A dynamic calibration and adaptive signal decoding method based on a psi5 interface

Abstract

The application relates to the technical field of signal decoding, in particular to a dynamic calibration and adaptive signal decoding method based on a PSI5 interface, which comprises the following steps: collecting prehistory data and sensor output analog signals under different working environmental conditions; performing feature extraction on the analog signals to obtain feature parameters; pre-processing historical signal data; dynamically calibrating the feature parameters of the analog signals by using calibration rules to obtain accurate analog signals; generating a decoding rule of a current signal by using a decision tree algorithm according to the accurate analog signals and a PSI5 protocol standard; decoding the accurate analog signals according to the decoding rule to obtain effective data; comparing the effective data with expected data to judge whether there is deviation; and if yes, adjusting the calibration rule and the decision tree algorithm parameters according to the deviation. The application reduces invalid data transmission and improves signal accuracy by dynamically calibrating the deviation of the signal and adjusting the decoding parameters.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of signal decoding, in particular to a dynamic calibration and adaptive signal decoding method based on a PSI5 interface. BACKGROUND

[0002] PSI5 is a widely used two-way digital communication protocol in the field of automotive electronics, mainly used for data transmission between sensors and electronic control units, which uses a two-wire current interface and Manchester coding to achieve efficient and reliable signal transmission.

[0003] Traditional signal calibration methods are often static and based on fixed parameter settings. Traditional decoding rules are based on fixed parameters and cannot adapt to signal changes under different working conditions. In the process of vehicle operation, it is difficult to adjust the calibration parameters flexibly in the face of various real-time changing working conditions, which may cause large deviations between calibrated data and actual signals. And the conventional signal decoding method is mostly designed for standard ideal signals, lacking the ability to adapt to signal changes in actual complex environments. Once the signal appears, such as waveform distortion caused by interference, signal characteristics change caused by sensor aging, etc., decoding errors or decoding failures will occur. Therefore, there is an urgent need for a signal decoding method that can dynamically calibrate and adapt to overcome these shortcomings and improve the accuracy and reliability of signal processing based on the PSI5 interface. SUMMARY

[0004] The present application provides a dynamic calibration and adaptive signal decoding method based on a PSI5 interface to solve the problems of low precision, poor robustness and resource waste caused by static calibration errors, fixed decoding rules and lack of feedback optimization in the prior art.

[0005] The present application provides a dynamic calibration and adaptive signal decoding method based on a PSI5 interface, comprising:

[0006] S1: Collecting historical signal data and sensor output analog signals under different working environment conditions.

[0007] S2: Feature extraction is performed on the analog signal to obtain feature parameters. The historical signal data is preprocessed to obtain prehistory data.

[0008] S3: Using calibration rules to dynamically calibrate the feature parameters to obtain accurate analog signals.

[0009] S4: According to the accurate analog signal and the PSI5 protocol standard, a decision tree algorithm is used to generate a decoding rule that adapts to the current signal.

[0010] S5: Decoding the accurate analog signal according to the decoding rule to obtain valid data.

[0011] S6: Comparing the effective data with the expected data, adjusting the calibration rule and the decision tree algorithm parameter according to the deviation, and obtaining accurate effective data.

[0012] According to the dynamic calibration and adaptive signal decoding method based on the PSI5 interface provided by the application, in step S1, the analog signal is converted into a digital signal through the PSI5 interface, and the digital signal contains multiple data frames, and each data frame contains a synchronization bit, a data bit and a check bit.

[0013] According to the dynamic calibration and adaptive signal decoding method based on the PSI5 interface provided by the application, in step S2, the characteristic parameters include the initial amplitude, frequency, phase and data frame format of the analog signal. The preprocessing includes: performing outlier processing on the historical signal data, One-Hot encoding, time encoding, data normalization and adding Gaussian noise processing.

[0014] According to the dynamic calibration and adaptive signal decoding method based on the PSI5 interface provided by the application, in step S3, the dynamic calibration of the characteristic parameters using the calibration rule includes:

[0015] S31: Set the structure parameters of the time series integrated model, and collect real-time environmental parameters.

