Fault-tolerant method of cockpit system and micro-control unit

Abstract

The embodiment of the invention provides a fault tolerance method of a cabin system and a micro-control unit, and relates to the technical field of intelligent cabins of vehicles. The fault tolerance method of the cabin system comprises the steps of obtaining index values of multiple health indexes of a main system chip of the cabin system in a current time window, and generating an index data set corresponding to the current time window; based on the index data set corresponding to the current time window, predicting the health state of the main system chip at the prediction time point corresponding to the current time window to obtain a state prediction result; the state prediction result comprises the health probability of the prediction time point; and if the health probability of the predicted time point in the state prediction result is smaller than or equal to a first probability threshold value, performing function takeover processing on the main system chip based on the health probability of the predicted time point. Based on the fault-tolerant method of the cockpit system and the micro-control unit provided by the embodiment of the invention, the reliability and the use experience of the cockpit domain function can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent cockpit, and in particular to a fault-tolerant method of a cockpit system and a micro control unit. BACKGROUND

[0002] In related technologies, a vehicle intelligent cockpit usually integrates various functions of a cockpit domain on a central control large screen of a cockpit system, and a user can use corresponding functions through a virtual function entry on the central control large screen. At present, the cockpit system usually relies on a highly integrated main system chip (System on Chip, SoC) to provide a function carrier for functions of each function entry of the central control large screen. Once the main system chip fails, the central control large screen will be black or unresponsive, and the user cannot use the corresponding functions of the cockpit domain, which affects the user experience. SUMMARY

[0003] Embodiments of the present application provide a fault-tolerant method of a cockpit system and a micro control unit, which are used to improve the reliability of cockpit domain functions and improve the user experience of cockpit domain functions.

[0004] In a first aspect, embodiments of the present application provide a fault-tolerant method of a cockpit system, which includes: obtaining index values of multiple health indexes of a main system chip of a cockpit system in a current time window, and generating an index data set corresponding to the current time window; wherein the current time window is a time window corresponding to a current time point, the time end point of which is the current time point, and the time length thereof is a predefined length; predicting a health state of the main system chip at a prediction time point corresponding to the current time window based on the index data set corresponding to the current time window, and obtaining a state prediction result; wherein the state prediction result includes a health probability of the main system chip at the prediction time point; the prediction time point corresponding to the current time window is a time point after the current time point, and the time interval between the prediction time point and the current time point is a predefined interval; and if the health probability of the prediction time point in the state prediction result is less than or equal to a first probability threshold, performing a function takeover process on the main system chip based on the health probability of the prediction time point.

[0005] In some embodiments of the present application, the function takeover process on the main system chip based on the health probability of the prediction time point includes: determining a health level of the prediction time point based on the health probability of the prediction time point; and performing a function takeover process corresponding to the health level on the main system chip.

[0006] In some embodiments of the present application, the determination of the health level of the prediction time point based on the health probability of the prediction time point includes: if the health probability of the predicted time point is less than or equal to the first probability threshold and greater than a second probability threshold, determining that the health level of the predicted time point is a second level; if the health probability of the predicted time point is less than or equal to the second probability threshold and greater than a third probability threshold, determining that the health level of the predicted time point is a third level; and if the health probability of the predicted time point is less than or equal to the third probability threshold, determining that the health level of the predicted time point is a fourth level.

[0007] In some embodiments of the present application, the function takeover processing of the main system chip corresponding to the health level comprises: if the health level of the predicted time point is the second level, performing a function preloading operation on a first function of the main system chip; the first function comprises a body control function; if the health level of the predicted time point is the third level, performing a function takeover operation on the first function of the main system chip; and if the health level of the predicted time point is the fourth level, performing a function takeover operation on a second function of the main system chip; the second function comprises a function other than the first function.

[0008] In some embodiments of the present application, the state prediction result further comprises a feature contribution degree of each feature in a feature data set corresponding to the index data set; and the method further comprises: based on the health probability of the predicted time point and the feature contribution degree of each feature in the state prediction result, performing a fault management processing on the main system chip.

[0009] In some embodiments of the present application, the fault management processing on the main system chip based on the health probability of the predicted time point and the feature contribution degree of each feature in the state prediction result comprises: determining a health level of the prediction time point based on the health probability of the prediction time point; wherein if the health probability is less than or equal to a first probability threshold and greater than a second probability threshold, the health level is determined as a second level, and if the health probability is less than or equal to the second probability threshold and greater than a third probability threshold, the health level is determined as a third level; if the health level is the second level or the third level, matching the feature contribution degree of each feature with a fault root cause in a fault knowledge base to obtain a matching result; the matching result includes: a fault root cause matched with the feature contribution degree of each feature, and a fault maintenance strategy and a fault tolerance strategy corresponding to the fault root cause; based on the matching result, determining a potential fault root cause and a fault maintenance strategy and a fault tolerance strategy corresponding to the potential fault root cause; if the health level is the second level, performing a fault maintenance operation on the main system chip based on the fault maintenance strategy corresponding to the potential fault root cause; and if the health level is the third level, performing a fault tolerance operation on the main system chip based on the fault tolerance strategy corresponding to the potential fault root cause.

[0010] In some embodiments of the present application, the method further comprises: generating a feature data set corresponding to the current time window based on the indicator data set corresponding to the current time window and a pre-defined feature engineering; the feature data set corresponding to the current time window corresponds to the indicator data set corresponding to the current time window; inputting the feature data set corresponding to the current time window into a health state prediction model; and predicting, by the health state prediction model, the health state of the prediction time point corresponding to the current time window based on the feature data set corresponding to the current time window, outputting a health probability of the prediction time point and a feature contribution degree of each feature in the feature data set to the health probability prediction.

[0011] In some embodiments of the present application, the method further comprises: obtaining a target training data set; the target training data set includes: a plurality of indicator data sets corresponding to historical time windows and health state labels of prediction time points corresponding to each historical time window; generating a feature data set corresponding to the historical time window based on the indicator data set corresponding to the historical time window and the pre-defined feature engineering; training a gradient boosting decision tree model based on the feature data set corresponding to the historical time window and the health state label of the prediction time point corresponding to the historical time window; and performing model lightening processing on the trained model to obtain the health state prediction model.

