A vehicle display instrument control system

Abstract

The application relates to the technical field of intelligent cockpits and function safety, in particular to a display instrument control system for vehicles, which comprises a forward-looking perception unit for generating multi-dimensional cockpit system state data; a risk prediction unit for calculating a system risk potential index; a scheduling decision unit for comparing and analyzing the system risk potential index with a preset risk threshold value; when the system risk potential index exceeds the risk threshold value, it is determined that the system enters a collaborative mode and a collaborative distribution factor is calculated; when the system risk potential index does not exceed the risk threshold value, it is determined that the system maintains an isolated mode; a rendering adjustment unit for generating an adjusted information entertainment domain target rendering load index; a closed-loop optimization unit for adaptively correcting the weight of a risk potential state quantitative prediction model; the application effectively solves the safety function display delay hidden danger caused by the resource competition of a graphics processing unit, and greatly improves the function safety and reliability of the intelligent cockpit.

Description

Technical Field

[0001] This invention relates to the field of intelligent cockpit and functional safety technology, specifically to a vehicle display instrument control system. Background Technology

[0002] In automotive intelligent cockpit systems, the instrument safety domain and the infotainment domain often share the same graphics processing unit (GPU) resources. This architecture can lead to GPU resource contention when both domains experience high load demands simultaneously, such as when rendering complex entertainment information while simultaneously displaying urgent Advanced Driver Assistance Systems (ADAS) warnings.

[0003] This resource competition may cause delays in the rendering of safety information used to warn of dangers, preventing it from being presented to the driver in a timely manner and thus posing a serious functional safety risk. Existing technologies often lack forward-looking prediction and dynamic avoidance mechanisms for such risks. The system cannot comprehensively consider multi-dimensional information such as vehicle dynamics, driver assistance status, and system load to predict the possibility of impending resource conflicts. Therefore, when conflicts occur, the system cannot intervene in advance and proactively adjust resource allocation strategies to prioritize the display of safety functions. This leaves the entire intelligent cockpit system with the potential for safety function failure due to display delays. Summary of the Invention

[0004] To solve the above-mentioned technical problems, the present invention provides a vehicle display instrument control system. Specifically, the technical solution of the present invention includes:

[0005] The forward-looking perception unit is used to acquire the warning level of the advanced driver assistance system and vehicle dynamic parameters, and to acquire the rendering task type and graphics processing unit load of the infotainment domain in order to generate multi-dimensional cockpit system status data.

[0006] The risk prediction unit is used to calculate the system risk potential index based on multi-dimensional cockpit system status data and a preset risk potential quantification prediction model.

[0007] The scheduling decision unit is used to compare and analyze the system risk potential index with the preset risk threshold; when the system risk potential index exceeds the risk threshold, it determines that the system enters the cooperative mode and calculates the cooperative allocation factor; when the system risk potential index does not exceed the risk threshold, it determines that the system maintains the isolation mode.

[0008] The rendering adjustment unit is used to respond to the cooperative mode and dynamically adjust the rendering task complexity of the infotainment domain according to the cooperative allocation factor to generate the adjusted target rendering load index of the infotainment domain.

[0009] The closed-loop optimization unit monitors the rendering latency of the safety domain and compares it with the maximum allowable latency required by functional safety requirements to generate a latency deviation. The weights of the risk potential quantification prediction model are then adaptively adjusted based on the latency deviation.

[0010] Furthermore, the process by which the risk prediction unit calculates the system risk potential index is as follows:

[0011] The warning levels of the advanced driver assistance system are quantified to obtain quantified warning level values; the vehicle dynamic instability is calculated based on vehicle dynamic parameters; the normalized rendering load index of the infotainment domain is obtained; and the quantified warning level values, vehicle dynamic instability, and normalized rendering load index are linearly weighted and summed to generate a system risk potential index.