[0016] S32: Train the time series integrated model using the prehistory data.

[0017] S33: Input the real-time environmental parameters into the trained time series integrated model to predict the current calibration parameter, convert the current calibration parameter into an operable calibration rule, and dynamically adjust the signal characteristic parameters according to the calibration rule.

[0018] According to the dynamic calibration and adaptive signal decoding method based on the PSI5 interface provided by the application, in step S32, the training of the time series integrated model includes:

[0019] S321: Divide the prehistory data into a training set one and a test set one.

[0020] S322: According to the structure characteristics of the time series integrated model and the characteristic parameters of the prehistory data, set the hyperparameters of the current time series integrated model.

[0021] S323: Convert the training set to a low-dimensional learnable space using the embedder of the time series integrated model.

[0022] S324: In the low-dimensional learnable space, the discriminator updates the hyperparameters of the current time series integrated model according to the gradient descent algorithm until the comprehensive evaluation index of the time series integrated model is continuously reduced for 2 rounds to stop updating the hyperparameters.

[0023] S325: test the time series integration model using the test set until the trained time series integration model meets the preset threshold, generate new data using the generator, and fuse the new data with the original prehistory data to obtain new prehistory data.

[0024] According to the dynamic calibration and adaptive signal decoding method based on the PSI5 interface provided by the application, in step S324, the training of the time series integration model includes: in the hyperparameter updating process, the generator continuously generates false data trying to deceive the discriminator, the discriminator finds out the false data output by the generator, and the two repeatedly perform adversarial training, and after completing each round of adversarial training, the time series integration model is tested using the test set, the model performance index and the change of the hyperparameter are monitored, and the time series integration model stops updating the hyperparameter.

[0025] According to the dynamic calibration and adaptive signal decoding method based on the PSI5 interface provided by the application, in step S4, the specific steps of generating an adaptive current signal decoding rule using a decision tree algorithm are as follows:

[0026] S41: Set the parameters of the decision tree algorithm according to the number of feature parameters.

[0027] S42: Construct a training set two and a test set two by combining the analog signal feature parameters and the correct decoding under the corresponding preset decoding rule, and input the training set two into the decision tree algorithm.

[0028] S43: Node splitting and branch road forming are performed on the best feature to form a decision tree, and the categories corresponding to the leaf nodes of the decision tree are the preliminary decoding rules.

[0029] S44: A group of current analog signals with known correct decoding results are used to decode the current analog signals according to the preliminary decoding rules to obtain the current analog signal decoding rule.

[0030] According to the dynamic calibration and adaptive signal decoding method based on the PSI5 interface provided by the application, in step S5, the decoding of the accurate analog signal includes:

[0031] S51: Find the starting frame in the accurate analog signal according to the definition of the signal frame structure in the decoding rule.

[0032] S52: From the starting frame position, check whether the accurate analog signal frame after the starting frame is complete according to the decoding rule, if yes, keep it unchanged, otherwise use a signal repair algorithm to repair the accurate analog signal frame.

[0033] S53: According to the position and length range of the data segment in the signal frame in the decoding rule, extract the analog signal data corresponding to the data segment from the complete accurate analog signal frame.

[0034] S54: Convert the analog signal data segment into digital signal data according to the decoding conversion mode of the decoding rule for the data segment.

[0035] S55: Parse and restore the digital signal data according to the data format in the decoding rule to obtain valid data.

[0036] According to the dynamic calibration and adaptive signal decoding method based on the PSI5 interface provided by the application, in step S53, the data segment includes:

[0037] Position-related data: used to indicate the specific position of the data segment in the entire signal frame.

[0038] Length range data: the length range occupied by the data segment itself.

[0039] Encoding feature data: the encoding feature of the data segment in the form of an analog signal.

[0040] Decoding conversion data: the specific conversion mode of how the data segment is converted from an analog signal to a digital signal.

[0041] Data format data: the data format requirement followed when the data segment is finally parsed and restored.

[0042] According to the dynamic calibration and adaptive signal decoding method based on the PSI5 interface provided by the application, in step S6, after adjustment, the parsed parameters are error checked, and it is judged whether the error checking result exceeds the threshold value, and if so, the valid data is compensated using redundant data.

