Signal processing method and device, electronic equipment and storage medium
By evaluating the confidence of the sensor and using nonlinear autoregressive neural network for data prediction, front-end signal fusion of the line-controlled steering system is realized, solving the problem of insufficient real-time data fusion in the prior art, and improving response speed and security.
Patent Information
- Application Number
- CN202411384258.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-30
- Publication Date
- 2025-01-10
AI Technical Summary
The existing wire-controlled steering system lacks real-time performance in data fusion, resulting in insufficient steering control response speed, affecting driving safety.
By evaluating the confidence of the sensor, selecting the sensor with the highest confidence for data fusion, using a nonlinear autoregressive neural network to predict the steering data at the next moment, realizing front-end signal fusion and reducing dependence on the central processing unit.
It improves the system's response speed and real-time performance, ensuring the steering system's fast response capability and driving safety.
Smart Images
Figure CN119293748A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of autonomous driving technology, and in particular to a signal processing method, device, electronic device and storage medium. Background Art
[0002] Steering By Wire (SBW) is an electronically controlled steering technology that eliminates the mechanical connection between the steering wheel and the steering wheel in the traditional steering system and uses electronic signal transmission to achieve steering control. This system replaces the physical mechanical linkage with the transmission of electronic signals, thus getting rid of the limitations of traditional mechanical structures and achieving more flexible steering angle transmission characteristics. The SBW system is considered to be one of the most advanced steering technologies currently, because it not only improves the vehicle's handling performance, but also lays the foundation for integrating higher-level autonomous driving functions.
[0003] To ensure accurate and timely steering control, the SBW system needs to process information from multiple sensors (such as cameras, radars, lidars, etc.) in real time. These sensors provide diverse data about the vehicle's surroundings, but since the data from a single sensor is often insufficient to make accurate judgments, data fusion is required to comprehensively analyze information from different sources. Given that the steering system is critical to driving safety, it must be able to quickly respond to the driver's operating instructions and input from other sensors, which places high demands on the real-time performance of data fusion. Summary of the invention
[0004] In order to solve the problems of the prior art, the embodiments of the present application provide a signal processing method, device, electronic device and storage medium. The technical solution is as follows:
[0005] In one aspect, a signal processing method is provided, the method comprising:
[0006] In response to a data update instruction for a target steering parameter at a current moment, acquiring historical steering data of the target steering parameter, predicted steering data of the target steering parameter at a current moment, and historical steering signals and current steering signals of a plurality of sensors corresponding to the target steering parameter;
[0007] For each of the sensors, based on the predicted steering data of the target steering parameter at the current moment, the historical steering signal of the sensor and the current steering signal, the confidence of the sensor is evaluated;
[0008] Based on the confidence of each of the sensors, a target sensor is selected from the multiple sensors corresponding to the target steering parameter; the confidence of the target sensor is higher than the confidence of the sensors not selected from the multiple sensors;
[0009] Determining current steering data of the target steering parameter based on a current steering signal of the target sensor;
[0010] Based on the historical steering data of the target steering parameter and the current steering data, the predicted steering data of the target steering parameter at the next moment is determined.
[0011] In another aspect, a signal processing device is provided, the device comprising:
[0012] A data acquisition module, configured to respond to a data update instruction for a target steering parameter at a current moment, acquire historical steering data of the target steering parameter, predicted steering data of the target steering parameter at a current moment, and historical steering signals and current steering signals of a plurality of sensors corresponding to the target steering parameter;
[0013] A confidence evaluation module, configured to evaluate the confidence of each sensor based on the predicted steering data of the target steering parameter at the current moment, the historical steering signal of the sensor, and the current steering signal;
[0014] A target selection module, configured to select a target sensor from the multiple sensors corresponding to the target steering parameter based on the confidence of each of the sensors; the confidence of the target sensor is higher than the confidence of the unselected sensors from the multiple sensors;
[0015] a data determination module, configured to determine current steering data of the target steering parameter based on a current steering signal of the target sensor;
[0016] The data prediction module is used to determine the predicted steering data of the target steering parameter at the next moment based on the historical steering data of the target steering parameter and the current steering data.
[0017] In an exemplary embodiment, the confidence assessment module includes:
[0018] A dimension evaluation module, configured to evaluate the confidence of the sensor from multiple evaluation dimensions based on the predicted steering data of the target steering parameter at the current moment, the historical steering signal and the current steering signal of the sensor, and obtain the score of the sensor under each of the evaluation dimensions;
[0019] The confidence determination module is used to determine the confidence of the sensor based on the score of the sensor in each of the evaluation dimensions.
[0020] In an exemplary embodiment, the data prediction module includes:
[0021] The model prediction module is used to input the historical steering data of the target steering parameter and the current steering data as observation values into a pre-trained nonlinear autoregressive neural network to obtain the predicted steering data of the target steering parameter at the next moment.