[0012] Furthermore, the calculation process for vehicle dynamic instability is as follows:

[0013] Obtain the current speed and current acceleration / deceleration; normalize the current speed according to the maximum design speed to obtain the normalized current speed; normalize the current acceleration / deceleration according to the maximum acceleration / deceleration to obtain the normalized current acceleration / deceleration; and perform a weighted summation of the normalized current speed and normalized current acceleration / deceleration to generate the vehicle dynamic instability.

[0014] Furthermore, the process by which the scheduling decision unit calculates the collaborative allocation factor is as follows:

[0015] The risk difference is calculated by subtracting the system risk potential index from the risk threshold. Based on the risk difference and the preset slope control parameters, the collaborative allocation factor is calculated using the logistic function model.

[0016] Furthermore, the process by which the rendering adjustment unit generates the adjusted target rendering load index for the infotainment domain is as follows:

[0017] Obtain the standard load index and the load index of the preset rendering task at the highest quality and the lowest acceptable quality; and based on the collaborative allocation factor, perform linear interpolation between the standard load index and the load index at the lowest acceptable quality to obtain the adjusted target rendering load index for the infotainment domain.

[0018] Furthermore, the adaptive correction process of the closed-loop optimization unit is as follows:

[0019] Continuously monitor the rendering latency of keyframes in the safety domain; compare the rendering latency with the maximum allowable latency to calculate the normalized latency deviation; and based on the normalized latency deviation, use the gradient descent algorithm to update the weights in the risk potential quantification prediction model.

[0020] Furthermore, the calculation process for the normalized delay bias is as follows:

[0021] Obtain the difference between the rendering latency and the maximum allowable latency; compare the difference with zero and take the larger value as the numerator; divide the numerator by the maximum allowable latency to obtain the normalized latency bias.

[0022] Furthermore, the weight update process is as follows:

[0023] Obtain the normalized delay bias, the preset learning rate, and the actual values ​​of the input parameters of the risk potential quantification prediction model corresponding to the moment that caused the rendering delay; multiply the normalized delay bias, the learning rate, and the actual values ​​of the input parameters to obtain the weight update amount; and subtract the weight update amount from the current weight value to obtain the updated weight.

[0024] Compared with the prior art, the present invention has the following beneficial effects:

[0025] 1. This system can integrate multi-dimensional information such as advanced driver assistance systems, vehicle dynamics, and system load to achieve forward-looking prediction of resource competition risks. It can identify and avoid conflicts that may cause delays in safety information display in advance, transforming safety assurance from post-event response to pre-event prevention, and significantly improving driving safety.

[0026] 2. Based on the real-time quantified risk index, this system dynamically and smoothly adjusts the rendering complexity of the infotainment domain. This refined resource scheduling strategy can not only prioritize the display of security information without delay when necessary, but also minimize the interference with the user experience of the entertainment system, thus achieving an effective balance between security and user experience.

[0027] 3. By constructing a complete control link from risk perception and quantitative prediction to dynamic decision-making, this system effectively solves the hidden danger of safety function display delay caused by the competition for graphics processing unit resources. It ensures that key safety warning information can be presented to the driver in a timely and reliable manner in any driving scenario, greatly improving the functional safety and reliability of the intelligent cockpit.

[0028] 4. This system has closed-loop optimization capabilities, which can continuously monitor the actual rendering performance of the safety domain and feed back latency deviations to the risk prediction model. This adaptive correction mechanism enables the system to continuously learn and evolve during use, dynamically optimizing its prediction accuracy, thereby ensuring that the safety strategy remains efficient and robust throughout the entire vehicle lifecycle. Attached Figure Description

[0029] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0030] Figure 1 This is a structural diagram of the system of the present invention. Detailed Implementation

[0031] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0032] Example 1:

[0033] Please see Figure 1 A vehicle display instrument control system, comprising:

[0034] The forward-looking perception unit is used to acquire the warning level of the advanced driver assistance system and vehicle dynamic parameters, and to acquire the rendering task type and graphics processing unit load of the infotainment domain in order to generate multi-dimensional cockpit system status data.

[0035] The risk prediction unit is used to calculate the system risk potential index based on multi-dimensional cockpit system status data and a preset risk potential quantification prediction model.

