Intelligent cabin processor frequency adjusting method, system, device and medium

Abstract

The invention discloses an intelligent cabin processor frequency adjusting method, system and device and a medium, and relates to the technical field of automobile intellectualization, and the method comprises the steps: obtaining a plurality of multi-mode sensor data of a vehicle-mounted intelligent cabin, and the current load data and the current temperature data of a processor; inputting the data into a load prediction model so as to output a load prediction result of the processor; classifying a plurality of pieces of sensor data into each priority queue, and if target sensor data classified into a target priority queue exists, obtaining delay time of the target priority queue; if the load prediction result is larger than a preset load threshold value, the current temperature data is smaller than a preset temperature safety threshold value, and the delay time is larger than or equal to a preset delay time threshold value, the processor is triggered to enter a temporary overclocking state; and after the processor enters the temporary overclocking state, adjusting the computing power frequency of the processor so as to increase the computing power frequency and maintain the computing power frequency for a target duration.

Description

Technical Field

[0001] The present application relates to the field of automobile intelligent technology, and in particular to a method, system, device and medium for adjusting the frequency of a smart cockpit processor. Background Art

[0002] At present, with the rapid development of automobile intelligence, the in-vehicle intelligent cockpit, as the core interactive platform of automobile intelligence, is evolving towards a more intelligent and humanized direction. The application of multimodal large models in the in-vehicle intelligent cockpit enables the cockpit to integrate multiple sensory data and provide users with richer and more convenient services, such as accurate voice interaction, intelligent environmental perception, and personalized entertainment experience. However, in the process of reasoning with multimodal large models in the in-vehicle intelligent cockpit, there are a series of problems that need to be solved.

[0003] However, the multimodal perception data of the smart cockpit has instantaneous peak characteristics, which seriously affects the real-time performance. The smart cockpit integrates many sensors, such as in-vehicle monitoring cameras, AR-HUD (Augmented Reality Head-Up Display) cameras, voice microphones, driver fatigue monitoring radars, etc. These devices will generate massive amounts of data at certain times. This causes the inference load to increase sharply in an instant, which cannot meet the computing power requirements of different tasks in the smart cockpit, and poses a great challenge to the system's processing capabilities. This may not only affect the user experience, but also reduce the overall performance and response speed of the system, thereby posing a potential threat to driving safety. Summary of the invention

[0004] A series of simplified concepts are introduced in the Summary of the Invention section, which will be further described in detail in the Detailed Description of the Invention section. The Summary of the Invention section of this application does not mean to attempt to define the key features and essential technical features of the claimed technical solution, nor does it mean to attempt to determine the scope of protection of the claimed technical solution.

[0005] In a first aspect, an embodiment of the present application provides a method for adjusting the frequency of a smart cockpit processor, the method comprising:

[0006] Acquire multiple multi-modal sensor data of the vehicle-mounted intelligent cockpit, as well as current load data and current temperature data of the processor;

[0007] Inputting the sensor data, the current load data, and the current temperature data into a load prediction model to output a load prediction result of the processor, wherein the load prediction model is obtained by training a deep learning network to be trained based on a plurality of historical samples, and the historical samples include: historical sensor data, historical load data, historical temperature data, and historical load prediction labels;

[0008] Classify the several sensor data into respective priority queues. If there is target sensor data classified into a target priority queue, obtain the latency time of the target priority queue, where the priorities of the respective priority queues are different, and the target priority queue has the highest priority among the respective priority queues;

[0009] If the load prediction result is greater than a preset load threshold, the current temperature data is less than a preset temperature safety threshold, and the latency time is greater than or equal to a preset latency time threshold, trigger the processor to enter a temporary overclocking state;

[0010] After the processor enters the temporary overclocking state, adjust the computing power frequency of the processor so that the computing power frequency increases and is maintained for a target duration, where the target duration is less than or equal to a preset duration threshold.

[0011] In an embodiment of the present invention, the step of, after the processor enters the temporary overclocking state, adjusting the computing power frequency of the processor so that the computing power frequency increases and is maintained for a target duration includes:

[0012] In the case of triggering temporary overclocking, determine the frequency increase ratio of the processor based on the load prediction result, the current temperature data, and the latency time of the target priority queue, and obtain the computing power improvement resources of the processor according to the frequency increase ratio;

[0013] Allocate the computing power improvement resources to the target priority queue to reduce the latency time of the target priority queue.

[0014] In an embodiment of the present invention, the method further includes:

[0015] After the processor enters the temporary overclocking state, pre-cool the processor or enhance the heat dissipation of the processor.

[0016] In an embodiment of the present invention, after the processor enters the temporary overclocking state, adjusting the computing power frequency of the processor so that the computing power frequency increases and is maintained for a target duration, where the target duration is less than or equal to a preset duration threshold, the subsequent steps include:

[0017] In the case where the current temperature data is greater than or equal to a preset temperature safety threshold, the load prediction result is less than or equal to a preset load threshold, the latency time of the target priority queue is less than a preset latency time threshold, or the duration of the temporary overclocking is greater than or equal to a preset overclocking upper limit duration, exit the temporary overclocking and restore to the normal frequency.