[0022] In an exemplary embodiment, the multiple evaluation dimensions include an accuracy index and a reliability index; the dimension evaluation module includes:
[0023] an accuracy evaluation module, for determining a deviation of a current steering signal of the sensor relative to the predicted steering data of the target steering parameter at the current moment, and obtaining a score of the sensor under the accuracy index;
[0024] The reliability evaluation module is used to determine the number of times the sensor fails based on the historical steering signal of the sensor, and obtain the score of the sensor under the reliability index.
[0025] In an exemplary embodiment, the confidence determination module includes:
[0026] A dimension scoring module, used to determine a plurality of scores corresponding to each evaluation dimension from the scores of each of the plurality of sensors under each evaluation dimension;
[0027] The first score processing module is used to perform weighted summation of the scores of the sensor under multiple evaluation dimensions based on the discreteness of the multiple scores corresponding to each evaluation dimension, so as to obtain the confidence of the sensor.
[0028] In an exemplary embodiment, the first score processing module includes:
[0029] An information entropy determination module is used to determine, for each of the evaluation dimensions, the discreteness of multiple scores corresponding to the evaluation dimension, and obtain the information entropy corresponding to the evaluation dimension;
[0030] An entropy weight determination module, used to determine the entropy weight corresponding to the evaluation dimension based on the information entropy corresponding to the evaluation dimension;
[0031] The second score processing module is used to perform weighted summation on the scores of the sensor under the multiple evaluation dimensions based on the entropy weight corresponding to each of the evaluation dimensions to obtain the confidence of the sensor.
[0032] On the other hand, an electronic device is provided, comprising a processor and a memory, wherein the memory stores at least one instruction or at least one program, and the at least one instruction or the at least one program is loaded and executed by the processor to implement the signal processing method of any one of the above aspects.
[0033] On the other hand, a computer-readable storage medium is provided, in which at least one instruction or at least one program is stored, and the at least one instruction or the at least one program is loaded and executed by a processor to implement a signal processing method as described in any of the above aspects.
[0034] On the other hand, a computer program product or a computer program is provided, the computer program product or the computer program includes computer instructions, the computer instructions are stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the electronic device performs the signal processing method of any of the above aspects.
[0035] The embodiment of the present application comprehensively analyzes the predicted steering data of the target steering parameter determined at the previous moment and the data acquired at the current moment, and evaluates the confidence of multiple sensors corresponding to the target steering parameter, so as to determine the current steering data of the target steering parameter. It is possible to complete the signal fusion of multiple sensors at the same time as signal acquisition, so that data processing and signal fusion are completed at the front end, that is, front-end fusion is realized, and there is no need to transmit data back to the central processing unit for processing, which greatly improves the response speed of the system, takes into account both real-time and accuracy, and ensures the rapid response capability and driving safety of the steering system. BRIEF DESCRIPTION OF THE DRAWINGS
[0036] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.
[0037] Figure 1 is a flowchart of a signal processing method provided in an embodiment of the present application;
[0038] Figure 2 is a schematic diagram of a sliding window provided in an embodiment of the present application;
[0039] Figure 3 is an example diagram of the evaluation dimensions provided in the embodiments of the present application;
[0040] Figure 4 It is a structural topological diagram of a nonlinear autoregressive neural network provided in an embodiment of the present application;
[0041] Figure 5 This is a single-step prediction error effect diagram of a neural network training using vehicle speed as an example provided in an embodiment of the present application;
[0042] Figure 6 is a flowchart of another signal processing method provided in an embodiment of the present application;
[0043] Figure 7 It is a schematic diagram of a framework of a signal processing method provided in an embodiment of the present application;
[0044] Figure 8 is a structural block diagram of a signal processing device provided in an embodiment of the present application;
[0045] Fig. 9 It is a hardware structure block diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0046] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.
[0047] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.
[0048] It is understandable that in the specific implementation of this application, related data such as user information is 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 relevant data need to comply with relevant laws, regulations and standards of relevant countries and regions.
[0049] See also Figure 1, which shows a flow chart of a signal processing method provided by an embodiment of the present application. It should be noted that this specification provides method operation steps such as the embodiments or flow charts, but may include more or fewer operation steps based on conventional or non-creative labor. The order of steps listed in the embodiments is only one way of executing the steps among many orders, and does not represent the only order of execution. When the system or product is executed in practice, it can be executed in the order shown in the embodiments or the drawings or in parallel (for example, in a parallel processor or multi-threaded processing environment). Specifically, Figure 1 As shown, the method may include:
[0050] S101, in response to a data update instruction for a target steering parameter at the current moment, obtain historical steering data of the target steering parameter, predicted steering data of the target steering parameter at the current moment, and historical steering signals and current steering signals of multiple sensors corresponding to the target steering parameter.
[0051] The current moment refers to the time point that the system is processing, and the system collects, analyzes and predicts data based on this time point.
[0052] The target steering parameters refer to the key parameters that can be detected by sensors in the wire steering system, which are used to monitor and control the steering behavior of the vehicle, such as vehicle speed, steering angle, steering wheel angle, etc.
[0053] The data update instruction is a command issued by the system to instruct sensors and systems to update the data they collect and process, ensuring that the data in the system is up to date so that the system can make accurate decisions based on the latest information. Specifically, the data update instruction is intended to update the current steering parameters of the target steering parameters.