[0036] The scheduling decision unit is used to compare and analyze the system risk potential index with the preset risk threshold; when the system risk potential index exceeds the risk threshold, it determines that the system enters the cooperative mode and calculates the cooperative allocation factor; when the system risk potential index does not exceed the risk threshold, it determines that the system maintains the isolation mode.

[0037] The rendering adjustment unit is used to respond to the cooperative mode and dynamically adjust the rendering task complexity of the infotainment domain according to the cooperative allocation factor to generate the adjusted target rendering load index of the infotainment domain.

[0038] The closed-loop optimization unit monitors the rendering latency of the safety domain and compares it with the maximum allowable latency required by functional safety requirements to generate a latency deviation. The weights of the risk potential quantification prediction model are then adaptively adjusted based on the latency deviation.

[0039] This embodiment provides a vehicle display instrument control system;

[0040] The system includes a forward-looking perception unit, a risk prediction unit, a scheduling decision-making unit, a rendering and adjustment unit, and a closed-loop optimization unit;

[0041] The purpose of the forward-looking perception unit is to comprehensively and in real-time collect multi-dimensional data required to build a risk assessment model, providing accurate input for subsequent risk prediction. In this embodiment, the unit obtains the warning level of the advanced driver assistance system and vehicle dynamic parameters in real time through the vehicle's controller area network CAN bus. At the same time, the unit obtains the rendering task type and graphics processing unit load of the infotainment domain through the application processor interface. The above data are fused and processed to generate multi-dimensional cockpit system status data.

[0042] In addition, the forward-looking sensing unit also includes a mechanism for validating the input data and handling outliers. For example, when the parameter value obtained from the CAN bus or application processor interface exceeds the preset reasonable physical range, such as when there is a maximum acceleration when the speed is 0 or when it fails to update within a certain time, the system will use the valid value of the previous cycle or the preset safe default value to ensure that the input of the risk prediction model is always stable and effective, and to prevent the system decision from fluctuating drastically due to sensor momentary failure or data communication interruption, thereby enhancing the robustness of the entire control system.

[0043] The risk prediction unit aims to proactively quantify the risk of a dual-peak load conflict in the graphics processing unit based on perceived multidimensional data. In this embodiment, the unit incorporates a risk potential quantification prediction model. This model receives multidimensional cockpit system state data generated by the proactive sensing unit, calculates and outputs a dimensionless scalar, namely the system risk potential index.

[0044] The scheduling decision unit aims to make core decisions regarding system resource scheduling based on quantified risk levels. In this embodiment, the unit compares the system risk potential index calculated by the risk prediction unit with a preset risk threshold. Perform real-time comparative analysis; risk threshold The settings are based on a balance between the vehicle functional safety level ASIL requirements and user experience, and are derived through offline simulation and real-world testing calibration. Specifically, during the calibration process, a cost function can be defined. This function imposes a high penalty on any situation exceeding the maximum permissible delay for functional safety, while imposing a smaller penalty proportional to the magnitude of the degradation in the infotainment domain rendering quality. By iteratively searching through a large number of simulation scenarios, the function that minimizes the total cost function is found. Value, a non-restrictive example threshold It can be set to 0.75. When the system risk potential index exceeds this risk threshold, the scheduling decision unit determines that the system should switch from the default isolation mode to the cooperative mode and calculates a cooperative allocation factor to guide specific resource adjustments. When the system risk potential index does not exceed the risk threshold, the system should maintain the isolation mode.

[0045] The rendering adjustment unit aims to execute instructions from the scheduling decision unit and dynamically adjust the resource consumption of the infotainment domain. In this embodiment, in response to the system entering a collaborative mode, this unit actively reduces the demand for graphics processing unit resources by lowering the quantifiable dimensions of the infotainment domain's rendering tasks, based on a collaborative allocation factor. This dynamically adjusts the complexity of the infotainment domain's rendering tasks. Specifically, the quantifiable dimensions of the rendering tasks may include: rendering resolution, model texture quality, lighting and shadow effects level, particle effect quantity, or animation refresh rate. For example, reducing complexity can be manifested by lowering the rendering resolution of the navigation map from 1080p to 720p, or disabling real-time lighting and shadow reflection effects in weather animations. The ultimate goal of the adjustment is to generate an adjusted target rendering load index for the infotainment domain. The rendering engine will then execute a degradation strategy based on this index, thereby prioritizing the use of the saved graphics processing unit resources by the security domain.