[0018] In one embodiment of the present invention, the load prediction model includes an LSTM model or a GRU model.

[0019] In one embodiment of the present invention, the method further includes:

[0020] After the processor enters the temporary overclocking state, obtain the real-time voltage of the processor;

[0021] In the case where the real-time voltage is greater than the threshold voltage set by the vehicle safety standard, reduce the frequency of the processor of the intelligent cockpit to below the rated frequency or switch to a standby processor.

[0022] In one embodiment of the present invention, the steps after inputting the sensor data, the current load data, and the current temperature data into the load prediction model to output the load prediction result of the processor include:

[0023] Based on a moving average or low-pass filtering algorithm, smooth the load prediction result.

[0024] In a second aspect, the present application provides an intelligent cockpit processor frequency adjustment system, the system includes: a prediction module, a classification module, and a trigger overclocking module;

[0025] The prediction module is configured to: obtain a plurality of multi-modal sensor data of the in-vehicle intelligent cockpit, as well as the current load data and the current temperature data of the processor; input the sensor data, the current load data, and the current temperature data into a load prediction model to output the load prediction result of the processor, wherein the load prediction model is trained by training a deep learning network to be trained based on a plurality of historical samples, and the historical samples include: historical sensor data, historical load data, historical temperature data, and historical load prediction labels;

[0026] The classification module is configured to: classify a plurality of the sensor data into respective priority queues, and if there is target sensor data classified into a target priority queue, obtain the delay time of the target priority queue, wherein the priorities of the respective priority queues are different, and the target priority queue has the highest priority among the respective priority queues;

[0027] The trigger overclocking module is configured to: if the load prediction result is greater than a preset load threshold, the current temperature data is less than a preset temperature safety threshold, and the delay time is greater than or equal to a preset delay time threshold, trigger the processor to enter a temporary overclocking state; after the processor enters the temporary overclocking state, adjust the computing power frequency of the processor so that the computing power frequency increases and maintains a target duration, and the target duration is less than or equal to a preset duration threshold.

[0028] In a third aspect, an electronic device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program stored in the memory, the steps of an intelligent cockpit processor frequency adjustment method according to any one of the first aspects described above are implemented.

[0029] In a fourth aspect, the present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of an intelligent cockpit processor frequency adjustment method according to any one of the first aspects are implemented.

[0030] In summary, for an intelligent cockpit processor frequency adjustment method according to an embodiment of the present application, after the processor enters the temporary overclocking state, the computing power frequency of the processor is adjusted so that the computing power frequency increases and maintains the target duration. Thus, when the multi-modal sensor data causes a large load on the processor, the computing power frequency of the processor is dynamically adjusted according to actual needs, enabling the processor to provide a larger computing power frequency in a short time, effectively meeting the requirements for high computing power when different tasks in the intelligent cockpit run concurrently, improving the overall system performance and response speed, and ensuring the smooth execution of queue tasks with higher priorities.

[0031] For the intelligent cockpit processor frequency adjustment method proposed in the present application, other advantages, objectives, and features of the present application will be partially reflected by the following description and partially understood by those skilled in the art through research and practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] By reading the detailed description of the preferred embodiments below, various other advantages and benefits will become clear to those of ordinary skill in the art. The drawings are only for the purpose of showing the preferred embodiments and are not considered to limit this specification. Moreover, throughout the drawings, the same reference numerals are used to represent the same components. In the drawings:

[0033] Figure 1 is a schematic flow chart of an intelligent cockpit processor frequency adjustment method provided by an embodiment of the present application;

[0034] Figure 2 is a schematic structural diagram of an intelligent cockpit processor frequency adjustment system provided by an embodiment of the present application;

[0035] Figure 3 is a schematic structural diagram of an electronic device for intelligent cockpit processor frequency adjustment provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0036] To better understand the technical solutions provided in the embodiments of this specification, the technical solutions of the embodiments of this specification will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific features in the embodiments of this specification and the embodiments are detailed descriptions of the technical solutions of the embodiments of this specification, rather than limitations on the technical solutions of this specification. Without conflict, the technical features in the embodiments of this specification and the embodiments can be combined with each other.

[0037] In this document, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprising", "including" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or device. Without further limitation, an element defined by the phrase "including a..." does not exclude the presence of additional identical elements in the process, method, article or device including the element. The term "more than two" includes two or more than two.

[0038] Please refer to Figure 1 , which is a schematic flow chart of a method for adjusting the frequency of an intelligent cockpit processor provided in an embodiment of this application, and specifically may include:

[0039] S110. Obtain a plurality of sensor data of multiple modalities of the in-vehicle intelligent cockpit, as well as the current load data and current temperature data of the processor;

[0040] Exemplarily, a plurality of sensor data of multiple modalities of the in-vehicle intelligent cockpit, as well as the current load data and current temperature data of the processor are obtained. There are various types of sensors in the in-vehicle intelligent cockpit, and these sensors collect various types of data of different modalities. For example, cameras are used to collect image data. For example, a driver fatigue monitoring camera collects the facial images of the driver, and an intelligent driving front-view camera collects the images of the road ahead. Microphones are used to collect voice data, and radars are used to obtain information such as distance. These multi-modal sensor data reflect various conditions inside and outside the cockpit and user operations.

