6 As shown, the device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements a data correction method. The display unit is used to form a visually visible image and can be a display screen, projection device, or virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the electronic device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the electronic device, or external keyboards, touchpads, or mice, etc.

[0128] Those skilled in the art will understand that Figure 6 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the electronic device to which the present application is applied. The specific electronic device may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0129] In some embodiments, an electronic device includes a memory, a processor, and a communication interface. The communication interface is used to interact with other devices to send and receive data. For example, in this embodiment, the communication interface may specifically be used to store computer program code, which includes computer instructions. These computer instructions run in the electronic device to implement the method shown in the above-described method embodiments. For example, the memory may include high-speed random access memory (RAM), and may also include non-volatile memory (NVM), such as at least one disk storage device, and may also be a USB flash drive, portable hard drive, read-only memory, disk, or optical disk, etc.

[0130] The processor can be a general-purpose processor, including a Central Processing Unit (CPU), a network processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. The processor can also be other general-purpose processors. A general-purpose processor can be a microprocessor or any conventional processor.

[0131] Memory, communication interfaces, and processor communication connections. For example, memory and communication interfaces can connect to the processor via the system bus and communicate with each other. The system bus can be a Peripheral Component Interconnect (PCI) bus, an Extended Industry Standard Architecture (EISA) bus, an Industry Standard Architecture (ISA) bus, etc. The system bus can be divided into address bus, data bus, control bus, etc. For ease of representation, only one thick line is used in the diagram, but this does not indicate that there is only one bus or one type of bus.

[0132] Alternatively, the memory can be either standalone or integrated with the processor. When the memory is set up independently, it is connected to the processor via the system bus.

[0133] This application also provides a chip for executing instructions, which is used to execute the data correction method described in the above embodiments.

[0134] This application also provides a computer-readable storage medium storing computer instructions. When these computer instructions are executed by a processor, they are used to implement the technical solution of the data correction method described in the above embodiments. Specifically, when the computer instructions are executed by a processor, the computer device can perform the technical solution of the data correction method provided in the above embodiments.

[0135] This application also provides a computer program product, which includes a computer program stored in a computer-readable storage medium. At least one processor can read the computer program from the computer-readable storage medium, and when the at least one processor executes the computer program, it can implement the technical solution of the data correction method provided in the above embodiments.

[0136] The aforementioned computer-readable storage media can be implemented by any type of volatile or non-volatile storage device or a combination thereof. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical storage, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), etc.

[0137] Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can take many forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). Computer-readable storage media may be any available medium accessible to general-purpose or special-purpose computers.

[0138] An exemplary computer-readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the storage medium. Of course, the computer-readable storage medium can also be a component of the processor. The processor and the computer-readable storage medium can reside in an application-specific integrated circuit (ASIC). Alternatively, the processor and the computer-readable storage medium can exist as discrete components in an electronic control unit or main control device; this application does not limit this.

[0139] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules may be combined or integrated into another system, or some parameters may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interfaces, devices, or modules, and may be electrical, mechanical, or other forms.

[0140] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules in the formula can be selected to implement the solution of this embodiment according to actual needs.

[0141] Furthermore, the functional modules in the various embodiments of this application can be integrated into one processing unit, or each module can exist physically separately, or two or more modules can be integrated into one unit. The unit composed of the above modules can be implemented in hardware or in the form of hardware plus software functional units.

[0142] The integrated modules described above, implemented as software functional modules, can be stored in a computer-readable storage medium. These software functional modules, stored in a storage medium, include several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute some steps of the methods of the various embodiments of this application.

[0143] It should be understood that the steps of the method disclosed in the embodiments of this application can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules in the processor.