[0054] Among them, historical steering data refers to the recorded values of the target steering parameters in the past period of time, which is used to help the system analyze trends, identify abnormal situations, and predict future steering behaviors. Specifically, historical steering data is the steering data of the target steering parameters at historical moments. The historical moments occur before the current moment, and the historical moments are continuous with the current moment. The historical moments can include multiple moments.
[0055] The predicted steering data of the target steering parameter at the current moment refers to the predicted value made by the system at the previous moment for the steering data of the target steering parameter at the current moment.
[0056] The multiple sensors corresponding to the target steering parameters are devices installed on the vehicle for detecting the target steering parameters, for example, multiple speed sensors corresponding to the vehicle speed.
[0057] The historical steering signal refers to the signal data collected from the sensor in the past, that is, the signal data output by multiple sensors corresponding to the target steering parameters at historical moments, which records the historical changes of the target steering parameters.
[0058] The current steering signal refers to signal data output by multiple sensors corresponding to the target steering parameters at the current moment, and these data reflect the current steering state of the vehicle.
[0059] For specific implementation, please refer to Figure 2 , which shows a schematic diagram of a sliding window provided by an embodiment of the present application. By selecting a suitable sliding window time length, information feature parameters in the corresponding time domain are collected. The total length of the sliding window should not be too short or too long. Too short will result in insufficient signal feature information being collected; but too long will increase the amount of calculation for real-time control. Specifically, within the total length of the sliding window, there is a part of the area called the historical data sampling window t-△T~t, and the historical steering data of the target steering parameter in this time period is collected for subsequent analysis and prediction. After the historical steering data is collected, the predicted steering data of the target steering parameter at the current time t, as well as the historical steering signals and current steering signals of multiple sensors, are combined to perform data analysis to determine the steering data of the target steering parameter at the current time t, that is, the current steering data; within the total length of the sliding window, there is another part of the area called the prediction window, in which the historical steering data obtained previously and the current steering data are used for prediction to generate the predicted value of the target steering parameter at the time t~t+△T, that is, the predicted steering parameter at the next time. As the current time t is updated, the sliding window moves rightward along the time axis. When the sliding window reaches a new position, the above steps are repeated, that is, collecting new historical data, analyzing this data, and then predicting future data. This process will continue to cycle, and each time the sliding window moves to a new position, historical data will be collected again, analyzed, and predicted again to ensure the real-time and accuracy of the system.
[0060] S103, for each sensor, based on the predicted steering data of the target steering parameter at the current moment, the historical steering signal of the sensor, and the current steering signal, the confidence of the sensor is evaluated.
[0061] The confidence level represents the reliability of the output signal of the sensor, and reflects the quality of the output signal of the sensor and the trust level of the system in the signal.
[0062] Specifically, each sensor has a certain measurement error range, which may be caused by the limitations of the sensor itself (such as accuracy, resolution) or external factors (such as temperature changes, electromagnetic interference). The smaller the measurement error, the higher the confidence. Highly consistent data indicates that the sensor can stably provide reliable measurement results. If multiple sensors corresponding to the target steering parameters provide similar readings in the same environment, these readings can be considered to have a high degree of confidence. Whether the data provided by the sensor at different time points is stable will also affect its confidence. A sensor with large fluctuations over time may not be as trustworthy as a sensor with stable output. Whether the sensor is properly calibrated will also affect its confidence. An uncalibrated or out-of-calibration sensor may give erroneous readings. For multiple sensors corresponding to the target steering parameters, the overall confidence can be improved by combining the information of multiple sensors through data fusion technology. For example, by comparing data from different sensors to identify outliers, or performing cross-checks, outputting sensor signals with higher reliability. The past performance record of the sensor is also an important indicator for evaluating confidence. If a sensor has always performed well in the past, its future readings may be more trustworthy.
[0063] In practice, the confidence of the sensor can be quantified by mathematical methods, such as using probability distribution to represent uncertainty, or by giving a confidence interval to represent the trustworthy range of the data. For example, the output signals of the sensor are evaluated to determine which output signals can be used for the decision-making of the steer-by-wire system and which signals should be ignored or require further verification.
[0064] In an exemplary embodiment, the above step S103 may include the following steps:
[0065] Based on the predicted steering data of the target steering parameter at the current moment, the historical steering signal of the sensor, and the current steering signal, the confidence of the sensor is evaluated from multiple evaluation dimensions to obtain the score of the sensor in each evaluation dimension;
[0066] The confidence of the sensor is determined based on the score of the sensor under each evaluation dimension.
[0067] For example, see Figure 3 , which shows an example diagram of the evaluation dimensions provided by the embodiment of the present application, and the multiple evaluation dimensions include but are not limited to reliability index, accuracy index, cost index, etc. Among them, the reliability index can be reflected as the number of sensor failures, where the failure generally refers to recoverable failures, usually excluding hardware failures; the accuracy index can be reflected as the deviation between the steering signal output by the sensor and the predicted value; the cost index can be reflected as a cost coefficient.