[0041] The current load data can reflect the busy degree of the processor in processing tasks at the moment. For example, it can be measured by collecting indicators such as the utilization rate of the CPU or GPU and the task queue length. The current temperature data is related to the operating state of the processor. Excessive temperature will affect the performance of the processor and even cause damage. Obtaining these data is the basis for subsequent reasonable adjustment of the processor frequency, and they provide the system with comprehensive information about the current system operating environment and the working state of the processor.

[0042] By collecting multi-modal sensor data, current load data, and current temperature data, the system can comprehensively and meticulously understand various situations inside the intelligent cockpit and the real-time working state of the processor. This provides rich and accurate information basis for subsequent accurately judging whether the processor frequency needs to be adjusted and how to adjust it, and helps to make decisions more in line with actual needs. The multi-dimensional data enables the system to more accurately grasp the working load and environmental conditions of the processor, avoiding the one-sidedness of making decisions based on only single data, improving the accuracy and rationality of the frequency adjustment decision, and thus enhancing the operating efficiency and stability of the entire intelligent cockpit system.

[0043] S120. Input the sensor data, the current load data, and the current temperature data into a load prediction model to output the load prediction result of the processor, where the load prediction model is obtained by training a deep learning network to be trained based on multiple historical samples, and the historical samples include: historical sensor data, historical load data, historical temperature data, and historical load prediction labels;

[0044] Exemplarily, input the obtained sensor data, current load data, and current temperature data into the load prediction model, which is a deep learning network trained with a large number of historical samples. The historical samples cover historical sensor data, historical load data, historical temperature data, and corresponding historical load prediction labels at different time points. By training the deep learning network with these rich historical data, the model can learn the internal correlations and laws between the data. For example, the load prediction model can discover the relationships between specific sensor data patterns, load levels, and temperature changes from the historical data, so as to predict the future load situation of the processor according to the currently input data. The output load prediction result can help the system know in advance the load pressure that the processor will face, so as to make preparations in advance.

[0045] Based on the load prediction results output by the load prediction model, the system can anticipate the future load requirements of the processor in advance. This enables the system to have sufficient time to plan resources, such as deciding in advance whether to adjust the processor frequency, to avoid a decline in system performance or lags caused by the inability to make timely adjustments when the load peak suddenly arrives, thereby improving the system's response speed and stability. Accurate load prediction helps to allocate system resources more rationally. The system can arrange some background tasks reasonably when the load is low according to the prediction results, and adjust the processor frequency in advance before the load peak arrives to ensure that critical tasks can obtain sufficient resources, thus improving the overall resource utilization rate of the system and reducing energy consumption.

[0046] S130: Classify several pieces of the sensor data into respective priority queues. If there is target sensor data classified into a target priority queue, obtain the latency time of the target priority queue, where the priorities of the respective priority queues are different, and the target priority queue has the highest priority among the respective priority queues.

[0047] Exemplarily, since different tasks in the intelligent cockpit have different requirements for system resources and real-time requirements, it is necessary to classify the sensor data. The sensor data is divided into queues with different priorities according to the importance and real-time requirements of the tasks. For example, the sensor data related to tasks such as driver fatigue monitoring related to safety and ADAS (Advanced Driver Assistance Systems) vision recognition is classified into a high-priority queue (target priority queue); the task data related to interactions such as AR-HUD rendering and voice assistants is classified into a medium-priority queue; the task data related to entertainment such as video playback and games is classified into a low-priority queue. For the target priority queue, obtain its latency time. The latency time reflects the timeliness of task processing in this queue. By paying attention to the latency time of the highest-priority queue, the real-time nature of critical tasks can be ensured with emphasis.

[0048] By classifying the sensor data by priority and focusing on the latency time of the target priority queue, the importance and real-time requirements of different tasks can be clarified. In the case of limited resources, the processing of high-priority tasks is prioritized to ensure the timeliness and accuracy of critical tasks (such as safety-related tasks). This helps to improve the reliability and security of the system, while rationally allocating resources, improving the overall resource utilization efficiency, and avoiding the execution of high-priority tasks being affected by low-priority tasks occupying too many resources. To ensure that high-priority tasks are guaranteed first and low-priority tasks can be downclocked or delayed when resources are limited or the temperature is too high.