[0012] In some embodiments of the present application, the generating of the feature data set corresponding to the historical time window based on the indicator data set corresponding to the historical time window and the pre-defined feature engineering comprises: Based on the pre-defined feature engineering, the instantaneous features, statistical features and derived features in the indicator data set are obtained; and based on the instantaneous features, statistical features and derived features in the indicator data set, the feature data set is generated.

[0013] In a second aspect, the embodiments of the present application provide a micro control unit, which is characterized in that the micro control unit is used to execute the fault-tolerant method of the cockpit system.

[0014] In the fault-tolerant method of the cockpit system and the micro control unit provided by the embodiments of the present application, the indicator values of multiple health indicators of a main system chip of the cockpit system in a current time window are obtained, an indicator data set is generated, the health status of the main system chip at a prediction time point corresponding to the current time window is predicted based on the indicator data set, a health probability of the main system chip at the prediction time point is obtained, and the function takeover processing of the main system chip is performed in a case where the health probability is less than or equal to a first probability threshold. Since the prediction time point is a time point after the current time point, the health probability of the prediction time point is the future health probability of the main system chip. When the health probability is less than the first probability threshold, the function takeover processing of the main system chip is performed, so that the function takeover of the main system chip is actively performed when it is predicted that the main system chip may be unhealthy in the future, that is, before the main system chip fails. Therefore, the active fault-tolerant processing of the main system chip is performed in advance before the main system chip fails, the change from “passive response after failure” to “prevention before failure” is realized, and the reliability and use experience of various cockpit domain functions that depend on the main system chip are improved.

[0015] The above description is only a summary of the technical solutions of the present application. In order to enable the technical means of the present application to be more clearly understood, the present application can be implemented according to the content of the description, and in order to enable the above and other purposes, features and advantages of the present application to be more obvious and easy to understand, the following specific embodiments of the present application are described. BRIEF DESCRIPTION OF DRAWINGS

[0016] Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The drawings are for purposes of illustration only and are not considered a limitation of the present application. Moreover, like reference numerals in the attached figures are intended to represent the same parts throughout the different drawings. In the drawings: Figure 1 is a schematic diagram of an application scenario of the fault-tolerant method of the cockpit system according to some embodiments of the present application; Figure 2 is a flowchart of the fault-tolerant method of the cockpit system according to some embodiments of the present application. Detailed Implementation

[0017] Exemplary embodiments of the present application will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present application are shown in the drawings, it should be understood that the present application may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this application will be thorough and complete, and will fully convey the scope of the present application to those skilled in the art.

[0018] Figure 1 This is a schematic diagram illustrating application scenarios of the fault-tolerant method for a cockpit system according to some embodiments of this application. For example... Figure 1 As shown, in some embodiments of this application, the fault-tolerant method of the cockpit system can be applied to the cockpit system 100, which is a primary / backup deployment architecture, including a primary system chip 110 and a microcontroller unit 120.

[0019] The Cockpit System 100 is an integrated in-vehicle electronic information system centered around an in-vehicle infotainment system, providing drivers and passengers with information, entertainment, comfort, and control services. The central control screen of the Cockpit System 100 can integrate various function entry points, such as entry points for in-vehicle entertainment functions and entry points for vehicle control functions, including window control, air conditioning control, seat control, ambient lighting control, etc.

[0020] The System on Chip (SoC) 110 is the system-on-a-chip of the cockpit system 100, which can provide a functional carrier for the functions corresponding to the various function entries on the central control screen.

[0021] The microcontroller unit (MCU) 120 is connected to the main system chip 110. It is a hardware module independent of the main system chip 110. It can ensure that the functions of various function entries on the central control screen continue to be available when the main system chip 110 fails or is about to fail.

[0022] The fault-tolerant method for the cockpit system provided in this application embodiment can be applied to the microcontroller unit 120 and executed by the microcontroller unit 120 to actively fault tolerate the main system chip 110 of the cockpit system 100. Figure 2 This is a flowchart illustrating a fault-tolerant method for a cockpit system according to some embodiments of this application. Figure 2 As shown, in some embodiments of this application, the fault-tolerant method for the cockpit system may include the following processing: S102: Obtain the values ​​of multiple health indicators of the main system chip of the cockpit system within the current time window, and generate the indicator dataset corresponding to the current time window.

[0023] Here, the current time window is the time window corresponding to the current time point, the time end point of the current time window is the current time point, and the time length is a predefined length. The current time window is a sliding time window, which changes as the current time point changes. For example, the current time window can be a time window with the time end point n of the current time point and a time length of N time points, i.e., [n-(N-1), n-(N-2), ..., n].

[0024] The cockpit system can be the aforementioned cockpit system 100, and the main system chip can be the aforementioned main system chip 110. Multiple health indicators are indicators related to the chip health of the main system chip, which may include multiple indicators from resource utilization, performance, and physical environment. Resource utilization indicators include, for example, CPU (Central Processing Unit) load rate, GPU (Graphics Processing Unit) load rate, and memory usage. Performance indicators include, for example, the response latency of critical tasks. Physical environment indicators include, for example, CPU temperature.

[0025] There can be multiple values ​​for a health indicator within the current time window. These values ​​can include the instantaneous values ​​of the health indicator of the main system chip at each point in time within the current time window. In other words, the values ​​of a health indicator within the current time window can be a sequence of values.

[0026] The current indicator dataset can be a dataset composed of the indicator values ​​of each health indicator among multiple health indicators, and it can be an indicator value matrix.

[0027] In implementation, the microcontroller unit can continuously collect the values ​​of multiple health indicators of the main system chip at set time intervals, such as CPU load rate, GPU load rate, memory usage, critical task response latency, and CPU temperature, to obtain the value of each health indicator. After collecting the values ​​of multiple health indicators of the main system chip at the current time point, the microcontroller unit can use a sliding time window (the current time window) with the current time point as the time endpoint and a time length of N time points to extract the values ​​of the N nearest time points from the current time point n of each health indicator, thus obtaining the value of each health indicator within the current time window. The values ​​of each health indicator within the current time window are then combined to generate the indicator dataset corresponding to the current time window.

[0028] S104, based on the indicator dataset corresponding to the current time window, predicts the health status of the main system chip at the predicted time point corresponding to the current time window, and obtains the status prediction result.