[0046] The purpose of the closed-loop optimization unit is to monitor the actual operation effect of the entire collaborative scheduling strategy and continuously optimize the initial risk prediction model through adaptive learning. In this embodiment, the unit continuously monitors the rendering latency of the safety domain during system operation and compares it with the maximum allowable latency required by functional safety requirements to generate a latency deviation. Based on this latency deviation, the unit adaptively corrects the weights of the risk potential quantification prediction model, thereby forming a complete closed-loop control.

[0047] Through the collaborative work of the aforementioned units, this invention constructs a complete control system encompassing risk perception, quantitative prediction, dynamic decision-making, and closed-loop optimization. It can proactively identify and mitigate the risk of delayed display of safety functions due to resource competition while ensuring the smooth operation of the infotainment system, thereby greatly enhancing the functional safety and reliability of the intelligent cockpit system.

[0048] Example 2:

[0049] The process by which the risk prediction unit calculates the system risk potential index is as follows:

[0050] The warning levels of the advanced driver assistance system are quantified to obtain quantified warning level values; the vehicle dynamic instability is calculated based on vehicle dynamic parameters; the normalized rendering load index of the infotainment domain is obtained; and the quantified warning level values, vehicle dynamic instability, and normalized rendering load index are linearly weighted and summed to generate a system risk potential index.

[0051] To further clarify the process of calculating the system risk potential index by the risk prediction unit, this embodiment discloses a linear weighted summation model that integrates multiple heterogeneous physical quantities into a single risk measure; the inputs of each dimension are standardized and then weighted and fused to obtain the final risk measure;

[0052] The calculation process involves processing the input parameters:

[0053] The warning levels of advanced driver assistance systems are quantified to obtain quantified warning level values. This value is a value in The standardized values ​​within the range are derived by mapping discrete, event-driven advanced driver assistance system states. For example, a mapping table can be established: no warning state is mapped to 0; general warnings, such as blind spot monitoring warnings, are mapped to 0.3; more urgent warnings, such as lane departure warning (LDW), are mapped to 0.7; and the most urgent warnings, such as forward collision warning (FCW) or automatic emergency braking (AEB) activation, are mapped to 1.0.

[0054] The vehicle's dynamic instability is calculated based on the vehicle's dynamic parameters. ;

[0055] Obtain the normalized rendering load index of the infotainment domain This value is a dimensionless value, derived from normalizing the real-time load of the graphics processing unit relative to its maximum theoretical load.

[0056] Based on the parameters processed above, the model quantifies the warning level. Vehicle dynamic instability and normalized rendering load index Perform a linear weighted summation to generate the system risk potential index. The calculation formula is as follows:

[0057]

[0058] in, : The systemic risk potential index at any given moment is dimensionless.

[0059] : The quantitative value of the warning level of the advanced driver assistance system at any time is dimensionless and its source is the real-time collection and quantification of the forward sensing unit;

[0060] : The vehicle dynamic instability at time t is dimensionless and its source is the calculation in subsequent steps;

[0061] : The normalized rendering load index of the infotainment domain at any given moment is dimensionless and originates from the real-time acquisition and normalization of the forward sensing unit.