[0049] S140. If the load prediction result is greater than a preset load threshold, the current temperature data is less than a preset temperature safety threshold, and the delay time is greater than or equal to a preset delay time threshold, then trigger the processor to enter a temporary overclocking state;

[0050] Exemplarily, if the load prediction result is greater than a preset load threshold, the current temperature data is less than a preset temperature safety threshold, and the delay time is greater than or equal to a preset delay time threshold, then trigger the processor to enter a temporary overclocking state. The load prediction result being greater than the preset load threshold means that according to the load prediction model, the processor is about to face high load pressure, and it may not be able to process tasks in a timely manner relying only on the current frequency and resources. The current temperature data being less than the preset temperature safety threshold indicates that the current temperature of the processor is within the safe range, meeting the temperature condition for overclocking. Because overclocking will increase the power consumption and temperature of the processor, if the current temperature is too high, overclocking may cause the temperature to exceed the safe range, affecting the processor performance or even damaging the hardware. And the delay time being greater than or equal to the preset delay time threshold shows that a certain degree of delay has occurred in high-priority tasks, and measures need to be taken to improve the processing speed. When these three conditions are met simultaneously, the system determines that it is necessary to trigger the processor to enter a temporary overclocking state to cope with the upcoming high load situation and ensure the smooth execution of high-priority tasks.

[0051] Among them, the temperature safety threshold is 90°C. If the current temperature data of the chip ≥ 90°C, overclocking is prohibited; if the current temperature data < 85°C, it can enter the overclocking preparation state. The load threshold is 80%, and the delay time threshold is 50 ms. If the load prediction result > 80% and the delay time of the target priority queue exceeds or is equal to the delay time threshold, it is determined that overclocking is necessary. This triggering mechanism comprehensively considers multiple factors such as load, temperature, and task delay, making the processor overclocking decision more scientific and reasonable. By precisely setting each threshold, it can be triggered only when overclocking is truly needed, avoiding unnecessary overclocking operations, reducing energy consumption and hardware wear. At the same time, ensuring overclocking under the premise of temperature safety guarantees the stability and reliability of the system, improves the processing efficiency and timeliness of high-priority tasks (especially safety-critical tasks), and effectively enhances the overall performance and user experience of the intelligent cockpit system.

[0052] S150. After the processor enters the temporary overclocking state, adjust the computing power frequency of the processor so that the computing power frequency increases and maintains for a target duration, and the target duration is less than or equal to a preset duration threshold.

[0053] Exemplarily, when the processor enters the temporary overclocking state, in order to meet the processing requirements of high-load tasks, it is necessary to adjust its computing power frequency. Increasing the computing power frequency means that the processor can process more computing tasks per unit time and improve the processing speed. And maintaining the target duration is to ensure that high-priority tasks can receive sufficient computing power support and be successfully processed during this period. At the same time, the target duration is limited within a preset duration threshold to prevent overclocking for too long. Overclocking for too long may cause the processor temperature to be too high, the power consumption to be too large, damage the hardware, and affect the stability and lifespan of the system. Therefore, it is necessary to reasonably control the overclocking duration while meeting the task requirements. Among them, the target duration is 200 ms, and the preset duration threshold is 3 s, because it is meaningful to overclock continuously for at least 200 ms; but the maximum cannot exceed 3 s to prevent overheating or power consumption spikes.

[0054] By adjusting the computing power frequency and controlling the overclocking duration, while meeting the processing requirements of high-load tasks, the security and stability of the system are ensured. A reasonable overclocking duration setting can not only provide sufficient computing power support when key tasks are needed, improve the system's response speed and processing ability, but also avoid various risks brought by overclocking for too long, and extend the service life of the processor. This precise adjustment mechanism further optimizes the resource utilization of the intelligent cockpit system, improves the overall system performance, and provides a more smooth and reliable user experience.

[0055] In summary, the intelligent cockpit processor frequency adjustment method proposed in the embodiment of the present application, after the processor enters the temporary overclocking state, adjusts the computing power frequency of the processor, so that the computing power frequency increases and maintains the target duration, so that when the multi-modal sensor data causes a large load on the processor, the computing power frequency of the processor is dynamically adjusted according to actual needs, so that the processor provides a larger computing power frequency in a short time, effectively meeting the high-computing power requirements when different tasks in the intelligent cockpit run concurrently, improving the overall system performance and response speed, and ensuring the smooth execution of the queue tasks with higher priorities.

[0056] In some examples, the step of adjusting the computing power frequency of the processor after the processor enters the temporary overclocking state so that the computing power frequency increases and maintains the target duration includes:

[0057] In the case of triggering temporary overclocking, based on the load prediction result, the current temperature data, and the delay time of the target priority queue, determine the frequency increase ratio of the processor, and obtain the computing power improvement resources of the processor according to the frequency increase ratio;

[0058] Allocate the computing power improvement resources to the target priority queue to reduce the delay time of the target priority queue.

[0059] Exemplarily, after triggering temporary overclocking, the frequency increase ratio of the processor is determined based on the load prediction result, the current temperature data, and the latency time of the target priority queue (the highest priority queue), and then the computing power improvement resources are obtained. Then these computing power improvement resources are allocated to the target priority queue to reduce the latency time of this queue and ensure the efficient processing of high-priority tasks. Among them, it can be increased by 10% - 30% based on the rated frequency. This way of dynamically determining the frequency increase ratio based on multiple factors and targeted allocation of computing power resources makes the overclocking process more reasonable and efficient. It can meet the real-time requirements of high-priority tasks while avoiding resource waste and increased energy consumption caused by excessive overclocking, optimizing the utilization of system resources, and improving the processing quality and timeliness of critical tasks.