[0029] Here, the predicted time point corresponding to the current time window is the time point after the current time point, and the time interval between it and the current time point is a predefined interval. This predefined time interval is, for example, a single time interval. For example, if the current time point is time point n, and the current time window is [nN, n-(N-1), n-(N-2)……n], the predicted time point corresponding to the current time window can be time point n+1.

[0030] The state prediction result can include the health probability of the main system chip at the predicted time point corresponding to the current time window. For example, if the predicted time point corresponding to the current time window is time point n+1, the state prediction result can include the health probability of the main system chip at time point n+1.

[0031] Health probability is used to represent the likelihood of the main system chip being healthy. The higher the health probability, the greater the likelihood of it being healthy, and the lower the health probability, the less likely it is to be healthy.

[0032] In implementation, a health status prediction model can be deployed on the microcontroller unit. After obtaining the indicator dataset corresponding to the current time window, the microcontroller unit can use the health status prediction model to predict the health status of the main system chip at the predicted time point corresponding to the current time window based on the indicator dataset.

[0033] In some embodiments, the above-described process S104, based on the indicator dataset corresponding to the current time window, predicts the health status of the main system chip at the predicted time point corresponding to the current time window, and obtains the status prediction result. This process may specifically include the following steps: The indicator dataset corresponding to the current time window is input into the first health prediction model. The first health prediction model then predicts the health probability at the predicted time point corresponding to the current time window based on the indicator dataset, thus obtaining the health probability at the predicted time point.

[0034] The first health prediction model can be obtained by training the time-series prediction model based on the target training dataset. The target training dataset may include: indicator datasets corresponding to multiple historical time windows and health status labels of the indicator datasets. The time-series prediction model is, for example, an LSTM (Long Short-Term Memory) model.

[0035] Similar to the current time window, the historical time window is a time window whose time endpoint is the historical time point, and its time length is a predefined length.

[0036] The indicator dataset corresponding to the historical time window has a similar data structure to the indicator dataset corresponding to the current time window, and records the same indicators. The indicator dataset corresponding to the historical time window may include: the indicator value of each health indicator among multiple health indicators of the main system chip within the historical time window.

[0037] The health status label of the indicator dataset corresponding to the historical time window can be the health status label of the prediction time point corresponding to the historical time window, which can be healthy or unhealthy. Similar to the prediction time point corresponding to the current time window, the prediction time point corresponding to the historical time window can be a time point after the historical time point, and the time interval between the historical time point and the historical time point is a predefined interval.

[0038] In implementation, the indicator dataset corresponding to the historical time window and the corresponding health status label can be obtained by testing the main system chip of the cockpit domain controller of the test vehicle in a test environment. The test environment can include various environments, such as high-temperature environments, parallel multi-application environments, and long-term operation environments.

[0039] In the above embodiments, the first health status prediction model is deployed on the microcontroller unit and can be a lightweight model. In implementation, a time-series prediction model can be trained based on the target training dataset to obtain the first health status prediction model. Then, the health status prediction model is lightweighted to obtain a lightweight first health status prediction model, which is then deployed on the microcontroller unit. After obtaining the indicator dataset corresponding to the current time window, the microcontroller unit can use the locally deployed lightweight first health status prediction model to predict the health status of the main system chip.

[0040] The lightweight model processing can include model quantization and model pruning. Model quantization includes, for example, converting floating-point numbers in the trained model to 8-bit integers, while model pruning includes, for example, removing model nodes that have little impact on the results. When deploying the lightweight first health state prediction model to the microcontroller unit, the microcontroller unit can allocate static memory for the health state prediction process of the first health state prediction model to avoid the uncertainty and overhead caused by dynamic memory allocation.

[0041] S106, if the health probability at the predicted time point in the state prediction result is less than or equal to the first probability threshold, perform functional takeover processing on the main system chip based on the health probability at the predicted time point.

[0042] Here, the predicted time point corresponding to the current time window is a time point after the current time point, i.e., a future time point. The health probability at the predicted time point is the future health probability of the main system chip. When this health probability is less than a first probability threshold, functional takeover processing is performed on the main system chip. This allows for proactive functional takeover of the main system chip even before it fails, or when future unhealthy conditions are predicted, achieving a shift from "passive response after a fault" to "prevention before a fault."

[0043] In the above embodiments, the values ​​of multiple health indicators of the main system chip of the cockpit system within the current time window are obtained to generate an indicator dataset. Based on this indicator dataset, the health status of the main system chip at the predicted time point corresponding to the current time window is predicted to obtain the health probability of the main system chip at the predicted time point. If the health probability is less than or equal to a first probability threshold, the main system chip's functions are taken over. Since the health probability at the predicted time point is the future health probability of the main system chip, taking over the main system chip's functions when the health probability is less than the first probability threshold allows for proactive takeover of the main system chip's functions even before it fails, and proactive fault tolerance processing is performed on the main system chip before it fails. This achieves a shift from "passive response after failure" to "prevention before failure," improving the reliability and user experience of various cockpit domain functions that rely on the main system chip.

[0044] In implementation, the above-mentioned process S106 performs functional takeover processing on the main system chip based on the predicted health probability at a given time point, which may specifically include the following processes: Based on the predicted health probability at a given time point, the health level at that time point is determined; and the main system chip is then subjected to the functional takeover processing corresponding to that health level.

[0045] The function takeover process can vary depending on the health level. By performing function takeover processing on the main system chip based on its health level, hierarchical function takeover can be implemented, improving the smoothness of function switching.

[0046] In some embodiments, the health level can be divided into four levels. The above-mentioned determination of the health level at a predicted time point based on the health probability at the predicted time point may specifically include the following processing: If the predicted health probability at a given time point is greater than the first probability threshold, the health level at that time point is determined to be Level 1; if the predicted health probability at a given time point is less than or equal to the first probability threshold but greater than the second probability threshold, the health level at that time point is determined to be Level 2; if the predicted health probability at a given time point is less than or equal to the second probability threshold but greater than the third probability threshold, the health level at that time point is determined to be Level 3; if the predicted health probability at a given time point is less than or equal to the third probability threshold, the health level at that time point is determined to be Level 4.