[0043] The dynamic calibration and adaptive signal decoding method based on the PSI5 interface provided by the application uses the calibration rule to dynamically calibrate the characteristic parameters of the analog signal to accurately decode the analog signal, solves the low precision defect caused by static calibration error and fixed decoding rule in the prior art, and achieves the beneficial effects of:

[0044] The application can monitor the deviation of the sensor output signal in real time through dynamic calibration, and adjust the decoding parameters through an algorithm to ensure signal accuracy. Traditional fixed calibration methods are difficult to cope with complex and variable actual working conditions, while the method obtains the original analog signal through differential signal acquisition technology, and adjusts the characteristic parameters in real time using a dynamic calibration rule. Differential signal acquisition can effectively suppress common-mode interference and ensure the integrity of the original signal. Dynamic calibration adjusts the calibration parameters according to real-time working conditions, so that the analog signal always maintains high precision.

[0045] The PSI5 interface in the application adopts Manchester coding and current modulation transmission, and has anti-interference capability. The adaptive decoding further optimizes through an algorithm, dynamically adjusts the decoding threshold or filtering parameters, and adapts to noise interference under different working conditions. The adaptive decoding can identify the signal characteristics of different sensors, realize "plug and play", and does not need to configure decoding parameters for each sensor, thereby reducing the system complexity. The adaptive decoding rule generation mechanism based on the PSI5 protocol can dynamically adjust the decoding strategy according to the current signal characteristics and channel state. The PSI5 protocol is a standard interface for automobile sensor networks, and the flexible frame structure and coding mode thereof provide a basis for adaptive decoding. The system automatically selects the optimal decoding algorithm by analyzing the frequency, duty cycle, noise level and other parameters of the signal, and combining the preset decoding rule library.

[0046] In the application, the adaptive decoding can dynamically adjust the decoding frequency or sampling rate according to the signal strength, avoid unnecessary calculation overhead, and prolong the service life of the vehicle-mounted battery. The dynamic calibration can compress signal redundancy, reduce invalid data transmission, and further improve the bus utilization rate in combination with the efficient coding mode of the PSI5 interface. By comparing the decoded data with the expected value in real time, the system can quickly identify the deviation and trigger closed-loop adjustment. The parameter modification of the dynamic calibration rule is based on the deviation analysis result, for example, adjusting the gain coefficient or compensation factor to offset environmental interference. The optimization of the decision tree algorithm parameters is updated by iteratively updating the model weight through machine learning technology, so that the decoding rule continuously approaches the optimal state. If the verification fails, the retransmission mechanism or the redundant data compensation strategy can ensure the data integrity and avoid system failure caused by single decoding error.

[0047] In the application, the sensor health status can be monitored in real time during the dynamic calibration process, potential faults can be found in advance, and warning information can be sent through the vehicle-mounted network. The automation characteristics of the adaptive decoding reduce the need for manual intervention and reduce maintenance costs, and are especially suitable for large-scale vehicle fleet management. BRIEF DESCRIPTION OF DRAWINGS

[0048] In order to more clearly illustrate the technical solutions in the application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor.

[0049] Fig. 1 is a step diagram of a dynamic calibration and adaptive signal decoding method based on a PSI5 interface provided by an embodiment of the application;

[0050] Fig. 2 is a flowchart of training a time series integrated model using prehistory data provided by an embodiment of the application. Detailed Implementation

[0051] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0052] The following is combined Figs. 1-2 This invention describes a dynamic calibration and adaptive signal decoding method based on the PSI5 interface.

[0053] like Fig. 1 As shown in the figure, an embodiment of the present invention provides a dynamic calibration and adaptive signal decoding method based on the PSI5 interface, comprising:

[0054] S1: Collects historical signal data under different working environment conditions. It uses differential signal acquisition to collect the analog signal output by the sensor and converts it into a digital signal stream through the PSI5 interface. The analog signal contains multiple data frames, and each data frame contains a synchronization bit, a data bit, and a parity bit.