[0060] In some examples, the method further includes:

[0061] After the processor enters the temporary overclocking state, pre-cool the processor or enhance the heat dissipation of the processor.

[0062] Exemplarily, in the case of triggering temporary overclocking, it is necessary to pre-cool the processor or enhance its heat dissipation. In the intelligent cockpit system, overclocking will increase the power consumption and temperature of the processor. Taking pre-cooling or enhancing heat dissipation measures in advance can effectively control the temperature of the processor and ensure the stability and reliability of the overclocking process. By pre-cooling or enhancing heat dissipation, the risk of performance degradation (thermal throttling) of the processor caused by excessive temperature is greatly reduced, ensuring that the processor can continuously and stably operate in the overclocking state, improving the overall stability and reliability of the system, extending the service life of the processor, and achieving the purpose of heat dissipation linkage. Specifically: dock with the in-vehicle air duct / liquid cooling system and synchronously adjust the heat dissipation device according to the temporary overclocking requirements. If in the overclocking state, quickly turn on the enhanced mode of the vehicle air duct or start the liquid cooling cycle. And pre-heat or pre-cool in advance according to the response latency of the air duct / liquid cooling to reduce the risk brought by sudden temperature increase.

[0063] In some examples, after the processor enters the temporary overclocking state, the computing power frequency of the processor is adjusted to increase the computing power frequency and maintain it for a target duration, and the steps after the target duration is less than or equal to a preset duration threshold include:

[0064] When the current temperature data is greater than or equal to a preset temperature safety threshold, the load prediction result is less than or equal to a preset load threshold, the latency time of the target priority queue is less than a preset latency time threshold, or the duration of the temporary overclocking is greater than or equal to a preset overclocking upper limit duration, exit the temporary overclocking and restore to the normal frequency.

[0065] Exemplarily, when one of the following situations occurs: the current temperature data is greater than or equal to the preset temperature safety threshold, the load prediction result is less than or equal to the preset load threshold, the delay time of the target priority queue is less than the preset delay time threshold, or the duration of temporary overclocking is greater than or equal to the preset overclocking upper limit duration, the temporary overclocking state is exited and the normal frequency is restored. Among them, the current temperature data being greater than or equal to the preset temperature safety threshold means that the temperature is too high, and continued overclocking may damage the hardware. The load prediction result being less than or equal to the preset load threshold indicates that the load has decreased and there is no need for overclocking to provide additional computing power. The delay time of the target priority queue being less than the preset delay time threshold indicates that high-priority tasks are processed smoothly and there is no need for overclocking support. The duration of temporary overclocking being greater than or equal to the preset overclocking upper limit duration is to prevent various risks brought by excessive overclocking time.

[0066] The clear exit conditions set reasonable boundaries for the processor overclocking operation, avoiding continuous overclocking of the processor under unnecessary circumstances, reducing energy consumption and hardware wear, and ensuring the stability and security of system operation. At the same time, flexibly exiting overclocking according to different situations can better adapt to the dynamic changes of task loads and environmental conditions in the intelligent cockpit.

[0067] In some examples, the load prediction model includes an LSTM model or a GRU model.

[0068] Exemplarily, the load prediction model adopts a single-layer or double-layer LSTM (Long Short-Term Memory) model or GRU (Gated Recurrent Unit) model, with the input sequence length N (such as 5 - 10), and the input features being CPU / GPU utilization rate, queue length, vehicle-mounted camera frame rate, microphone voice activation frequency, historical temperature difference, etc. The load prediction model is used for load prediction. Both LSTM (Long Short-Term Memory) and GRU (Gated Recurrent Unit) are good at processing sequence data and can capture the time series features in the data. By learning the sequence information of historical sensor data, load data, temperature data, etc., it predicts the future load situation of the processor, providing a basis for subsequent decisions such as whether to trigger overclocking. A prediction update is performed approximately every 100 - 200 ms to ensure sufficient real-time performance while not overloading the system.

[0069] LSTM and GRU models can handle long-term dependencies in data and, compared with some traditional models, can learn the patterns in data more accurately, thereby improving the accuracy of processor load prediction. Accurate load prediction allows the system to more precisely determine when overclocking is needed, avoiding unnecessary overclocking or failure to overclock in a timely manner due to misjudgment and improving the efficiency of system resource utilization. These two models have good adaptability to the complex and changing intelligent cockpit environment. The data sources in the intelligent cockpit are extensive and dynamic. LSTM and GRU models can effectively process this complex sequential data, providing reliable prediction support for frequency adjustment in different scenarios and ensuring the stable operation of the system under various usage conditions.

[0070] In some examples, the method further includes:

[0071] After the processor enters the temporary overclocking state, obtain the real-time voltage of the processor;

[0072] In the case where the real-time voltage is greater than the threshold voltage set by the vehicle safety standard, reduce the frequency of the processor of the intelligent cockpit to below the rated frequency or switch to a standby processor.