[0068] In specific implementation, multiple evaluation dimensions may also include real-time indicators, robustness indicators, adaptability indicators, etc. Among them, the real-time indicator is reflected in the speed of sensor data update and the time required for data transmission to the processing unit; the robustness indicator is reflected in the sensor's resistance to external interference (such as electromagnetic interference, temperature changes, etc.); the adaptability indicator is reflected in the sensor's ability to adapt to different environmental conditions.
[0069] Specifically, in the case of multiple evaluation dimensions including accuracy index and reliability index, the deviation of the current steering signal of the sensor relative to the predicted steering data of the target steering parameter at the current moment is determined to obtain the score of the sensor under the accuracy index; based on the historical steering signals of the sensor, the number of times the sensor has failed is determined to obtain the score of the sensor under the reliability index.
[0070] Wherein, deviation = |current steering signal - predicted steering data|.
[0071] Specifically, define what kind of data points are considered faults. For example, define data points that obviously deviate from the normal working range or exhibit abnormal behavior as faults. Use an anomaly detection algorithm to identify faulty data points in historical turn signals, and then count the number of faults that occurred in historical turn signals.
[0072] Specifically, among the scores of each sensor in each evaluation dimension, multiple scores corresponding to each evaluation dimension are determined; based on the discreteness of the multiple scores corresponding to each evaluation dimension, the scores of the sensor in the multiple evaluation dimensions are weighted and summed to obtain the confidence of the sensor.
[0073] The degree of dispersion indicates the performance differences of different sensors under the same evaluation dimension. Specifically, the greater the degree of dispersion of the evaluation dimension, the greater the impact (i.e., weight) of the evaluation dimension on the comprehensive evaluation. If multiple scores corresponding to a certain evaluation dimension are all equal, the evaluation dimension will not play a role in the comprehensive evaluation.
[0074] Exemplarily, if multiple evaluation dimensions include indicators such as accuracy, consistency, real-time, robustness, reliability, and adaptability, the score of each sensor under each evaluation dimension is calculated. For each evaluation dimension, the scores of all sensors under this dimension are collected. Specifically, the accuracy scores of all sensors are collected, the consistency scores of all sensors are collected, the real-time scores of all sensors are collected, the robustness scores of all sensors are collected, the reliability scores of all sensors are collected, the adaptability scores of all sensors are collected, and the discrete degree of the scores corresponding to each evaluation dimension is calculated.
[0075] Specifically, for each evaluation dimension, the discrete degree of multiple scores corresponding to the evaluation dimension is determined to obtain the information entropy corresponding to the evaluation dimension; based on the information entropy corresponding to the evaluation dimension, the entropy weight corresponding to the evaluation dimension is determined; based on the entropy weight corresponding to each evaluation dimension, the scores of the sensor under multiple evaluation dimensions are weighted and summed to obtain the confidence of the sensor.
[0076] Among them, information entropy is an indicator to measure uncertainty, which is used to describe the uncertainty of multiple score distributions corresponding to the evaluation dimensions.
[0077] Among them, the entropy weight reflects the relative importance of the evaluation dimension in the evaluation, and is used as the weight of the evaluation dimension to calculate the confidence of the sensor.
[0078] Specifically, for each evaluation dimension, the information entropy is calculated based on the discrete degree of the score. In the specific implementation, the scores under each evaluation dimension are standardized so that they are at the same level, and the proportion of each sensor score to the total score is calculated. The information entropy can be calculated using formula (1):
[0079]
[0080] Among them, H j is the information entropy of the jth evaluation dimension, p ij is the score ratio of the i-th sensor on the j-th evaluation dimension, and n represents the number of sensors.
[0081] Specifically, the entropy weight is calculated based on the information entropy of each evaluation dimension. In a specific implementation, the entropy weight can be calculated using formula (2):
[0082]
[0083] where w j is the entropy weight of the jth evaluation dimension, H j is the information entropy of the jth evaluation dimension, and m represents the number of evaluation dimensions.
[0084] It can be seen from the above technical solutions of the embodiments of the present application that the embodiments of the present application evaluate the confidence of the sensor from multiple evaluation dimensions through the predicted steering data of the target steering parameter at the current moment, the historical steering signal of the sensor and the current steering signal. Specifically, by determining the entropy weight corresponding to the evaluation dimension based on the discrete degree of multiple scores corresponding to each evaluation dimension, and then performing weighted summation to obtain the confidence of the sensor, the performance of the sensor can be comprehensively evaluated to ensure the comprehensiveness and objectivity of the evaluation results; by determining the entropy weight based on the information entropy of the evaluation dimension, the weight can be adaptively adjusted according to the performance of the sensor in different evaluation dimensions, thereby improving the flexibility and applicability of the evaluation results; by real-time evaluation of the current steering signal and predicted steering data, the system needs can be quickly responded to, and the response speed and real-time performance of the system can be improved.
[0085] S105 , selecting a target sensor from a plurality of sensors corresponding to the target steering parameter based on the confidence level of each sensor.
[0086] The confidence level of the target sensor is higher than the confidence levels of the sensors that are not selected from the multiple sensors.