[0055] Synchronization bits are crucial components in data frames used to achieve clock and data synchronization between the transmitting and receiving ends. During signal transmission, the transmitting end sends data according to a specific clock rhythm, and the receiving end also needs to accurately receive and parse the data according to the same rhythm. Synchronization bits play a vital role in ensuring the receiving end "finds the right rhythm." They help the receiving end determine the start position of a data frame, enabling subsequent data bits to be correctly identified and processed, avoiding data misalignment, misinterpretation, and other issues, thus ensuring the accuracy and orderliness of data transmission.

[0056] Data bits are the core part of a data frame, carrying the actual information to be transmitted. They contain various useful data that the sender wants to convey to the receiver, and these data can represent different meanings depending on the specific application scenario. For example, in a sensor network, data bits may contain numerical codes of physical quantities such as temperature, humidity, and pressure collected by sensors. In a communication system, data bits may be the encoded binary representation of text, images, audio, or other information to be transmitted. After receiving the data bits, the receiver will restore them to the original information content according to the corresponding decoding rules to achieve effective information transmission.

[0057] A checksum bit is used to detect errors in data transmission. Since signals may be subject to various interferences during transmission, such as electromagnetic interference and noise interference, some bits in the data may change. The checksum bit helps the receiver determine whether the received data is correct. Using a specific checksum algorithm, the sending end generates a checksum bit based on the data bit content and adds it to the data frame. The receiving end uses the same algorithm to perform a checksum calculation on the received data, comparing the calculated result with the received checksum bit to determine if there are any transmission errors, thus improving the reliability of data transmission.

[0058] S2: Extract features from the acquired initial signal to obtain characteristic parameters such as the initial amplitude, frequency, phase, and data frame format of the signal.

[0059] Amplitude is a key indicator describing the strength or magnitude of a signal. For different types of signals, such as analog electrical signals and mechanical vibration signals, amplitude represents the magnitude of the signal in the corresponding physical quantity dimension.

[0060] Frequency refers to the number of times a signal repeats itself periodically per unit time, measured in Hertz (Hz). It is an important parameter characterizing the periodicity of a signal; signals of different frequencies often have different physical origins and functions. In communications, electromagnetic waves of different frequencies are used to carry and transmit different types of information. In mechanical systems, the frequency of vibration of equipment components can reflect their operating status and whether there are potential faults such as loosening or wear. Normally operating equipment typically has vibration frequencies within a specific, reasonable range.

[0061] Phase describes the position of a signal relative to a reference point at a given moment, and is usually represented by angles. Phase plays a crucial role when comparing or combining multiple signals.

[0062] Data frame format primarily concerns the organization of digital signals during transmission and storage. It specifies how signal data is segmented, identified, and what additional information it includes. Different application scenarios often have different data frame format requirements. For example, in communication networks, the data frame format determines the methods of data transmission, reception, and parsing, ensuring accurate information transmission between different devices. In the process of sensor network data acquisition and transmission to a host computer, the data frame format clarifies how the data collected by each sensor is arranged, facilitating accurate identification and processing of data from different sensors by the host computer.

[0063] S3: Using calibration rules and real-time monitored environmental parameters, the characteristic parameters of the signal are dynamically calibrated to obtain accurate signal characteristic parameters.

[0064] like Fig. 2 As shown, dynamic calibration includes:

[0065] S31: Set the structure parameters of the time series integration model, and collect real-time environmental parameters.

[0066] S32: Train the time series integration model using pre-forecast data.

[0067] S321: Divide the pre-forecast data into training set one and test set one in a ratio of 8:2, or other ratios.

[0068] S322: Set the hyperparameters of the model according to the structure characteristics of the time series integration model and the characteristic parameters of the pre-forecast data, including the dimension of the hidden layer, the number of attention heads, and the learning rate. The dimension of the hidden layer determines the ability of the model to abstractly represent features. The number of attention heads affects the degree of attention paid by the model to different features and time steps. The learning rate determines the step size of parameter updates during model training, and a suitable learning rate helps the model converge faster and more stably.

[0069] S323: Use the embedder of the time series integration model to convert the training set to a low-dimensional learnable space.