[0073] Exemplarily, in a specific case where temporary overclocking is triggered, the system will monitor the voltage of the processor in real time. Vehicle safety standards such as ISO26262 set a threshold voltage to ensure the safe operation of automotive electronic devices. When the real-time voltage of the monitored processor is greater than this threshold voltage, it means that there is a safety risk in the current operating state of the processor. At this time, the system will take corresponding measures to reduce the frequency of the processor below the rated frequency to reduce power consumption and voltage, or directly switch to a standby processor to ensure that the system can continue to operate safely and stably.

[0074] By monitoring the real-time voltage of the processor and implementing corresponding measures, it is ensured that the processor is always within a safe voltage range during the operation of the intelligent cockpit system. This is crucial for an environment like an automobile where safety requirements are extremely high, avoiding processor damage, system failures, and even safety accidents caused by excessive voltage, and ensuring the safety of the entire intelligent cockpit system and vehicle driving. Adjusting the processor status in a timely manner when the voltage is abnormal can prevent system instability caused by voltage problems. For example, it can avoid problems such as processor operation errors and data loss caused by excessive voltage, ensuring that the system can continuously and stably provide services to users in various situations and improving the user experience. Once overclocking fails, the temperature exceeds the limit abnormally, or other failures occur (such as the liquid cooling system cannot start), immediately reduce the frequency and send a warning to the upper-level control system, or switch to a standby core to execute critical tasks to ensure vehicle safety.

[0075] In some examples, the steps after inputting the sensor data, the current load data, and the current temperature data into the load prediction model to output the load prediction result of the processor include:

[0076] Based on a moving average or low-pass filtering algorithm, smooth the load prediction result.

[0077] Exemplarily, after inputting the sensor data, the current load data, and the current temperature data into the load prediction model and obtaining the load prediction result, the load prediction result will be smoothed based on a moving average or low-pass filtering algorithm. The moving average algorithm smooths the data by calculating the average value of the data within a certain time window, and the low-pass filtering algorithm allows low-frequency signals to pass through and suppresses high-frequency noise. Both of these algorithms can reduce the noise and fluctuations in the load prediction result, making the prediction result more stable and reliable.

[0078] The smoothed load prediction result is more stable, reducing the overclocking decision-making errors caused by the instantaneous fluctuations of the prediction result. For example, it avoids frequent overclocking or downclocking operations caused by occasional abnormal data, enabling the system to make reasonable frequency adjustment decisions based on a more reliable load prediction, improving the stability and reliability of the system operation. Reducing unnecessary frequent frequency adjustments reduces the wear and tear of the processor and related hardware devices, helps extend the service life of the hardware devices, and reduces the system maintenance cost. At the same time, the stable frequency adjustment strategy can also reduce the system performance jitter caused by frequent adjustments, improving the overall system performance.

[0079] The present invention will be described in detail below with reference to embodiments, but they should not be construed as limiting the protection scope of the present invention.

[0080] Embodiment:

[0081] The vehicle is driving on the highway. Before starting the intelligent driving front-view camera for forward road navigation, there are rear-seat passengers in the car watching high-definition videos and running multiplayer online games.

[0082] For the acquisition of sensor / data stream data: Intelligent driving front-view camera: It is necessary to perform augmented reality rendering on the forward road (with high computational requirements and extremely high real-time requirements). In-vehicle monitoring camera: Collect the driver's expression and attention state to prevent fatigue driving (a key safety task). Voice microphone: The in-vehicle voice assistant is always on standby, with relatively frequent but medium real-time requirements. Rear-seat entertainment system: High-definition video decoding + game screen rendering, with high computing power requirements but relatively low priority.

[0083] When an instantaneous peak occurs: When the AR-HUD (Augmented Reality Head-Up Display) simultaneously loads high-resolution map data + driver facial expression recognition + entertainment system, strongly occupying the GPU / NPU, the overall computing power demand surges rapidly within a short period. If the system does not have an effective temporary overclocking and heat dissipation linkage, it is extremely likely to occur operation jams or frequency reduction due to overheating.

[0084] In this application, through priority scheduling, driver attention monitoring and AR-HUD recognition rendering are set as high priorities, and rear-seat entertainment is listed as a low priority. When the system load prediction model finds that the inference delay of critical tasks may exceed the limit and the chip temperature is acceptable, it immediately triggers temporary overclocking. At this time, the fan speed is quickly increased and liquid cooling is started to stabilize the temperature, so that the system can meet the real-time inference requirements of high priorities within the 2-second peak period. The in-vehicle entertainment may be affected by a small amount of frequency reduction but does not affect the safety function. After the peak, when it is detected that the task pressure is relieved, the overclocking mode is exited, and it gradually returns to the normal frequency and reduces the heat dissipation intensity.