[0087] Specifically, the target sensor refers to a sensor used to provide a steering signal, and the current steering signal of the target sensor is used by the steer-by-wire system for a decision-making process, such as determining an actual steering angle of the vehicle.
[0088] In practice, selecting target sensors is an ongoing process because the performance of sensors will change over time and with environmental conditions. Therefore, it is necessary to continuously evaluate the confidence of sensors and update the selection of target sensors in real time.
[0089] S107, determining current steering data of the target steering parameter based on the current steering signal of the target sensor.
[0090] Among them, the current steering data refers to the specific value of the target steering parameter measured by the target sensor at the current moment, which is used to reflect the current steering state of the vehicle and is one of the important bases for the wire control system to make decisions.
[0091] In a specific implementation, it can be determined through the above steps that the data of the target sensor is sufficiently accurate and reliable relative to other sensors, and the current steering signal of the target sensor can be directly used as the current steering data. If necessary, the current steering signal can be fine-tuned or corrected according to the calibration parameters of the sensor. The current steering data is used for further decision-making or control processes.
[0092] S109, determining predicted steering data of the target steering parameter at the next moment based on the historical steering data and the current steering data of the target steering parameter.
[0093] The predicted steering data of the target steering parameter at the next moment refers to the predicted value of the steering data of the target steering parameter at the next moment made by the system at the current moment. Specifically, the predicted steering data of the target steering parameter at the next moment is used to update the steering data of the target steering parameter at the next moment.
[0094] It can be seen from the above technical solutions of the embodiment of the present application that the embodiment of the present application comprehensively analyzes the predicted steering data of the target steering parameter determined at the previous moment and the data acquired at the current moment, and evaluates the confidence of multiple sensors corresponding to the target steering parameter, so as to determine the current steering data of the target steering parameter. It is possible to complete the signal fusion of multiple sensors at the same time as signal acquisition, so that data processing and signal fusion are completed at the front end, that is, front-end fusion is realized, and there is no need to transmit data back to the central processing unit for processing, which greatly improves the response speed of the system, takes into account both real-time and accuracy, and ensures the rapid response capability and driving safety of the steering system.
[0095] In an exemplary embodiment, the above step S109 may include the following steps:
[0096] The historical steering data and current steering data of the target steering parameters are taken as observation values and input into a pre-trained nonlinear autoregressive neural network (NARNN) to obtain the predicted steering data of the target steering parameters at the next moment.
[0097] For details, please refer to Figure 4 , which shows the structural topology diagram of the nonlinear autoregressive neural network provided by the embodiment of the present application. In this network, the input node represents the historical steering data, the hidden layer node is used to capture the complex relationship in the input data, and the output node generates the predicted steering data. The specific NAR neural network uses past observations to predict future time series data. The input node contains a series of historical observations, X 1 ~X n Represent the first n observation points of the time series, and each hidden layer node (H 1 ~H n) receive data from the input nodes and form an interactive relationship. The purpose of the hidden layer nodes is to extract features from the input data and convert them into a representation that can be used for prediction. The output of the hidden layer nodes is aggregated to a single output node, which is responsible for generating the final prediction result, that is, the predicted steering data. Based on this, the NAR neural network is able to learn patterns in historical steering data and use these patterns to predict the steering data at the next moment. Due to its nonlinear nature, it can handle time series data with complex relationships, which is very useful for many real-world application scenarios.
[0098] In specific implementation, the NAR neural network can be trained through the back propagation algorithm to minimize the error between the predicted value and the true value. Through continuous iteration and adjustment of weights, the network can gradually improve the prediction accuracy. Specifically, historical steering data, including steering data of target steering parameters at different time points, are collected and used as training sets and test sets; a NAR neural network model is created, including an input layer, a hidden layer, and an output layer, and initial weights are randomly assigned to each connection in the network; the historical steering data is input into the network, and the values of the hidden layer and the output layer are calculated by the activation function. This step generates a preliminary predicted steering data; the predicted steering data is compared with the actual steering data, and the difference between the two is calculated. The mean square error or other suitable loss function can be used; according to the result of the loss function, the weights and biases of the network are adjusted by the gradient descent method or a similar optimization algorithm to reduce the prediction error; the forward propagation and backpropagation steps are continued to be looped until a predetermined stopping criterion is reached, such as the maximum number of iterations, the loss function converges, or the performance on the validation set is no longer improved; the performance of the model is evaluated using a test data set that is not involved in the training to check its generalization ability; if the model performs poorly, it may be necessary to adjust the network architecture (increase / decrease the number of hidden layer nodes), change the activation function, try different optimizers, or change other hyperparameters; once satisfactory model performance is obtained, it can be deployed in a practical environment for real-time prediction of steering data of target steering parameters.
[0099] See also Figure 5 , which shows a single-step prediction error effect diagram of the neural network training using vehicle speed as an example provided in the embodiment of the present application. It can be seen from the figure that in the entire time series, the predicted value has a certain similarity with the actual value and the simulated value, but they do not completely overlap, especially in certain time periods, there is a large gap between the predicted value and the actual value, indicating that the prediction of the neural network does not always perfectly match the actual data. However, in terms of the overall trend, the predicted value can still follow the changes in the actual value well, indicating that the neural network has successfully captured the basic laws of the steering data to a certain extent.