[0070] S324: In the low-dimensional learnable space, the discriminator updates the hyperparameters of the current time series integration model according to the accuracy of discriminating whether the received data is from the training set one or the fake data generated by the generator, according to the gradient descent algorithm, until the comprehensive evaluation index of the time series integration model decreases continuously for 2 rounds, and the update of the hyperparameters stops. During the update of the hyperparameters, the generator continuously generates fake data to try to deceive the discriminator, and the discriminator finds out the fake data output by the generator, and they repeatedly perform adversarial training. After completing each round of adversarial training, the test set one is used to test the time series integration model to monitor the performance indicators of the model and the changes of the hyperparameters, until the time series integration model stops updating the hyperparameters.

[0071] S325: Test the time series integration model using the test set one to determine whether the trained time series integration model meets the preset threshold. If yes, use the generator of the time series integration model to generate new data and fuse it with the original data to obtain new pre-forecast data. If not, continue the adversarial training. To prevent overfitting of the integrated model, the new data set is divided according to the 5-fold cross-validation method.

[0072] S33: Input the real-time environmental parameters into the trained time series integration model to predict the current calibration parameters, convert the current calibration parameters into operable calibration rules, and dynamically adjust the signal feature parameters according to the calibration rules.

[0073] S4: Generate decoding rules suitable for the current signal using decision tree algorithm according to the calibrated signal feature parameters and combining with PSI5 protocol standard.

[0074] S41: Set the evaluation criteria for the decision tree algorithm node splitting and the maximum depth of the tree and other parameters, provide a reasonable initial framework for the operation of the algorithm, so that it can effectively carry out the learning and adjustment process towards the generation of adaptive decoding rules.

[0075] S42: Build the analog signal feature parameters and the correct decoding under the corresponding preset decoding rule into training set two and test set two, and input the training set two into the decision tree algorithm.

[0076] S43: Start the decision tree algorithm to learn and infer, so that it can generate a preliminary decoding rule based on the input data. Based on the input training data set, the best features are selected for node splitting, branch path construction and other operations, and a decision tree is gradually formed, and the categories corresponding to each leaf node are the preliminary decoding rules.

[0077] S44: Prepare a set of independent analog signal test samples with known correct decoding results, and decode these test samples according to the generated preliminary decoding rule to obtain the decoding rule.

[0078] S5: Decode the calibrated signal according to the decoding rule to obtain the effective data output by the sensor.

[0079] The specific steps of decoding the accurate analog signal according to the decoding rule are as follows:

[0080] S51: According to the definition of signal frame structure in the decoding rule, find the characteristic features representing the start of the frame in the analog signal, such as a certain amplitude, a certain frequency pulse, etc. as the frame header mark, accurately lock the frame header position through signal analysis software or algorithm, and then determine the range of the entire signal frame, and prepare for the subsequent extraction of the region where the effective data is located.

[0081] S52: Along the determined frame header, check whether the subsequent signal part is complete and meets the expectations according to the frame length, sequence of each part and other requirements specified in the decoding rule. If there is a missing or abnormal part, analyze whether it is a problem in the acquisition and transmission link or an abnormality in the signal itself, and try to solve it by re-acquiring or using signal repair algorithm, etc. to ensure the availability of the entire signal frame.

[0082] S53: According to the position and length range of the data segment in the signal frame specified in the decoding rule, accurately extract the analog signal part corresponding to the data segment from the complete signal frame, for example, the rule indicates that the data segment starts from the 8th sampling point after the frame header and has a length of 20 sampling points, which is strictly cut according to this standard.

[0083] Position-related data: It explicitly indicates the specific position information of the data segment in the entire signal frame. For example, taking a simple analog signal frame structure as an example, the signal frame contains a frame header, a data segment, a frame tail, etc., and the position data of the data segment indicates that it starts from the first sampling point of the signal frame and ends at the first sampling point. By such accurate position definition, the subsequent steps can accurately lock the target range to be extracted from the complete analog signal frame, just like in a signal frame with a length of 100 sampling points, the data segment is specified to be located between the 20th and 80th sampling points, which is the embodiment of its position-related data.