[0085] This application aims at the computing power peak that appears when multi-modal data concurs in the intelligent cockpit. By means of temporary overclocking + linked heat dissipation, it provides peak performance in a short time, significantly reducing the inference delay of large models. Compared with simple "fixed overclocking + passive heat dissipation", this application automatically determines whether to overclock and when to exit according to real-time load and temperature prediction, combining the "pre-cooling" or "quick efficiency increase" mode of air duct / liquid cooling, greatly reducing the risk of overheating frequency reduction and avoiding wasting energy consumption during unnecessary periods. Compared with "conventional temperature-triggered heat dissipation", this application can intervene in heat dissipation in advance when the temperature has not soared to the limit, with higher response speed and better safety margin. By distinguishing tasks related to safety, interaction and entertainment, it can effectively guarantee the inference quality and timeliness of critical tasks in sudden high-load scenarios, while minimizing the impact on low-priority tasks. It supports vehicle safety requirements such as ISO 26262, comprehensively detects temperature, power supply, voltage, etc. during the overclocking process, and immediately switches to the safety mode and issues an alarm once an abnormality occurs. The shortest / longest time of temporary overclocking and the heat dissipation mechanism are quantitatively set, taking into account the functional feasibility and the stability of in-vehicle systems.

[0086] As Figure 2 shown, this application proposes an intelligent cockpit processor frequency adjustment system, and the system includes: a prediction module 21, a classification module 22 and a trigger overclocking module 23;

[0087] The prediction module 21 is configured to: obtain a plurality of multi-modal sensor data of the in-vehicle intelligent cockpit, as well as the current load data and the current temperature data of the processor; input the sensor data, the current load data, and the current temperature data into a load prediction model to output a load prediction result of the processor, where the load prediction model is obtained by training a deep learning network to be trained based on a plurality of historical samples, and the historical samples include: historical sensor data, historical load data, historical temperature data, and historical load prediction labels;

[0088] The classification module 22 is configured to: classify the plurality of sensor data into respective priority queues, and if there is target sensor data classified into a target priority queue, obtain the delay time of the target priority queue, where the priorities of the respective priority queues are different, and the target priority queue has the highest priority among the respective priority queues;

[0089] The trigger overclocking module 23 is configured to: if the load prediction result is greater than a preset load threshold, the current temperature data is less than a preset temperature safety threshold, and the delay time is greater than or equal to a preset delay time threshold, trigger the processor to enter a temporary overclocking state; after the processor enters the temporary overclocking state, adjust the computing power frequency of the processor so that the computing power frequency increases and maintains for a target duration, and the target duration is less than or equal to a preset duration threshold.

[0090] For the effects of the above system when applying the foregoing method, reference may be made to the description in the foregoing method embodiments, and details are not repeated herein.

[0091] As Figure 3 shown, an embodiment of the present application further provides an electronic device 300, including a memory 310, a processor 320, and a computer program 311 stored on the memory 310 and executable on the processor. When the processor 320 executes the computer program 311, the steps of any of the foregoing methods for adjusting the frequency of the intelligent cockpit processor are implemented.

[0092] Since the electronic device introduced in this embodiment is the device adopted for implementing an intelligent cockpit processor frequency adjustment device in an embodiment of the present application, based on the method introduced in the embodiment of the present application, those skilled in the art can understand the specific implementation manners and various variations of the electronic device in this embodiment. Therefore, the specific implementation of how this electronic device implements the method in the embodiment of the present application will not be described in detail herein. As long as the device adopted by those skilled in the art to implement the method in the embodiment of the present application belongs to the scope to be protected by the present application.

[0093] In the specific implementation process, when the computer program 311 is executed by the processor, it can implementFigure 1 Any implementation mode in the corresponding embodiment.

[0094] It should be noted that in the above embodiments, the descriptions of each embodiment have their own focuses. For the parts not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0095] Those skilled in the art should understand that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-readable storage media (including but not limited to disk memories, CD-ROMs, optical memories, etc.) containing computer-readable program codes.

[0096] The present application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of the processes and / or blocks in the flowchart and / or block diagram can also be implemented. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0097] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including instruction means, and the instruction means implements the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0098] These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0099] The embodiments of the present application also provide a computer program product, which includes computer software instructions. When the computer software instructions run on a processing device, the processing device is caused to execute the process of the LDPC decoding method of the solid-state drive controller.

[0100] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions according to the embodiments of the present application are wholly or partially generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from a website, a computer, a server, or a data center to another website, computer, server, or data center in a wired manner (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or a wireless manner (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be stored by a computer or a data storage device such as a server or a data center that integrates one or more available media. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium (for example, a DVD), or a semiconductor medium (for example, a solid-state disk (SSD)), etc.

[0101] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0102] In the several embodiments provided by the present application, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of units is only a logical function division, and there can be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed mutual coupling, direct coupling, or communication connection can be an indirect coupling or communication connection through some interfaces, devices, or units, and can be in an electrical, mechanical, or other form.

[0103] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they can be located in one place, or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0104] In addition, in each embodiment of the present application, each functional unit can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.

[0105] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods in each embodiment of the present application. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes.

[0106] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of each embodiment of the present application.

[0107] Although the preferred embodiments of this specification have been described, those skilled in the art can make additional changes and modifications once they know the basic creative concept. Therefore, the appended claims are intended to be construed as including the preferred embodiments and all changes and modifications that fall within the scope of this specification.