[0100] It can be seen from the above technical solutions of the embodiments of the present application that the embodiments of the present application can capture the complex relationships in the steering data and improve the prediction accuracy by utilizing the nonlinear modeling capability of the nonlinear autoregressive neural network. The nonlinear autoregressive neural network can adaptively learn the patterns in the steering data and maintain a high prediction accuracy even when the steering characteristics change. By inputting the current steering data in real time, the steering data at the next moment can be quickly predicted, thereby improving the response speed of the system at the next moment.
[0101] In order to facilitate a full understanding of the solution of this application, the following Figure 6 , 7 Take the signal processing process as an example to explain the overall process.
[0102] See also Figure 6 , which is a flow chart of another signal processing method provided by an embodiment of the present application, and the signal processing process may specifically include:
[0103] (1) Determine whether the initial confidence levels of each sensor are the same;
[0104] Specifically, if the result of the judgment is yes, step (2) can be executed; otherwise, if the result of the judgment is no, step (6) can be executed.
[0105] (2) determining whether the sliding window used to collect data meets the sliding window duration;
[0106] Specifically, if the result of the judgment is yes, then step (3) can be executed; otherwise, if the result of the judgment is no, then step (7) can be executed.
[0107] (3) Divide the sliding window and extract features from the steering signal output by the sensor within the sliding window.
[0108] Specifically, the signal is low-pass filtered, the sensor signal within the prediction window is predicted using a nonlinear autoregressive neural network (NAR), and the deviation of the sensor's current steering signal relative to the predicted steering data of the target steering parameter at the current moment is calculated; as well as the number of sensor failures within the statistical window.
[0109] (4) Use the entropy weight method to calculate the comprehensive score of each sensor.
[0110] (5) Select the current steering signal of the sensor with the highest score as the current steering data of the target steering parameter.
[0111] (6) The current steering signal of the sensor with the highest initial confidence is used as the current steering data of the target steering parameter.
[0112] (7) Using the current steering signal of a default sensor among the multiple sensors corresponding to the target steering parameter as the current steering data of the target steering parameter.
[0113] See also Figure 7 , which shows a schematic diagram of the framework of a signal processing method provided by an embodiment of the present application. The entire system continuously updates and fuses sensor data through a sliding window, uses the NAR neural network for prediction, and stores the results in a storage module. Specifically, the sliding window feature extraction module is responsible for extracting features from the steering data at multiple times, including the original value of the signal and the fault count, and these features will be passed to the post-signal filtering module; after receiving the original value of the signal, the signal filtering module filters the signal to obtain the filtered signal value, and the filtered signal value will be sent to the data storage module; the sensor fault counting module is responsible for recording the number of sensor failures, and the number of failures will also be transmitted to the data storage module, and the data storage module will pass the received data to the data fusion module; the data fusion module combines the number of failures and the filtered signal value to perform data fusion, and the fused output value will be sent to the data storage module and the NAR neural network prediction module; the NAR neural network prediction module predicts the fused data, and the prediction result will be fed back to the data storage module to update the prediction deviation.
[0114] Corresponding to the signal processing methods provided in the above-mentioned embodiments, an embodiment of the present application also provides a signal processing device. Since the signal processing device provided in the embodiment of the present application corresponds to the signal processing methods provided in the above-mentioned embodiments, the implementation methods of the aforementioned signal processing methods are also applicable to the signal processing device provided in this embodiment and will not be described in detail in this embodiment.
[0115] See also Figure 8 , which is a schematic diagram of the structure of a signal processing device provided in an embodiment of the present application, the device has the function of implementing the signal processing method in the above method embodiment, and the function can be implemented by hardware or by hardware executing corresponding software. Figure 8 As shown, the device may include:
[0116] A data acquisition module 810 is used to respond to a data update instruction for a target steering parameter at a current moment, acquire historical steering data of the target steering parameter, predicted steering data of the target steering parameter at a current moment, historical steering signals of multiple sensors corresponding to the target steering parameter, and current steering signals;
[0117] A confidence evaluation module 820, for evaluating the confidence of each sensor based on the predicted steering data of the target steering parameter at the current moment, the historical steering signal of the sensor, and the current steering signal;
[0118] A target selection module 830 is used to select a target sensor from a plurality of sensors corresponding to a target steering parameter based on the confidence of each sensor; the confidence of the target sensor is higher than the confidence of the unselected sensor from the plurality of sensors;
[0119] A data determination module 840, for determining current steering data of a target steering parameter based on a current steering signal of a target sensor;
[0120] The data prediction module 850 is used to determine the predicted steering data of the target steering parameter at the next moment based on the historical steering data and the current steering data of the target steering parameter.
[0121] In an exemplary embodiment, the confidence assessment module includes:
[0122] A dimension evaluation module is used to evaluate the confidence of the sensor from multiple evaluation dimensions based on the predicted steering data of the target steering parameter at the current moment, the historical steering signal of the sensor, and the current steering signal, and obtain the score of the sensor in each evaluation dimension;
[0123] The confidence determination module is used to determine the confidence of the sensor based on the score of the sensor in each evaluation dimension.