[0047] Corresponding to the four health levels mentioned above, the main system chip can be subject to the corresponding health level's functional takeover processing.

[0048] In some embodiments, the function takeover process corresponding to the first level may be no operation. The function takeover process corresponding to the second level may be a function preloading operation on the first function. The function takeover process corresponding to the third level may be a function takeover operation on the first level. The function takeover process corresponding to the fourth level may be a function takeover operation on the second function. The above-mentioned function takeover process corresponding to the health level of the main system chip may specifically include the following processes: If the predicted health level at the time point is Level 1, no function takeover operation will be performed on the main system chip; if the predicted health level at the time point is Level 2, the first function of the main system chip will be preloaded; if the predicted health level at the time point is Level 3, the first function of the main system chip will be taken over; if the predicted health level at the time point is Level 4, the second function of the main system chip will be taken over.

[0049] The first probability threshold is greater than the second probability threshold, and the second probability threshold is greater than the third probability threshold. For example, the first probability threshold could be 0.8, the second probability threshold could be 0.6, and the third probability threshold could be 0.3. A higher health probability for the main system chip indicates a greater likelihood of it being healthy and a lower likelihood of it malfunctioning. Conversely, a lower health probability indicates a lower likelihood of it being healthy and a higher likelihood of it malfunctioning.

[0050] The first function can be a key function on the main system chip, which may include vehicle control functions that affect user safety and comfort, such as window control, air conditioning control, seat control, ambient lighting control, etc. There can be multiple first functions. The second function can be any function on the main system chip other than the first function; that is, the remaining functions on the main system chip besides the first function, such as audio functions, video functions, and other entertainment functions. There can be multiple second functions.

[0051] In implementation, while performing corresponding function takeover processing on the main system chip based on the health level of the predicted time point corresponding to the current time window, the microcontroller unit can continuously collect the index values ​​of each health indicator of the main system chip at the next time point, obtain the index dataset corresponding to the next time window of the main system chip, and continuously predict the health status of the main system chip at the corresponding predicted time point based on this index dataset. Based on the health probability at the predicted time point, it continuously determines whether to perform active function takeover operation on the main system chip. Thus, based on the real-time index data of the main system chip, the real-time health status of the main system chip is predicted, and real-time function takeover processing is performed on the main system chip according to the real-time predicted health status. Furthermore, if the detected health probability of the main system chip is greater than a first probability threshold, the microcontroller unit can perform function rollback, returning control of each function to the main system chip.

[0052] In particular, when performing function preloading or function takeover operations, operations can be performed at the function granularity to ensure the atomicity of function switching and avoid intermediate states.

[0053] For example, if the health level at the predicted time point is Level 2, when the microcontroller performs a function preloading operation on the main system chip's first function, if there are multiple first functions, the microcontroller can preload each first function sequentially according to its functional granularity. For instance, if the multiple first functions are window control, air conditioning control, and seat control, the microcontroller can perform function preloading operations sequentially for window control, air conditioning control, and seat control. Performing a function preloading operation may include: preloading the function's functional logic, pre-establishing the function's communication channel, and pre-allocating resources for the function.

[0054] For example, if the health level at the predicted time point is level three, when the microcontroller takes over the primary function of the main system chip, if there are multiple primary functions, the microcontroller can take over each primary function sequentially according to its functional granularity. For instance, if the multiple primary functions are window control, air conditioning control, and seat control, the microcontroller can take over these functions sequentially. After the takeover, the microcontroller and the main system chip can operate collaboratively. Simultaneously, the microcontroller can continuously monitor the main system chip and continuously predict its health probability.

[0055] For example, if the health level at the predicted time point is level four, when the microcontroller takes over the secondary functions of the main system chip, if there are multiple secondary functions, the microcontroller can take over each secondary function sequentially according to its functional granularity. For instance, if multiple secondary functions are remaining functions on the main system chip, such as voice interaction and video entertainment functions, the microcontroller can sequentially take over these remaining functions, thus achieving complete control over all functions on the main system chip.

[0056] In the above embodiments, based on the health level of the main system chip at the predicted time point, the main system chip is subjected to corresponding level of function takeover processing. This can be achieved by adopting a gradual switching strategy to smoothly transfer the control of various functions on the main system chip, avoiding abrupt switching that could cause a break in the user experience.

[0057] In application, the central control screen of the cockpit system can provide virtual function entry points for vehicle body control functions. The cockpit system can rely on the main system chip to provide the functional carrier for vehicle body control functions. It can respond to user operations on the corresponding vehicle body control function entry points on the central control screen, and rely on the main system chip to send corresponding vehicle body control commands to the vehicle domain controller to control the corresponding vehicle body components. In the above embodiment, the first function includes vehicle body control functions. When the health probability of the main system chip at the predicted time point is at level two, the microcontroller unit takes over the vehicle body control functions on the main system chip. Therefore, when the predicted future health level of the main system chip is low, the cockpit system can rely on the microcontroller unit to provide the functional carrier for the corresponding vehicle body control function entry points on the central control screen, ensuring the availability of the virtual function entry points for the corresponding vehicle body control functions on the central control screen. This improves the reliability of the vehicle body control functions and ensures a better user experience.

[0058] Vehicle body control functions include window control, air conditioning control, seat control, and ambient lighting control. These functions are fundamental to user safety and comfort. To further improve the reliability and availability of these fundamental functions, physical function entry points for vehicle body control can be provided in the vehicle. The microcontroller unit can communicate with these physical function entry points. The fault tolerance methods for the aforementioned cockpit system may also include the following processing: After taking over the body control function of the main system chip, it sends adjustment commands to the body domain controller based on the adjustment signals sent by the physical function entry of the body control function, instructing the body domain controller to adjust the corresponding body components.

[0059] Physical function entry points for body control functions include hardware buttons and mechanical knobs. These physical function entry points can send electro-regulation signals to the microcontroller unit in response to user adjustment operations. The microcontroller unit can then generate corresponding adjustment commands based on this electro-regulation information and send these commands to the body domain controller.

[0060] For example, if the microcontroller unit (MCU) takes over the air conditioning control function, the user can adjust the air conditioning temperature using the mechanical adjustment buttons. In response to the user's adjustment, the mechanical adjustment button can send an electrical adjustment signal to the MCU. The MCU can then generate a corresponding air conditioning adjustment command based on this signal and send it to the vehicle domain controller. The vehicle domain controller can then respond to this air conditioning adjustment command and adjust the air conditioning temperature accordingly.