[0108] Obviously, those skilled in the art can make various changes and modifications to this specification without departing from the spirit and scope of this specification. Thus, if these modifications and variations of this specification fall within the scope of the claims of this specification and their equivalent technologies, this specification is also intended to include these modifications and variations.

Claims

1. A method for adjusting the frequency of a smart cockpit processor, characterized in that: The method comprises: Acquire multiple multi-modal sensor data of the vehicle-mounted intelligent cockpit, as well as current load data and current temperature data of the processor; Inputting the sensor data, the current load data, and the current temperature data into a load prediction model to output a load prediction result of the processor, wherein the load prediction model is obtained by training a deep learning network to be trained based on a plurality of historical samples, and the historical samples include: historical sensor data, historical load data, historical temperature data, and historical load prediction labels; Classify the plurality of sensor data into respective priority queues, and if there is target sensor data classified into a target priority queue, obtain the delay time of the target priority queue, wherein the priorities of the priority queues are different, and the target priority queue has the highest priority among the priority queues; If the load prediction result is greater than a preset load threshold, the current temperature data is less than a preset temperature safety threshold, and the delay time is greater than or equal to a preset delay time threshold, the processor is triggered to enter a temporary overclocking state; After the processor enters the temporary overclocking state, the computing frequency of the processor is adjusted to increase the computing frequency and maintain a target duration, where the target duration is less than or equal to a preset duration threshold.

2. The method for adjusting the frequency of a smart cockpit processor according to claim 1, characterized in that: After the processor enters the temporary overclocking state, the step of adjusting the computing frequency of the processor so that the computing frequency increases and maintains the target duration includes: In the case of triggering temporary overclocking, determining the frequency increase ratio of the processor based on the load prediction result, the current temperature data and the delay time of the target priority queue, and obtaining the computing power increase resource of the processor according to the frequency increase ratio; The computing power enhancement resources are allocated to the target priority queue to reduce the delay time of the target priority queue.

3. The method for adjusting the frequency of a smart cockpit processor according to claim 1, characterized in that: The method further comprises: After the processor enters the temporary overclocking state, the processor is precooled or the heat dissipation of the processor is enhanced.

4. The method for adjusting the frequency of a smart cockpit processor according to claim 1, characterized in that: After the processor enters the temporary overclocking state, the computing frequency of the processor is adjusted to increase the computing frequency and maintain the target duration. The steps after the target duration is less than or equal to the preset duration threshold include: When the current temperature data is greater than or equal to the preset temperature safety threshold, the load prediction result is less than or equal to the preset load threshold, the delay time of the target priority queue is less than the preset delay time threshold, or the temporary overclocking duration is greater than or equal to the preset overclocking upper limit duration, exit the temporary overclocking and restore to the normal frequency.

5. The method for adjusting the frequency of a smart cockpit processor according to claim 1, characterized in that: The load prediction model includes an LSTM model or a GRU model.

6. The method for adjusting the frequency of a smart cockpit processor according to claim 1, characterized in that: The method further comprises: After the processor enters a temporary overclocking state, obtaining a real-time voltage of the processor; When the real-time voltage is greater than the threshold voltage set by the vehicle safety standard, the frequency of the processor of the smart cockpit is reduced to below the rated frequency or switched to a backup processor.

7. The method for adjusting the frequency of a smart cockpit processor according to claim 1, characterized in that: The steps after inputting the sensor data, the current load data and the current temperature data into the load prediction model to output the load prediction result of the processor include: The load forecast results are smoothed based on a sliding average or low-pass filtering algorithm.

8. An intelligent cockpit processor frequency adjustment system, characterized in that: The system comprises: a prediction module, a classification module and a trigger overclocking module; The prediction module is configured to: obtain a plurality of multi-modal sensor data of the vehicle-mounted intelligent cockpit, and current load data and current temperature data of the processor; input the sensor data, the current load data and the current temperature data into a load prediction model to output a load prediction result of the processor, wherein the load prediction model is obtained by training a deep learning network to be trained based on a plurality of historical samples, and the historical samples include: historical sensor data, historical load data, historical temperature data and historical load prediction labels; The classification module is configured to: classify the plurality of sensor data into respective priority queues, and if there is target sensor data classified into a target priority queue, obtain the delay time of the target priority queue, wherein the priorities of the priority queues are different, and the target priority queue has the highest priority among the priority queues; The trigger overclocking module is configured as follows: if the load prediction result is greater than a preset load threshold, the current temperature data is less than a preset temperature safety threshold, and the delay time is greater than or equal to a preset delay time threshold, then the processor is triggered to enter a temporary overclocking state; after the processor enters the temporary overclocking state, the computing power frequency of the processor is adjusted to increase the computing power frequency and maintain a target duration, and the target duration is less than or equal to a preset duration threshold.

9. An electronic device, comprising: A memory and a processor, characterized in that the processor is used to implement the steps of a smart cockpit processor frequency adjustment method as described in any one of claims 1 to 7 when executing a computer program stored in the memory.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of a method for adjusting the frequency of an intelligent cockpit processor as described in any one of claims 1 to 7 are implemented.

Images