[0124] In an exemplary embodiment, the data prediction module includes:
[0125] The model prediction module is used to input the historical steering data and current steering data of the target steering parameter as observation values into a pre-trained nonlinear autoregressive neural network to obtain the predicted steering data of the target steering parameter at the next moment.
[0126] In an exemplary embodiment, the multiple evaluation dimensions include an accuracy index and a reliability index; the dimension evaluation module includes:
[0127] An accuracy evaluation module is used to determine the deviation of the current steering signal of the sensor relative to the predicted steering data of the target steering parameter at the current moment, and obtain the score of the sensor under the accuracy index;
[0128] The reliability evaluation module is used to determine the number of sensor failures based on the historical steering signals of the sensor and obtain the score of the sensor under the reliability index.
[0129] In an exemplary embodiment, the confidence determination module includes:
[0130] A dimension scoring module, used to determine a plurality of scores corresponding to each evaluation dimension from the scores of each sensor in each evaluation dimension among the plurality of sensors;
[0131] The first score processing module is used to perform weighted summation of the scores of the sensor under multiple evaluation dimensions based on the discreteness of the multiple scores corresponding to each evaluation dimension to obtain the confidence of the sensor.
[0132] In an exemplary embodiment, the first score processing module includes:
[0133] An information entropy determination module is used to determine the discreteness of multiple scores corresponding to each evaluation dimension, and obtain the information entropy corresponding to the evaluation dimension;
[0134] An entropy weight determination module, used to determine the entropy weight corresponding to the evaluation dimension based on the information entropy corresponding to the evaluation dimension;
[0135] The second score processing module is used to perform weighted summation of the scores of the sensor under multiple evaluation dimensions based on the entropy weight corresponding to each evaluation dimension to obtain the confidence of the sensor.
[0136] It should be noted that the device provided in the above embodiment, when implementing its functions, is only illustrated by the division of the above functional modules. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the device and method embodiments provided in the above embodiment belong to the same concept, and the specific implementation process is detailed in the method embodiment, which will not be repeated here.
[0137] An embodiment of the present application provides an electronic device, which includes a processor and a memory, wherein the memory stores at least one instruction or at least one program, and the at least one instruction or the at least one program is loaded and executed by the processor to implement any one of the signal processing methods provided in the above method embodiments.
[0138] The memory can be used to store software programs and modules. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, application programs required for functions, etc.; the data storage area can store data created according to the use of the device, etc. In addition, the memory can include a high-speed random access memory and can also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other volatile solid-state storage devices. Accordingly, the memory can also include a memory controller to provide the processor with access to the memory.
[0139] The method embodiments provided in the embodiments of the present application may be executed in a computer terminal, a server or a similar computing device, that is, the above-mentioned electronic device may include a computer terminal, a server or a similar computing device. Fig. 9 is a hardware structure block diagram of a computer device running a signal processing method provided by an embodiment of the present invention, such as Fig. 9 As shown, the internal structure of the computer device may include but is not limited to: a processor, a network interface and a memory. The processor, network interface and memory in the computer device may be connected via a bus or other means. Fig. 9 The example of connecting through bus is taken in the following.
[0140] Among them, the processor (or CPU (Central Processing Unit)) is the computing core and control core of the computer device. The network interface may optionally include a standard wired interface, a wireless interface (such as WI-FI, a mobile communication interface, etc.). Memory is a memory device in a computer device for storing programs and data. It is understandable that the memory here can be a high-speed RAM storage device or a non-volatile memory device, such as at least one disk storage device; optionally, it can also be at least one storage device located away from the aforementioned processor. The memory provides a storage space, which stores the operating system of the electronic device, which may include but is not limited to: Windows system (an operating system), Linux (an operating system), Android (Android, a mobile operating system) system, IOS (a mobile operating system) system, etc., and the present invention is not limited to this; and, in the storage space, one or more instructions suitable for being loaded and executed by the processor are also stored, and these instructions may be one or more computer programs (including program codes). In the embodiment of this specification, the processor loads and executes one or more instructions stored in the memory to implement the signal processing method provided in the above method embodiment.
[0141] An embodiment of the present application also provides a computer-readable storage medium, which can be set in an electronic device to store at least one instruction or at least one program related to implementing a signal processing method. The at least one instruction or the at least one program is loaded and executed by the processor to implement any one of the signal processing methods provided in the above method embodiments.
[0142] Optionally, in this embodiment, the above-mentioned storage medium may include but is not limited to: a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk, and other media that can store program codes.
[0143] It should be noted that the above-mentioned sequence of the embodiments of the present application is for description only and does not represent the advantages and disadvantages of the embodiments. The above-mentioned specific embodiments of this specification are described. Other embodiments are within the scope of the attached claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0144] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.
[0145] A person skilled in the art will understand that all or part of the steps to implement the above embodiments may be accomplished by hardware or by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a disk or an optical disk, etc.