[0061] In the above embodiments, by setting a physical function entry point for the vehicle body function, the microcontroller unit is connected to the physical function entry point for the vehicle body control function. The microcontroller unit sends adjustment commands to the vehicle domain controller based on the adjustment signals sent by the physical function entry point for the vehicle body control function. This can further improve the availability of the vehicle body control function, a basic function that affects user safety and comfort, and ensure the user experience of the basic function.

[0062] In the above embodiments, a first health status prediction model is obtained by training a time-series prediction model based on the target training dataset. Based on the indicator dataset corresponding to the current time window and the first health status prediction model, the health status at the predicted time point is predicted, yielding the health probability at the predicted time point. In other embodiments, other health status prediction models can also be used to predict the health status at the predicted time point and obtain the health probability at the predicted time point. For example, a second health status prediction model can be obtained by training a gradient boosting decision tree model based on the target training dataset. Based on the indicator dataset corresponding to the current time window and the second health status prediction model, the health status at the predicted time point is predicted, yielding the health probability at the predicted time point.

[0063] In some embodiments, health status prediction can be performed based on a second health status prediction model. The above process S104, based on the indicator dataset corresponding to the current time window, predicts the health status of the main system chip within the current time window to obtain a status prediction result, and may include the following processing: Based on the indicator dataset corresponding to the current time window and the predefined feature engineering, a feature dataset corresponding to the current time window is generated; the feature dataset corresponding to the current time window is input into the second health status prediction model, which predicts the health status at the prediction time point and outputs the health probability at the prediction time point.

[0064] The second health status prediction model is obtained by training the second base model on the target training dataset. The target training dataset includes: indicator datasets corresponding to multiple historical time windows and health status labels for prediction time points corresponding to each indicator dataset. The second base model includes: Gradient Boosting Decision Tree (GBDT) models, such as XGBoost (extreme gradient boosting) models or LightGBM (light gradient boosting machine) models.

[0065] In some embodiments, the fault-tolerant method for a cockpit system may further include the following processing: Obtain the target training dataset; the target training dataset includes: indicator datasets corresponding to multiple historical time windows and health status labels for the prediction time points corresponding to each historical time window; generate feature datasets corresponding to historical time windows based on the indicator datasets corresponding to historical time windows and predefined feature engineering; train the gradient boosting decision tree model based on the feature datasets corresponding to historical time windows and the health status labels for the prediction time points corresponding to historical time windows; perform model lightweighting on the trained model to obtain the second health status prediction model.

[0066] Specifically, the target training dataset may include indicator datasets corresponding to multiple historical time windows and health status labels for each indicator dataset. The historical time windows have the same duration as the current time window. Similar to the indicator dataset corresponding to the current time window, the indicator dataset corresponding to the historical time window may include the indicator value of each health indicator among multiple health indicators of the main system chip within the historical time window. The health status labels of the indicator dataset corresponding to the historical time window can be the health status labels for the predicted time point corresponding to the historical time window, and can be either healthy or unhealthy. The target training dataset can be obtained by testing the main system chip of the cockpit domain controller of a test vehicle in a test environment.

[0067] During training, predefined feature engineering can be used to convert the historical time window index dataset in the target training dataset into a corresponding feature dataset, resulting in a feature dataset labeled with health status. Based on this health status-labeled feature dataset, a gradient boosting decision tree model can be trained to obtain a second health status prediction model. Subsequently, the second health status prediction model can be lightweighted and deployed on a microcontroller unit.

[0068] Similar to the indicator dataset corresponding to the current time window, the indicator dataset corresponding to the historical time window includes: a sequence of indicator values ​​for each health indicator among multiple health indicators of the main system chip within the historical time window. This includes the indicator value of each health indicator at each time point within the historical time window, i.e., the instantaneous value of each health indicator. The indicator dataset is a collection of instantaneous features. To enable better training and prediction of the second health state prediction model, a feature dataset generated through predefined feature engineering can further include statistical features and derived features, based on the instantaneous features of the corresponding indicator dataset.

[0069] In some embodiments, the generation of the feature dataset corresponding to the historical time window based on the indicator dataset and predefined feature engineering can specifically include the following processing: Based on predefined feature engineering, instantaneous features, statistical features, and derived features in the indicator dataset corresponding to the historical time window are obtained; based on the instantaneous features, statistical features, and derived features in the indicator dataset corresponding to the historical time window, a feature dataset corresponding to the historical time window is generated.

[0070] The instantaneous features in the indicator dataset corresponding to the historical time window can be the indicator values ​​of each health indicator at each time point within the historical time window, reflecting the instantaneous state of each health indicator. The statistical features in the indicator dataset corresponding to the historical time window can be features obtained by statistically calculating the indicator values ​​of each health indicator within the historical time window, reflecting the overall behavioral pattern and distribution of each health indicator during that period; examples include mean and variance. The derived features in the indicator dataset corresponding to the historical time window can be new features calculated based on the instantaneous and statistical features of the indicator dataset, used to reveal deeper relationships or patterns in the data; examples include heat load and pressure.

[0071] For example, if multiple health indicators include CPU load rate, GPU load rate, memory usage rate, and CPU temperature, predefined feature engineering can be used to further calculate the mean, standard deviation, slope, and other statistical characteristics of each health indicator based on the indicator values ​​of each health indicator at each time point in the indicator dataset. Based on the indicator values ​​of health indicators such as CPU load rate and CPU temperature at each time point in the indicator dataset, derived features such as thermal load pressure at each time point can be calculated.

[0072] The thermal load pressure at a given time point can be equal to the product of the CPU load rate and CPU temperature at that time point. A feature dataset corresponding to the indicator dataset can be generated based on the indicator dataset and its statistical and derived features. This feature dataset can include the instantaneous features of each health indicator in the indicator dataset, as well as the calculated statistical and derived features. By generating a feature dataset containing various features such as instantaneous features, statistical features, and derived features based on predefined feature engineering, and training a gradient boosting decision tree model based on this feature dataset and the corresponding health status labels, the model can learn deeper features and improve its predictive ability.