[0062] These are the preset weight coefficients corresponding to the three parameters mentioned above, dimensionless. These weight coefficients are derived from offline training and calibration using a large amount of actual road test and simulation datasets. To ensure variable independence during the calibration process, this calibration dataset contains a series of system state snapshots and corresponding annotations of the probability of occurrence of dual-peak load scenarios. The contribution of each input parameter to the risk is fitted using regression analysis to determine the weight values. For example, a multiple linear regression model can be used, trained on a snapshot dataset containing thousands of typical driving scenarios. Each snapshot in this dataset is labeled with whether a dual-peak load conflict has occurred or is about to occur, marked as 1 or 0. By minimizing the prediction error, each weight value can be fitted. A set of non-restrictive example weight values ​​could be: , , ;

[0063] This formula ensures strict consistency of physical dimensions on both sides of the equation. This model successfully integrates information from three different dimensions—driving safety, vehicle dynamics, and software load—into a unified, forward-looking system risk potential index, providing a solid and reliable basis for subsequent scheduling decisions. It should be noted that the linear weighted model used in this embodiment is an effective simplification and engineering approximation of complex real-world problems. In actual scenarios, there may be more complex nonlinear coupling relationships between various risk factors. However, through adaptive correction of the weights by the subsequent closed-loop optimization unit, this model can continuously iterate in a dynamic environment, constantly approaching the optimal risk assessment effect, achieving a good balance between computational efficiency and prediction accuracy.

[0064] Example 3:

[0065] The calculation process for vehicle dynamic instability is as follows:

[0066] Obtain the current speed and current acceleration / deceleration; normalize the current speed according to the maximum design speed to obtain the normalized current speed; normalize the current acceleration / deceleration according to the maximum acceleration / deceleration to obtain the normalized current acceleration / deceleration; and perform a weighted summation of the normalized current speed and normalized current acceleration / deceleration to generate the vehicle dynamic instability.

[0067] This embodiment illustrates vehicle dynamic instability. The calculation process aims to more accurately reflect driving intentions and potential risks.

[0068] This calculation process requires acquiring and processing basic vehicle dynamic parameters:

[0069] The current speed and current acceleration / deceleration are obtained from the vehicle's CAN bus.

[0070] The current rate is normalized based on the maximum design speed to obtain the normalized current rate. ;

[0071] The current acceleration / deceleration is normalized based on the maximum acceleration / deceleration to obtain the normalized current acceleration / deceleration. ;

[0072] The model normalizes the current rate. With normalized current acceleration / deceleration A weighted summation is performed to generate the vehicle's dynamic instability. The calculation formula is as follows:

[0073]

[0074] in, : The vehicle's dynamic instability at any given time is dimensionless.

[0075] : The normalized current rate at time step is dimensionless and is obtained by normalizing the current rate.

[0076] : The normalized current acceleration / deceleration at any given moment is dimensionless and is obtained by normalizing the current acceleration / deceleration.

[0077] : These are the preset weights corresponding to the normalized rate and normalized acceleration / deceleration, respectively. They are dimensionless, and their determination method is similar to the aforementioned calibration process for weight coefficients. The calibration process is similar to that described above. A set of non-limiting example weight values ​​could be: , ;

[0078] This formula ensures that the physical dimensions on both sides of the equation are strictly consistent; by introducing the dimension of acceleration and deceleration and integrating it with the speed weighting, the accuracy of driving risk assessment is enhanced, and it can more sensitively capture violent driving behaviors such as emergency acceleration and emergency braking, thereby more accurately predicting the instability of vehicle dynamics.

[0079] Example 4:

[0080] The process by which the scheduling decision unit calculates the collaborative allocation factor is as follows:

[0081] The risk difference is calculated by subtracting the system risk potential index from the risk threshold. Based on the risk difference and the preset slope control parameters, the collaborative allocation factor is calculated using the logistic function model.

[0082] This embodiment illustrates how the scheduling decision unit calculates the cooperative allocation factor. The process; to address the technical problem of how to smoothly and non-linearly generate resource scheduling instructions when system risks exceed a threshold, this embodiment introduces a logistic function;

[0083] The calculation process uses the system risk potential index. With risk threshold Perform the difference calculation to obtain the risk difference. This difference is the core variable driving resource allocation; subsequently, based on the risk difference and preset slope control parameters... The collaborative allocation factor is calculated using the logistic function model. The calculation formula is as follows:

[0084]

[0085] in, : The time-based co-allocation factor is dimensionless.