[0146] The above are only preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.
Claims
1. A signal processing method, characterized in that: The method comprises: In response to a data update instruction for a target steering parameter at a current moment, acquiring historical steering data of the target steering parameter, predicted steering data of the target steering parameter at a current moment, and historical steering signals and current steering signals of a plurality of sensors corresponding to the target steering parameter; For each of the sensors, based on the predicted steering data of the target steering parameter at the current moment, the historical steering signal of the sensor and the current steering signal, the confidence of the sensor is evaluated; Based on the confidence of each of the sensors, a target sensor is selected from the multiple sensors corresponding to the target steering parameter; the confidence of the target sensor is higher than the confidence of the sensors not selected from the multiple sensors; Determining current steering data of the target steering parameter based on a current steering signal of the target sensor; Based on the historical steering data of the target steering parameter and the current steering data, the predicted steering data of the target steering parameter at the next moment is determined.
2. The signal processing method according to claim 1, characterized in that: The evaluating the confidence of the sensor based on the predicted steering data of the target steering parameter at the current moment, the historical steering signal of the sensor and the current steering signal includes: Based on the predicted steering data of the target steering parameter at the current moment, the historical steering signal and the current steering signal of the sensor, the confidence of the sensor is evaluated from multiple evaluation dimensions to obtain the score of the sensor in each evaluation dimension; The confidence level of the sensor is determined based on the score of the sensor under each of the evaluation dimensions.
3. The signal processing method according to claim 2, characterized in that: The step of determining predicted steering data of the target steering parameter at a next moment based on the historical steering data of the target steering parameter and the current steering data includes: The historical steering data of the target steering parameter and the current steering data are used as observation values and input into a pre-trained nonlinear autoregressive neural network to obtain the predicted steering data of the target steering parameter at the next moment.
4. The signal processing method according to claim 3, characterized in that: The multiple evaluation dimensions include an accuracy index and a reliability index; the confidence of the sensor is evaluated from multiple evaluation dimensions based on the predicted steering data of the target steering parameter at the current moment, the historical steering signal of the sensor, and the current steering signal, to obtain the score of the sensor under each evaluation dimension, including: Determining a deviation of a current steering signal of the sensor relative to predicted steering data of the target steering parameter at a current moment, and obtaining a score of the sensor under the accuracy index; Based on the historical turning signals of the sensor, the number of times the sensor fails is determined, and the score of the sensor under the reliability index is obtained.
5. The signal processing method according to claim 2, characterized in that: The determining the confidence of the sensor based on the score of the sensor under each of the evaluation dimensions includes: Among the scores of each of the multiple sensors under each evaluation dimension, determine a plurality of scores corresponding to each evaluation dimension; Based on the discreteness of the multiple scores corresponding to each of the evaluation dimensions, the scores of the sensor under the multiple evaluation dimensions are weighted and summed to obtain the confidence of the sensor.
6. The signal processing method according to claim 5, characterized in that: The step of performing weighted summation of the scores of the sensor under the multiple evaluation dimensions based on the discreteness of the multiple scores corresponding to each evaluation dimension to obtain the confidence of the sensor includes: For each of the evaluation dimensions, determining the discreteness of multiple scores corresponding to the evaluation dimension, and obtaining the information entropy corresponding to the evaluation dimension; Based on the information entropy corresponding to the evaluation dimension, determining the entropy weight corresponding to the evaluation dimension; Based on the entropy weight corresponding to each of the evaluation dimensions, the scores of the sensor under the multiple evaluation dimensions are weighted and summed to obtain the confidence of the sensor.
7. A signal processing device, characterized in that: The device comprises: a data acquisition module, configured to respond to a data update instruction for a target steering parameter at a current moment, acquire historical steering data of the target steering parameter, predicted steering data of the target steering parameter at a current moment, historical steering signals and current steering signals of a plurality of sensors corresponding to the target steering parameter; A confidence evaluation module, configured to evaluate the confidence of each sensor based on the predicted steering data of the target steering parameter at the current moment, the historical steering signal of the sensor, and the current steering signal; A target selection module, configured to select a target sensor from the multiple sensors corresponding to the target steering parameter based on the confidence of each of the sensors; the confidence of the target sensor is higher than the confidence of the unselected sensors from the multiple sensors; a data determination module, configured to determine current steering data of the target steering parameter based on a current steering signal of the target sensor; The data prediction module is used to determine the predicted steering data of the target steering parameter at the next moment based on the historical steering data of the target steering parameter and the current steering data.
8. An electronic device, characterized in that: The method comprises a processor and a memory, wherein the memory stores at least one instruction or at least one program, and the at least one instruction or at least one program is loaded and executed by the processor to implement the signal processing method according to any one of claims 1 to 6.
9. A computer-readable storage medium, wherein at least one instruction or at least one program is stored in the storage medium, and the at least one instruction or the at least one program is loaded and executed by a processor to implement the signal processing method according to any one of claims 1 to 6.
10. A computer program, characterized in that When the computer program is executed by a processor, the signal processing method according to any one of claims 1 to 6 is implemented.