[0073] During inference, after the microcontroller obtains the indicator dataset corresponding to the current time window of the main system chip, it can convert the indicator dataset corresponding to the current time window into the corresponding feature dataset through the same feature engineering as in the training phase. The feature dataset is then input into the second health status prediction model deployed locally by the microcontroller. The second health status prediction model predicts the health status at the prediction time point corresponding to the current time window based on the feature dataset and outputs the health probability at the prediction time point corresponding to the current time window.

[0074] In the above embodiments, the second health status prediction model is trained by a gradient boosting decision tree model, which is a gradient boosting decision tree model. During the inference process, the gradient boosting decision tree model can traverse the decision tree, record the information gain of each feature at the separation node, and normalize it into a contribution.

[0075] In some embodiments, the state prediction result may include not only the health probability of the main system chip at the predicted time point corresponding to the current time window, but also the feature contribution of each feature in the feature dataset corresponding to the indicator dataset of the current time window. When performing inference based on the second health state prediction model, the second health state prediction model can, while predicting the health probability at the predicted time point corresponding to the current time window based on the feature dataset, simultaneously calculate the feature contribution of each feature in the feature dataset to the health probability prediction, and output both the health probability of the main system chip at the predicted time point corresponding to the current time window and the feature contribution of each feature in the feature dataset to the health probability prediction.

[0076] The above process S104 predicts the health status of the main system chip in the current time window based on the indicator dataset corresponding to the current time window, and obtains the status prediction result. It may specifically include the following processes: Based on the indicator dataset corresponding to the current time window and the predefined feature engineering, a feature dataset corresponding to the current time window is generated; the feature dataset corresponding to the current time window is input into the second health status prediction model, which predicts the health status at the prediction time point and outputs the health probability at the prediction time point and the feature contribution of each feature in the feature dataset.

[0077] For example, if the features in the feature dataset include CPU load rate, CPU temperature, memory usage, and other features, the second health status prediction model can output the health probability at the predicted time point, as well as the feature contribution of CPU load rate, CPU temperature, memory usage, and other features to that health probability. For example, the feature contribution of CPU load rate might be 45%, CPU temperature 30%, memory usage 30%, and other features 10%.

[0078] When the state prediction result includes the health probability at the prediction time point and the feature contribution of each feature in the feature dataset corresponding to the indicator dataset, the fault tolerance method of the above cockpit system may also include the following processing: Based on the health probability at the predicted time point and the feature contribution of each feature in the state prediction results, fault management is performed on the main system chip.

[0079] Specifically, the microcontroller unit can determine the health level at the predicted time point based on the health probability at the predicted time point in the state prediction results; determine the potential root causes of failure based on the health level and the feature contribution of each feature in the state prediction results; and perform fault management processing on the main system chip based on the health level and the potential root causes of failure.

[0080] The determination of the health level at the predicted time point based on the health probability at the predicted time point in the state prediction results can specifically include the following processing: If the predicted health probability at a given time point is greater than a first probability threshold, the health level at that time point can be determined as Level 1; if the predicted health probability at a given time point is less than or equal to the first probability threshold and greater than a second probability threshold, the health level at that time point is determined as Level 2; if the predicted health probability at a given time point is less than or equal to the second probability threshold and greater than a third probability threshold, the health level at that time point is determined as Level 3; if the predicted health probability at a given time point is less than or equal to the third probability threshold, the health level at that time point is determined as Level 4.

[0081] The determination of potential root causes of failures based on the feature contribution of each feature in the health level and state prediction results can specifically include the following processing: If the health level is level two or level three, the feature contribution of each feature is matched with the root causes of failure in the fault knowledge base to obtain the matching result; the matching result includes: the root causes of failure that match the feature contribution of each feature, and the fault maintenance strategy and fault tolerance strategy corresponding to the root causes of failure; based on the matching result, potential root causes of failure and the fault maintenance strategy and fault tolerance strategy corresponding to the potential root causes of failure are determined.

[0082] The fault knowledge base can be mined from historical data and includes mapping relationships between various feature contribution patterns and potential fault root causes, as well as corresponding fault maintenance and fault tolerance strategies. The fault knowledge base can be deployed in the cloud or on a microcontroller unit. When matching the feature contribution degree of each feature in the state prediction result with the fault root causes in the fault knowledge base, the similarity between the feature contribution degree of each feature in the state prediction result and the feature contribution degree in the feature contribution pattern of the fault knowledge base can be calculated. If the similarity is greater than a similarity threshold, the fault root cause corresponding to that feature contribution pattern can be identified as the fault root cause matching the feature contribution degree of each feature in the state prediction result.

[0083] When determining potential root causes of failures and corresponding fault maintenance and fault tolerance strategies based on matching results, the matched root causes of failures can be identified as potential root causes of failures, the fault maintenance strategies corresponding to the matched root causes of failures can be identified as fault maintenance strategies corresponding to potential root causes of failures, and the fault tolerance strategies corresponding to the matched root causes of failures can be identified as fault tolerance strategies corresponding to potential root causes of failures.

[0084] The above-mentioned fault management processing of the main system chip based on health level and potential root causes of failure can specifically include the following processes: If the predicted health level at the time point is Level 2, fault maintenance operations are performed on the main system chip based on the fault maintenance strategy corresponding to the potential root cause of the fault. If the predicted health level at the time point is Level 3, fault tolerance operations are performed on the main system chip based on the fault tolerance strategy corresponding to the potential root cause of the fault.

[0085] Among them, fault maintenance operations can correct corresponding potential faults, i.e., corrective fault maintenance operations. Fault tolerance operations are operations that ensure the availability of basic functions when corresponding potential faults exist.

[0086] For example, if the predicted health level at a given time is Level 2, and the potential root cause is a memory leak, the main system chip's primary functions can be pre-loaded. Based on the memory leak fault maintenance strategy, the microcontroller unit can send application cleanup and restart commands to the main system chip, causing it to clean up and restart certain functions, thus performing corrective maintenance for potential memory leaks. If the predicted health level at a given time is Level 2, and the potential root cause is poor heat dissipation, the microcontroller unit can send a cooling command to the cockpit domain controller's thermal management module. In response to this command, the thermal management model can increase the cockpit domain controller's fan speed, thereby performing corrective maintenance for potential heat dissipation problems.