[0086] The current risk potential index is the core input of this formula, which is generated by the risk potential model in the previous steps.

[0087] The risk threshold for triggering the collaborative mode is dimensionless, and its setting logic has been explained above.

[0088] The slope control parameter is dimensionless and is derived from optimization through simulation testing to determine the sensitivity to changes in the allocation factor. The optimization goal of this parameter is to achieve smooth resource scheduling intervention when the risk index just exceeds the threshold, avoiding abrupt changes in user experience, while enabling rapid and significant resource allocation when the risk index far exceeds the threshold. A non-limiting example value could be... ;

[0089] Due to the exponential term Since it is dimensionless, this formula is physically correct; by employing the logistic function, a linearly changing risk difference is mapped to a... The interval-based, S-shaped curve-based collaborative allocation factor maximizes the preservation of user experience continuity while ensuring safety.

[0090] Example 5:

[0091] The process by which the rendering adjustment unit generates the adjusted target rendering load index for the infotainment domain is as follows:

[0092] Obtain the standard load index and the load index of the preset rendering task at the highest quality and the lowest acceptable quality; and based on the collaborative allocation factor, perform linear interpolation between the standard load index and the load index at the lowest acceptable quality to obtain the adjusted target rendering load index for the infotainment domain.

[0093] This embodiment illustrates how the rendering adjustment unit generates the adjusted infotainment domain target rendering load index. The process; to establish a clear path from the collaborative allocation factor to the specific rendering load target, this embodiment uses linear interpolation;

[0094] This calculation process requires obtaining the performance boundaries of the task:

[0095] Obtain the standard load index of the preset rendering task at the highest quality. Load index at the lowest acceptable image quality These two parameters are derived from offline performance profiling tools, which test and calibrate specific rendering tasks to define the adjustment range of the rendering complexity of the task.

[0096] Subsequently, the model is based on the collaborative allocation factor. In standard load index Load index at the lowest acceptable image quality Linear interpolation is performed between the two values ​​to obtain the final adjusted target rendering load index for the infotainment domain. The calculation formula is as follows:

[0097]

[0098] in, Adjusted infotainment domain target rendering load index, dimensionless;

[0099] The highest load boundary of the task, dimensionless, derived from offline calibration;

[0100] The minimum load boundary of the task, dimensionless, derived from offline calibration;

[0101] : Cooperative allocation factor, dimensionless, used as the weight of the linear interpolation method, calculated from the preceding steps;

[0102] All terms in the formula are dimensionless loading exponents or scaling factors, with consistent dimensions on both sides; the calculation logic of this method lies in... The value is based on The changes, in Linear variation within the interval; this method ensures that the adjustment of rendering load is precisely proportional to system risk, achieving fine-grained control.

[0103] Example 6:

[0104] The adaptive correction process of the closed-loop optimization unit is as follows:

[0105] Continuously monitor the rendering latency of keyframes in the safety domain; compare the rendering latency with the maximum allowable latency to calculate the normalized latency deviation; and based on the normalized latency deviation, use the gradient descent algorithm to update the weights in the risk potential quantification prediction model.

[0106] This embodiment illustrates the adaptive correction process of the closed-loop optimization unit. To address the technical challenge of open-loop control systems being unable to cope with changes in operating conditions and model drift, this embodiment introduces a correction mechanism based on actual performance feedback.

[0107] The adaptive correction process is a closed-loop link that feeds back the system's final output to its initial input. The process includes:

[0108] Continuously monitor the rendering latency of keyframes in the security domain. ;

[0109] Delay rendering With maximum allowable delay Perform a comparison to calculate the normalized delay bias. Maximum allowable delay The source is the absolute time limit defined for a specific safety task according to the functional safety standard;

[0110] Based on normalized delay bias The gradient descent algorithm is used to update the weights in the risk potential quantification prediction model;

[0111] When the detected rendering latency Exceeding the maximum allowed delay This indicates that the current risk prediction model failed to predict the conflict in a timely and accurate manner; at this point, the delay bias... This will be a positive value; the closed-loop optimization unit then uses this deviation signal to fine-tune the weights corresponding to the risk model input parameters that caused the delay. By iterating through this monitoring-comparison-correction process, the risk prediction model can gradually learn, so that when encountering similar situations in the future, it can calculate a higher risk potential index and trigger cooperative scheduling earlier. This closed-loop feedback optimization mechanism gives the system the ability to learn adaptively, greatly enhancing the robustness and reliability of the system throughout its entire lifecycle.