[0084] Length range data: It defines the length range of the data segment itself, which, in combination with the position information, further accurately determines the data segment. For example, it is specified that the length of the data segment is fixed at 60 sampling points, or it is variable within a certain interval, such as a minimum length of 50 sampling points and a maximum length of 70 sampling points. Length range data helps accurate extraction on the one hand, and on the other hand, it can adapt to the certain elastic changes of the data segment in different situations, ensuring that the corresponding data segment can be correctly obtained in various scenarios that meet the rules.

[0085] Encoding feature data: It describes the encoding features of the data segment in the form of analog signal, that is, how it carries and represents information at the analog signal level. For example, a specific modulation method is used to encode the original information to form this data segment, or different values are represented according to certain level change rules. These encoding feature data are important basis for subsequent decoding conversion, and only by clearly understanding the encoding characteristics of the data segment in the analog state, can it be accurately converted into the corresponding digital form.

[0086] Decoding conversion-related data: It contains specific conversion method data for how to convert the data segment from analog signal to digital signal. For example, it is specified that a specific quantization algorithm is used to convert the amplitude value of the analog signal into a digital value according to every 0.1V as a quantization unit. Or according to a certain mapping relationship, such as the analog signal level in a certain interval corresponds to a certain specific code in the digital signal, etc. These explicit conversion rule-related data guide the conversion operation of analog signal to digital signal in step S54.

[0087] Data format-related data: It concerns the data format requirements followed by the data segment in the final parsing and restoration. For example, it is specified that the converted digital signal data segment should be arranged according to a specific byte order, or should follow a certain specific protocol format for further parsing, to ensure that the final effective data can be restored as expected, for subsequent use and processing.

[0088] S54: According to the specific decoding conversion mode in the decoding rule for the data segment, the extracted analog signal data segment is converted into a corresponding digital form or other form that meets the subsequent analysis requirements. For example, if the data segment is encoded using pulse width modulation, it is converted into the corresponding digital code according to the correspondence between pulse width and specific value, completing the key data conversion step.

[0089] S55: After the signal data is verified and error processed, it is parsed and restored in detail according to the provisions of the decoding rule for data format and the meaning of each part. For example, if the rule specifies that the first few digits represent the temperature measurement value of the sensor and the middle few digits represent the humidity measurement value, the data is restored to the specific measurement value according to this correspondence to generate the valid data content output by the sensor. The valid data parsed is checked for reasonableness, such as whether the value is within a reasonable physical range. If there is unreasonable data, further investigation is conducted to determine whether the decoding process or the sensor itself is abnormal. If necessary, appropriate processing is performed.

[0090] S6: Compare the valid data obtained by decoding with the expected data. If there is a deviation, adjust the calibration rule in the dynamic calibration step and the decision tree algorithm parameter in the adaptive decoding rule generation step according to the deviation. If the error check fails, trigger the retransmission mechanism or enable the redundant data compensation.

[0091] The valid data and the expected data are compared one by one according to the corresponding dimensions. For example, in the time series, the valid temperature data collected at the same time and the expected temperature data at that time are calculated by difference, or the change trend is observed to determine whether they are consistent. Common comparison methods include calculating the difference, relative error, and mean square error. For example, the difference between the valid temperature data and the expected temperature data at each time point is calculated. If the absolute value of the difference is within a certain allowed range, the data is considered to be basically consistent. If the difference exceeds this range, there is a deviation.

[0092] When a deviation is determined, the cause of the deviation is analyzed. The causes include zero drift of the measurement device, influence of environmental factors, aging of internal system components, etc. For example, in a weighing system, if the measured weight data is always larger than the actual expected weight, it is found that the sensor has a zero drift after long-term use, causing the initial measurement reference to deviate. Based on the deviation reason, the calibration rule is modified accordingly. Still taking the weighing system as an example, if it is a zero drift problem, the sensor zero point needs to be recalibrated, that is, the related parameters about zero correction in the calibration rule are modified to ensure that the subsequent measurement is based on the correct reference, so that the valid data is closer to the expected data.