[0087] If the predicted health level at the specified time is Level 3, and the potential root cause of the failure is GPU overload, the microcontroller unit (MCU) can take over the primary function of the main system chip (MSC) and, based on the fault tolerance strategy for GPU overload, send a rendering degradation command to the MSC. In response to this rendering degradation command, the MSC can proactively disable anti-aliasing and reduce the rendering resolution to perform rendering degradation. If the predicted health level at the specified time is Level 3, and the potential root cause of the failure is CPU overload, the MCU can take over the primary function and, based on the fault tolerance strategy for CPU overload, switch the taken-over functional logic to simpler functional logic, thereby performing functional degradation for the taken-over function.

[0088] In the above embodiments, by further obtaining the feature contribution of each feature, determining the potential root cause of failure, and further performing fault management processing based on the potential root cause of failure, the cockpit system's risk resistance capability and the reliability of cockpit domain functions can be further improved.

[0089] Furthermore, this application also provides a microcontroller unit applied to a cockpit system for performing the following processes: The system acquires the values ​​of multiple health indicators of the main system chip of the cockpit system within the current time window, generating an indicator dataset corresponding to the current time window. The current time window is the time window corresponding to the current time point, with the current time point as its end point and a predefined time length. Based on the indicator dataset corresponding to the current time window, the health status of the main system chip at the predicted time point corresponding to the current time window is predicted, yielding a status prediction result. The status prediction result includes: the health probability of the main system chip at the predicted time point; the predicted time point corresponding to the current time window is a time point after the current time point, with a predefined time interval between it and the current time point; if the health probability at the predicted time point in the status prediction result is less than or equal to a first probability threshold, functional takeover processing is performed on the main system chip based on the health probability at the predicted time point.

[0090] In some embodiments of this application, the microcontroller is specifically used to perform the following processes: Based on the health probability at the predicted time point, the health level at the predicted time point is determined; and the main system chip is subjected to the function takeover processing corresponding to the health level.

[0091] In some embodiments of this application, the microcontroller is specifically used to perform the following processes: If the health probability at the predicted time point is less than or equal to the first probability threshold and greater than the second probability threshold, the health level at the predicted time point is determined to be the second level; if the health probability at the predicted time point is less than or equal to the second probability threshold and greater than the third probability threshold, the health level at the predicted time point is determined to be the third level; if the health probability at the predicted time point is less than or equal to the third probability threshold, the health level at the predicted time point is determined to be the fourth level.

[0092] In some embodiments of this application, the microcontroller is specifically used to perform the following processes: If the health level at the predicted time point is Level 2, a function preloading operation is performed on the first function of the main system chip; the first function includes: vehicle body control function; if the health level at the predicted time point is Level 3, a function takeover operation is performed on the first function of the main system chip; if the health level at the predicted time point is Level 4, a function takeover operation is performed on the second function of the main system chip; the second function includes: functions other than the first function.

[0093] In some embodiments of this application, the state prediction result further includes: the feature contribution of each feature in the feature dataset corresponding to the index dataset; the microcontroller unit is also used to perform the following processing: Based on the health probability at the predicted time point and the feature contribution of each feature in the state prediction results, fault management processing is performed on the main system chip.

[0094] In some embodiments of this application, the microcontroller is specifically used to perform the following processes: Based on the health probability at the predicted time point, the health level at the predicted time point is determined; wherein, if the health probability is less than or equal to a first probability threshold and greater than a second probability threshold, the health level is determined to be level two; if the health probability is less than or equal to the second probability threshold and greater than a third probability threshold, the health level is determined to be level three; if the health level is level two or level three, the feature contribution of each feature is matched with the root causes of faults in the fault knowledge base to obtain a matching result; the matching result includes: the root causes of faults matched with the feature contribution of each feature, and the fault maintenance strategy and fault tolerance strategy corresponding to the root causes of faults; based on the matching result, potential root causes of faults and the fault maintenance strategy and fault tolerance strategy corresponding to the potential root causes of faults are determined; if the health level is level two, fault maintenance operations are performed on the main system chip based on the fault maintenance strategy corresponding to the potential root causes of faults; if the health level is level three, fault tolerance operations are performed on the main system chip based on the fault tolerance strategy corresponding to the potential root causes of faults.

[0095] In some embodiments of this application, the microcontroller is specifically used to perform the following processes: Based on the indicator dataset corresponding to the current time window and the predefined feature engineering, a feature dataset corresponding to the current time window is generated; the feature dataset corresponding to the current time window corresponds to the indicator dataset corresponding to the current time window; the feature dataset corresponding to the current time window is input into the health status prediction model, and the health status prediction model predicts the health status at the prediction time point corresponding to the current time window based on the feature dataset corresponding to the current time window, and outputs the health probability at the prediction time point and the feature contribution of each feature in the feature dataset to the prediction of the health probability.

[0096] In some embodiments of this application, the microcontroller unit is also configured to perform the following processes: Obtain the target training dataset; the target training dataset includes: indicator datasets corresponding to multiple historical time windows and health status labels for the predicted time points corresponding to each historical time window; Based on the indicator dataset corresponding to the historical time window and the predefined feature engineering, a feature dataset corresponding to the historical time window is generated; The gradient boosting decision tree model is trained based on the feature dataset corresponding to the historical time window and the health status label of the predicted time point corresponding to the historical time window. The trained model is then subjected to model lightweighting to obtain the health status prediction model.

[0097] In some embodiments of this application, the microcontroller is specifically used to perform the following processes: Based on the predefined feature engineering, instantaneous features, statistical features, and derived features in the indicator dataset are obtained; based on the instantaneous features, statistical features, and derived features in the indicator dataset, the feature dataset is generated.

[0098] When using an integrated unit, the microcontroller unit may include a processing module and a storage module. The processing module can be used to control and manage the actions of the environmental sensing device. The storage module can store program code and data used to support the execution of corresponding actions by the environmental sensing device.

[0099] In some embodiments of this application, the microcontroller unit may include: a memory and a processor, wherein a computer program is stored in the memory, and the processor is used to execute the computer program to implement the above. Figure 2 The processing shown.