[0144] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0145] The technical parameters in the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical parameters in the above embodiments are described. However, as long as these combinations of technical parameters do not contradict each other, they should be considered within the scope of this specification. The above embodiments only illustrate several implementation methods of this application, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of this application's patent. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

[0146] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical parameters in the formula. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. A data correction method, characterized in that, The method includes: Acquire first sensor data from multiple dimensions, and determine multidimensional fused sensor data based on the first sensor data from multiple dimensions; Based on the multidimensional fused sensor data, the first multidimensional difference parameter is determined; The overall confidence level of the multidimensional fusion sensor data is determined based on the data quality of the multidimensional fusion sensor data and the reliability of the parameter estimation of the multidimensional fusion sensor data. Based on the overall confidence level of the multidimensional fused sensor data, a sensor data correction strategy is determined; Based on the sensor data correction strategy and the first multidimensional difference parameter, the first sensor data of multiple dimensions is corrected to obtain the second sensor data of multiple dimensions. Based on the second sensor data from the multiple dimensions, a second multidimensional difference parameter is determined. If the second multidimensional difference parameter is less than or equal to the multidimensional difference parameter threshold, the second sensor data is determined as the target sensor data. The multidimensional difference parameter is used to characterize the degree of difference between the sensor data in the real environment and the simulation environment.

2. The data correction method according to claim 1, characterized in that, The step of determining the first multidimensional difference parameter based on the multidimensional fused sensor data includes: The data deviation of the first sensor data in each dimension is determined based on the difference between the first sensor data and the calibration value in each dimension. Based on the reliability of the information from the first sensor data in each dimension, determine the bias fusion weight of the first sensor data in each dimension; The first multidimensional difference parameter is determined based on the data deviation of the first sensor data in multiple dimensions and the multiplication and addition result of the corresponding deviation fusion weights.

3. The data correction method according to claim 1, characterized in that, The step of determining the sensor data correction strategy based on the comprehensive confidence level of the multidimensional fused sensor data includes: If the overall confidence level is greater than the first overall confidence threshold, the sensor data correction is determined to be the first sensor data correction strategy. If the overall confidence level is greater than the second overall confidence level threshold and less than or equal to the first overall confidence level threshold, the sensor data correction is determined to be the second sensor data correction strategy. If the overall confidence level is less than or equal to the second overall confidence level threshold, the sensor data correction is determined to be a third sensor data correction strategy; the first overall confidence level threshold is greater than the second overall confidence level threshold.

4. The data correction method according to claim 1, characterized in that, The step of correcting the first sensor data according to the sensor data correction strategy and the first multidimensional difference parameter to obtain the second sensor data includes: When the sensor data correction strategy is the first sensor data correction strategy, the correction frequency and correction order of the physical parameters in the first sensor data are determined according to the matching relationship between the physical parameters in the first sensor data and the preset data information. Based on the correction frequency and correction order, the physical parameters in the first sensor data are corrected using a first data adjustment step size to obtain the second sensor data; the data information includes physical parameters, the physical parameters and their corresponding correction frequencies, and the correction order corresponding to the physical parameters; the physical parameters and their corresponding correction frequencies and correction order are determined based on the degree of influence and identification difficulty of the physical parameters.

5. The data correction method according to claim 1, characterized in that, The step of correcting the first sensor data according to the sensor data correction strategy and the first multidimensional difference parameter to obtain the second sensor data includes: When the sensor data correction strategy is the second sensor data correction strategy, the correction frequency and correction order of the physical parameters in the first sensor data are determined according to the matching relationship between the physical parameters in the first sensor data and the preset data information. Based on the correction frequency and correction order, the physical parameters in the first sensor data are corrected using a second data adjustment step size to obtain the second sensor data; the first data adjustment step size is larger than the second data adjustment step size.

6. The data correction method according to claim 1, characterized in that, The step of correcting the first sensor data according to the sensor data correction strategy and the first multidimensional difference parameter to obtain the second sensor data includes: When the sensor data correction strategy is the third sensor data correction strategy, the physical parameters in the first sensor data are corrected with the optimization objectives of maximizing information gain and minimizing risk to obtain the third sensor data. Based on the third sensor data, the process continues to execute the step of determining multi-dimensional fused sensor data based on the third sensor data from multiple dimensions.

7. The data correction method according to claim 1, characterized in that, The first sensor data includes at least one of the following: Sensor data in a simulated environment or sensor data in a real environment.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the data correction method as described in any one of claims 1-7.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that, when executed by a processor, are used to implement the data correction method as described in any one of claims 1-7.

10. A computer program product, characterized in that, When the computer program product is run on a computer / executed by the computer's processor, it implements the data correction method as described in any one of claims 1-7.

Images