[0093] Decision tree algorithm can dynamically adjust its parameters according to the changes of input data to adapt to new situations and better achieve the expected goal. For example, in adaptive learning rate algorithm in machine learning, the learning rate is adjusted according to the error of each training, so that the model converges faster and better.

[0094] When the deviation between the effective data and the expected data is found, the size, direction and other information of the deviation are input into the decision tree algorithm as feedback. For example, in an intelligent flow control system, if it is detected that the effective data deviates greatly from the expected flow data and the actual flow is small, the decision tree algorithm will appropriately increase the related parameters of the control valve opening degree according to the degree of deviation, so that the flow can be adjusted in the expected direction. By continuously updating the algorithm parameters according to the deviation feedback, the effective data can eventually match the expected data as much as possible.

[0095] The application provides a dynamic calibration and adaptive signal decoding method based on a PSI5 interface, which uses a calibration rule to dynamically calibrate the characteristic parameters of an analog signal to decode the accurate analog signal, solves the defects of low precision, poor robustness and resource waste caused by static calibration error, fixed decoding rule and lack of feedback optimization in the prior art, and achieves the beneficial effects that:

[0096] Dynamic calibration can monitor the deviation of sensor output signal in real time, and adjust the decoding parameters through algorithm to ensure the signal accuracy. Traditional fixed calibration method is difficult to cope with complex and variable actual working conditions, while the method can obtain the original analog signal through differential signal acquisition technology, and real-time correct the characteristic parameters by using dynamic calibration rule. Differential signal acquisition can effectively suppress common mode interference and ensure the integrity of the original signal. Dynamic calibration adjusts the calibration parameters according to the real-time working condition, so that the analog signal always maintains high precision.

[0097] The adaptive decoding rule generation mechanism based on the PSI5 protocol can dynamically adjust the decoding strategy according to the current signal characteristics and channel state. As the standard interface of automobile sensor network, the PSI5 protocol provides a basis for adaptive decoding with its flexible frame structure and encoding mode. The system automatically selects the optimal decoding algorithm by analyzing the frequency, duty cycle, noise level and other parameters of the signal, combined with the preset decoding rule library.

[0098] The device embodiments described above are merely illustrative, wherein the units illustrated as separate components can or can not be physically separate, and the components illustrated as units can or can not be physical units, i.e., can be located in one place, or can be distributed to multiple network units. Part or all of the modules can be selected to achieve the purposes of the embodiments according to actual needs. Those skilled in the art can understand and implement without creative labor.

[0099] Through the description of the above embodiments, those skilled in the art can clearly understand that the embodiments can be realized by means of software and the necessary general hardware platform, and of course can also be realized by hardware. Based on such understanding, the above technical solutions can be embodied in the form of software products, and the computer software products can be stored in a computer readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and include a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods of the embodiments or some parts of the embodiments.

[0100] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit it. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement to part of the technical features. And these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for dynamic calibration and adaptive signal decoding based on a PSI5 interface, characterized in that, The method comprises the following steps: S1: collecting historical signal data and analog signals of sensor outputs under different working environmental conditions; S2: extracting features of the analog signals to obtain feature parameters; S3: using a calibration rule to dynamically calibrate the feature parameters to obtain accurate analog signals; S31: setting structure parameters of a time series integrated model and collecting real-time environmental parameters; S32: training the time series integrated model using the prehistory data; S33: inputting the real-time environmental parameters into the trained time series integrated model to predict current calibration parameters, converting the current calibration parameters into operable calibration rules, and dynamically adjusting the signal feature parameters according to the calibration rules; S4: generating decoding rules suitable for the current signal by using a decision tree algorithm according to the accurate analog signals and the PSI5 protocol standard; S5: decoding the accurate analog signals according to the decoding rules to obtain effective data; S6: comparing the effective data with expected data, adjusting the calibration rules and the decision tree algorithm parameters according to the deviation, and obtaining accurate effective data. In step S1, the analog signals are converted into digital signals through a PSI5 interface, and the digital signals contain multiple data frames, each of which contains a synchronization bit, a data bit and a check bit. In step S2, the feature parameters include the initial amplitude, frequency, phase and data frame format of the analog signals; the preprocessing includes: performing outlier processing, One-Hot encoding, time encoding, data normalization and adding Gaussian noise processing on the historical signal data.