[0100] Furthermore, embodiments of this application also provide a computer-readable storage medium (including but not limited to disk storage, CD-ROM, optical storage, etc.) storing computer program code. When the computer program code is executed on a computer, it causes the computer to perform the aforementioned related methods to achieve the above-mentioned... Figure 2 The processing shown.

[0101] Furthermore, this application also provides a computer program product that, when run on a computer, causes the computer to perform the aforementioned related processes to achieve the above-mentioned goals. Figure 2 The processing shown.

[0102] In addition, this application also provides a cockpit system including the aforementioned microcontroller unit.

[0103] In addition, this application also provides a vehicle including the above-described cockpit system.

[0104] The beneficial effects of the above embodiments can be referred to the beneficial effects of the corresponding methods provided above, and will not be repeated here.

[0105] Through the above description of the embodiments, those skilled in the art will understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0106] In the embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0107] In the description of this application, it should be understood that if the terms "upper", "lower", "front", "rear", "left" and "right" are used to indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the position or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this application.

[0108] It should be noted that, in this application, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0109] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A fault-tolerant method for a cockpit system, characterized in that, The method includes: The system obtains the values ​​of multiple health indicators of the main system chip of the cockpit system within the current time window and generates an indicator dataset corresponding to the current time window; wherein, the current time window is the time window corresponding to the current time point, the time end point is the current time point, and the time length is a predefined length. Based on the indicator dataset corresponding to the current time window, the health status of the main system chip at the predicted time point corresponding to the current time window is predicted to obtain a status prediction result; wherein, the status prediction result includes: the health probability of the main system chip at the predicted time point; the predicted time point corresponding to the current time window is a time point after the current time point, and the time interval between it and the current time point is a predefined interval; If the health probability at the predicted time point in the state prediction result is less than or equal to the first probability threshold, the main system chip is subjected to functional takeover processing based on the health probability at the predicted time point.

2. The method according to claim 1, characterized in that, The process of taking over the main system chip based on the predicted health probability at the predicted time point includes: Based on the health probability at the predicted time point, the health level at the predicted time point is determined; The main system chip is then subjected to functional takeover processing corresponding to the health level.

3. The method according to claim 2, characterized in that, Determining the health level at the predicted time point based on the predicted health probability includes: If the health probability at the predicted time point is less than or equal to the first probability threshold and greater than the second probability threshold, the health level at the predicted time point is determined to be the second level. If the health probability at the predicted time point is less than or equal to the second probability threshold and greater than the third probability threshold, the health level at the predicted time point is determined to be the third level. If the health probability at the predicted time point is less than or equal to the third probability threshold, the health level at the predicted time point is determined to be the fourth level.

4. The method according to claim 3, characterized in that, The step of taking over the functions corresponding to the health level of the main system chip includes: If the health level at the predicted time point is level two, a function preloading operation is performed on the first function of the main system chip; the first function includes: vehicle body control function; If the health level at the predicted time point is level three, a function takeover operation is performed on the first function of the main system chip. If the health level at the predicted time point is level four, a function takeover operation is performed on the second function of the main system chip; the second function includes: functions other than the first function.

5. The method according to claim 1, characterized in that, The state prediction result further includes: the feature contribution of each feature in the feature dataset corresponding to the indicator dataset; the method further includes: Based on the health probability at the predicted time point and the feature contribution of each feature in the state prediction results, fault management processing is performed on the main system chip.

6. The method according to claim 5, characterized in that, The fault management process for the main system chip, based on the health probability at the predicted time point and the feature contribution of each feature in the state prediction result, includes: Based on the health probability at the predicted time point, the health level at the predicted time point is determined; wherein, if the health probability is less than or equal to a first probability threshold and greater than a second probability threshold, the health level is determined to be a second level; if the health probability is less than or equal to a second probability threshold and greater than a third probability threshold, the health level is determined to be a third level. If the health level is level two or level three, the feature contribution of each feature is matched with the root causes of failure in the fault knowledge base to obtain a matching result; the matching result includes: the root causes of failure that match the feature contribution of each feature, as well as the fault maintenance strategy and fault tolerance strategy corresponding to the root causes of failure. Based on the matching results, potential root causes of failures and corresponding fault maintenance and fault tolerance strategies are determined. If the health level is the second level, the main system chip is subjected to fault maintenance operation based on the fault maintenance strategy corresponding to the potential root cause of the fault. If the health level is level three, the main system chip is subjected to fault tolerance operation based on the fault tolerance strategy corresponding to the potential root cause of the fault.

7. The method according to claim 5, characterized in that, The method of predicting the health status of the main system chip at the predicted time point corresponding to the current time window based on the indicator dataset corresponding to the current time window, and obtaining the status prediction result, includes: Based on the indicator dataset corresponding to the current time window and the predefined feature engineering, a feature dataset corresponding to the current time window is generated; the feature dataset corresponding to the current time window corresponds to the indicator dataset corresponding to the current time window. The feature dataset corresponding to the current time window is input into the health status prediction model. The health status prediction model predicts the health status at the prediction time point corresponding to the current time window based on the feature dataset corresponding to the current time window, and outputs the health probability at the prediction time point and the feature contribution of each feature in the feature dataset to the prediction of the health probability.

8. The method according to claim 7, characterized in that, The method further includes: Obtain the target training dataset; the target training dataset includes: indicator datasets corresponding to multiple historical time windows and health status labels for the predicted time points corresponding to each historical time window; Based on the indicator dataset corresponding to the historical time window and the predefined feature engineering, a feature dataset corresponding to the historical time window is generated; The gradient boosting decision tree model is trained based on the feature dataset corresponding to the historical time window and the health status label of the predicted time point corresponding to the historical time window. The trained model is then subjected to model lightweighting to obtain the health status prediction model.

9. The method according to claim 8, characterized in that, Based on the indicator dataset corresponding to the historical time window and the predefined feature engineering, a feature dataset corresponding to the historical time window is generated, including: Based on the predefined feature engineering, instantaneous features, statistical features, and derived features in the indicator dataset are obtained; The feature dataset is generated based on the instantaneous features, statistical features, and derived features in the indicator dataset.

10. A microcontroller unit, characterized in that, The microcontroller unit is used to perform the method according to any one of claims 1 to 9.

Images