[0112] Example 7:

[0113] The calculation process for the normalized delay bias is as follows:

[0114] Obtain the difference between the rendering latency and the maximum allowable latency; compare the difference with zero and take the larger value as the numerator; divide the numerator by the maximum allowable latency to obtain the normalized latency bias.

[0115] This embodiment illustrates the normalized delay bias. The purpose of this calculation process is to generate a dimensionless, non-negative error signal to provide a standardized input for the subsequent gradient descent algorithm.

[0116] This calculation process obtains the rendering latency. With maximum allowable delay The difference ,in Representative at the The maximum rendering latency of the safety domain monitored within each correction cycle is then compared with zero, and the larger value is taken as the numerator. The numerator is then divided by the maximum allowable latency. The normalized delay bias is obtained. The calculation formula is as follows:

[0117]

[0118] in, : No. The normalized delay bias within each correction period is dimensionless.

[0119] The maximum rendering latency of the security domain monitored during this period is expressed in time and is sourced from real-time monitoring.

[0120] Maximum permissible delay, in time, is derived from functional safety standards.

[0121] The error signal generated by this method is proportional to the relative severity of the delay exceeding the limit, ensuring the accuracy of subsequent model corrections.

[0122] Example 8:

[0123] The weight update process is as follows:

[0124] Obtain the normalized delay bias, the preset learning rate, and the actual values ​​of the input parameters of the risk potential quantification prediction model corresponding to the moment that caused the rendering delay; multiply the normalized delay bias, the learning rate, and the actual values ​​of the input parameters to obtain the weight update amount; and subtract the weight update amount from the current weight value to obtain the updated weight.

[0125] This embodiment illustrates the weight update process; to achieve efficient and directional adaptive learning, this embodiment discloses a weight update rule based on the idea of ​​gradient descent;

[0126] The update process uses any one of the weights in the model. For example, its calculation process requires obtaining the normalized delay bias. Preset learning rate And the actual values ​​of the input parameters of the risk potential quantization prediction model corresponding to the moment that causes rendering delay. Among them, the learning rate This is a pre-defined dimensionless small positive number, determined through experimental tuning to ensure that the weight update process converges quickly while avoiding oscillations in the learning process caused by excessive deviations in a single delay. An unrestricted example value could be... Actual value of input parameter This refers to the moment when rendering delay occurs. Used to calculate the risk potential index The actual values ​​of each input parameter; specifically:

[0127] When updating weights ,correspond hour, For that moment value;

[0128] When updating weights ,correspond hour, For that moment value;

[0129] When updating weights ,correspond hour, For that moment value;

[0130] Similarly, if the optimization is extended to the vehicle dynamic instability model, when updating the weights... ,correspond hour, For that moment value;

[0131] Multiply the above three factors to obtain the weight update amount; from the current weight value Subtract the weight update amount from the middle to get the updated weight. The formula for its update rule is:

[0132]

[0133] in, : No. The updated value of each weight in the next period;

[0134] : No. The current value of each weight;

[0135] Learning rate, dimensionless;

[0136] Normalized delay bias, dimensionless, is calculated from the preceding steps;

[0137] The moment that causes the delay deviation The first risk potential index used in the calculation The actual values ​​of each input parameter are derived from a snapshot of the parameters at a corresponding point in time during system runtime;

[0138] By correlating the delay bias with the actual input parameter values ​​that cause the delay, it achieves a highly intelligent targeted correction, which greatly improves learning efficiency and model convergence speed.