2. The method of claim 1, wherein the PSI5 interface-based dynamic calibration and adaptive signal decoding method is characterized by, In step S32, the training of the time series integrated model comprises:

3. The method of claim 1, wherein the PSI5 interface-based dynamic calibration and adaptive signal decoding method is characterized by, S321: dividing the prehistory data into a training set one and a test set one; 4. The method of claim 1, wherein the PSI5 interface-based dynamic calibration and adaptive signal decoding method is characterized by, S322: setting hyperparameters of the current time series integrated model according to the structure characteristics of the time series integrated model and the feature parameters of the prehistory data; S323: converting the training set to a low-dimensional learnable space using an embedder of the time series integrated model; S324: in the low-dimensional learnable space, the discriminator updates the hyperparameters of the current time series integrated model according to the gradient descent algorithm until the comprehensive evaluation index of the time series integrated model decreases continuously for 2 rounds, and stops updating the hyperparameters; S325: testing the time series integrated model using the test set one until the time series integrated model obtained by training meets a preset threshold, generating new data using the generator, and fusing the new data with the original prehistory data to obtain new prehistory data. In step S324, the training of the time series integrated model comprises: in the hyperparameter updating process, the generator continuously generates false data to deceive the discriminator, the discriminator finds out the false data output by the generator, and the two repeatedly perform adversarial training; after completing each round of adversarial training, the time series integrated model is tested using the test set one to monitor the model performance index and the change of the hyperparameters, until the time series integrated model stops updating the hyperparameters. ​ 5. The method of dynamic calibration and adaptive signal decoding based on the PSI5 interface according to claim 4, characterized in that, ​ 6. The method of dynamic calibration and adaptive signal decoding based on the PSI5 interface according to claim 1, characterized in that, In step S4, the specific steps for generating the decoding rule adapted to the current signal by using the decision tree algorithm are as follows: S41: setting parameters of the decision tree algorithm according to the number of feature parameters; S42: constructing a training set two and a test set two by using the analog signal feature parameters and the correct decoding under the corresponding preset decoding rule, and inputting the training set two into the decision tree algorithm; S43: performing node splitting on the optimal feature to construct a branch road to form a decision tree, and the category corresponding to each leaf node of the decision tree is a preliminary decoding rule; S44: using a group of current analog signals with known correct decoding results to perform decoding operation on the current analog signals according to the preliminary decoding rule to obtain a current analog signal decoding rule.

7. The method of claim 1, wherein the PSI5 interface-based dynamic calibration and adaptive signal decoding method is characterized by, In step S5, the decoding of the accurate analog signal includes: S51: finding the starting frame in the accurate analog signal according to the definition of the signal frame structure in the decoding rule; S52: checking whether the accurate analog signal frame after the starting frame is complete according to the decoding rule from the position of the starting frame, and if yes, keeping unchanged, otherwise, repairing the accurate analog signal frame by using a signal repair algorithm; S53: extracting the analog signal data corresponding to the data segment from the complete accurate analog signal frame according to the position and length range of the data segment in the signal frame in the decoding rule; S54: converting the data segment of the analog signal into digital signal data according to the decoding conversion mode of the data segment in the decoding rule; S55: parsing and restoring the digital signal data according to the data format in the decoding rule to obtain valid data.

8. The method according to claim 7, wherein, In step S53, the data segment includes: Position-related data: used to indicate the specific position of the data segment in the entire signal frame; Length range data: the length range occupied by the data segment itself; Encoding feature data: the encoding feature of the data segment in the analog signal form; Decoding conversion data: the specific conversion mode for describing how the data segment is converted from the analog signal to the digital signal; Data format data: the data format requirement followed by the data segment in the final parsing and restoration.

9. The method of claim 1, wherein the PSI5 interface-based dynamic calibration and adaptive signal decoding method is characterized by, In step S6, after the adjustment, the parsed parameters are subjected to error checking, and it is judged whether the error checking result exceeds a threshold value, and if yes, the valid data is supplemented by using redundant data.

Images