[0139] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A control system for a vehicle display instrument, characterized by, The method comprises the following steps: A forward-looking perception unit is used to obtain the pre-warning level of an advanced driving assistance system and vehicle dynamic parameters, and to obtain the rendering task type of an infotainment domain and the GPU load, so as to generate multi-dimensional cabin system state data; A risk prediction unit is used to calculate a system risk potential index by a preset risk potential quantification prediction model according to the multi-dimensional cabin system state data; A scheduling decision unit is used to compare and analyze the system risk potential index with a preset risk threshold value; when the system risk potential index exceeds the risk threshold value, it is determined that the system enters a cooperative mode and a cooperative allocation factor is calculated; When the system risk potential index does not exceed the risk threshold value, it is determined that the system maintains an isolated mode; A rendering adjustment unit is used to dynamically adjust the rendering task complexity of the infotainment domain according to the cooperative allocation factor in response to the cooperative mode, so as to generate an adjusted infotainment domain target rendering load index; A closed-loop optimization unit is used to monitor the rendering delay of the safety domain, compare it with the maximum allowed delay of the functional safety requirement to generate a delay deviation, and then adaptively correct the weight of the risk potential quantification prediction model according to the delay deviation.

2. The display instrument control system for vehicle according to claim 1, characterized by The process of calculating the system risk potential index by the risk prediction unit is as follows: The pre-warning level of the advanced driving assistance system is quantified to obtain a pre-warning level quantization value; the vehicle dynamic instability is calculated according to the vehicle dynamic parameters; and the normalized rendering load index of the infotainment domain is obtained; The pre-warning level quantization value, the vehicle dynamic instability, and the normalized rendering load index are linearly weighted and summed to generate the system risk potential index.

3. The display instrument control system for a vehicle according to claim 2, characterized by The process of calculating the vehicle dynamic instability is as follows: The current speed and the current acceleration / deceleration are obtained; the current speed is normalized according to the maximum design speed to obtain a normalized current speed; The current acceleration / deceleration is normalized according to the maximum acceleration / deceleration to obtain a normalized current acceleration / deceleration; The normalized current speed and the normalized current acceleration / deceleration are weighted and summed to generate the vehicle dynamic instability.

4. The display instrument control system for vehicle according to claim 1, characterized by The process of calculating the cooperative allocation factor by the scheduling decision unit is as follows: The risk difference value is calculated by subtracting the risk threshold value from the system risk potential index; and the cooperative allocation factor is calculated by a logistic function model according to the risk difference value and a preset slope control parameter.

5. The display instrument control system for a vehicle according to claim 1, characterized by The process of generating the adjusted infotainment domain target rendering load index by the rendering adjustment unit is as follows: The standard load index under the highest quality and the load index under the lowest acceptable quality of the preset rendering task are obtained; and the adjusted infotainment domain target rendering load index is calculated by linear interpolation between the standard load index and the load index under the lowest acceptable quality based on the cooperative allocation factor.

6. The display instrument control system for a vehicle according to claim 1, characterized by The process of adaptive correction by the closed-loop optimization unit is as follows: The rendering delay of the safety domain key frame is continuously monitored; the normalized delay deviation is calculated by comparing the rendering delay with the maximum allowed delay; and the weight in the risk potential quantification prediction model is updated by using a gradient descent algorithm based on the normalized delay deviation.

7. The display instrument control system for a vehicle according to claim 6, wherein The process of calculating the normalized delay deviation is as follows: A difference between the rendering delay and the maximum allowed delay is obtained; the difference is compared with zero, and the larger one is taken as a numerator; and the numerator is divided by the maximum allowed delay to obtain a normalized delay deviation.

8. The display instrument control system for a vehicle according to claim 6, wherein The updating process of the weight is as follows: The normalized delay deviation, a preset learning rate, and actual values of input parameters of a risk state quantitative prediction model corresponding to the moment of the rendering delay are obtained; the normalized delay deviation, the learning rate, and the actual values of the input parameters are multiplied to obtain a weight updating amount; and the weight updating amount is subtracted from a current weight value to obtain an updated weight.

Images