Dynamic scheduling method for computing power of Android car-infotainment system based on Rockchip chip

Abstract

The invention discloses a method for dynamically scheduling computing power of an Android vehicle-mounted infotainment system based on a Rockchip chip, relates to the technical field of vehicle-mounted electronics, and realizes scenarized and elastic scheduling of the computing power through three core modules of'dynamic modeling ', 'real-time perception' and'heterogeneous scheduling '. The method comprises the following specific steps: establishing a task priority system by using a dynamic modeling module, and performing real-time adjustment based on a vehicle state signal to provide a decision basis for scheduling; the real-time sensing module establishes a computing power dynamic graph, quantifies task computing power requirements and evaluates residual computing power; the heterogeneous scheduling module adopts a computing power elastic telescoping and directional distribution algorithm, realizes efficient elastic telescoping of computing power resources, realizes linkage of a computing power scheduling layer and a Rockchip bottom layer power consumption control module, and dynamically adjusts the chip operation frequency, the balance performance and the power consumption according to a real-time load. According to the method, the core function response speed is increased, the computing power resource utilization rate is optimized, the vehicle engine power is reduced, and the system stability is improved.

Description

Technical Field

[0001] This invention relates to the field of automotive electronics technology, specifically to a method for dynamic scheduling of computing power in an Android vehicle infotainment system based on a Rockchip chip. Background Technology

[0002] With the rapid evolution of automotive intelligence and connectivity, Android-based in-vehicle infotainment systems have become the core carrier of in-vehicle interaction. Their functional boundaries continue to expand, encompassing diverse application scenarios such as navigation and positioning, 360° surround-view parking assistance, natural voice interaction, high-definition multimedia entertainment, and vehicle control parameter settings. In actual driving, these functions often exhibit multi-tasking concurrent operation. For example, navigation and music playback may occur simultaneously at high speeds, while surround-view image rendering and voice command response may be processed in parallel during parking scenarios. This places extremely high demands on the flexibility and real-time performance of the in-vehicle infotainment system's computing power allocation.

[0003] Rockchip chips, as the mainstream hardware support for Android in-vehicle infotainment systems, feature a high-performance heterogeneous computing architecture, integrating multi-core CPUs (such as a heterogeneous combination of A76 large cores and A55 small cores), a dedicated neural network processor (NPU, with computing power covering 1-6 TOPS), and a graphics processing unit (GPU), providing a sufficient hardware computing power foundation for multi-task parallel processing. However, current Android in-vehicle infotainment systems generally adopt a fixed computing power allocation model, which fails to fully adapt to the heterogeneous architecture advantages of Rockchip chips and cannot meet the dynamic computing power requirements in various scenarios, leading to a series of core problems.

[0004] First, there is both waste and contention for computing resources. In the fixed allocation mode, the system does not perform differentiated computing power scheduling based on task priority. High-priority critical tasks (such as real-time rendering of parking images and emergency navigation route planning) compete with low-priority non-critical tasks (such as background music playback and silent application updates) for limited computing power resources. This often leads to stuttering and response delays (delay time ≥300ms) in high-priority tasks, which seriously affects driving safety and user experience. In low-load scenarios (such as running only a single navigation function), the remaining computing power of the chip cannot be effectively utilized, resulting in redundant and wasted computing power.

[0005] Secondly, there is insufficient adaptability to different scenarios. The computing power requirements vary significantly in different driving scenarios (parking, highway driving, urban commuting, etc.). The fixed computing power allocation mode does not dynamically adjust to the computing power requirements of different driving scenarios, and still allocates computing power to each function in a fixed proportion, resulting in the core computing power requirements in specific scenarios not being met.

[0006] Third, the power consumption control is ineffective. Under the fixed computing power allocation mode, the chip needs to maintain a high computing power output to cover all possible concurrent tasks. Even in low-load scenarios, it cannot dynamically reduce computing power consumption, resulting in high power consumption of the vehicle system (normal operating current ≥1A), which does not meet the low power consumption requirements of the vehicle scenario.

[0007] Fourth, the hardware performance is not fully utilized. The heterogeneous computing architecture of Rockchip chips was originally designed to achieve efficient processing of different types of tasks through the collaborative work of CPU, NPU, and GPU (such as NPU handling AI-related calculations, GPU handling graphics rendering, and CPU handling general task scheduling). However, the existing fixed computing power allocation mode relies solely on the CPU to process multi-sensor data and various application tasks, failing to fully leverage the dedicated computing advantages of NPU and GPU. This results in the overall computing power potential of the chip not being released, forming a significant computing power bottleneck and restricting further improvement of the functionality and experience of Android in-vehicle infotainment systems.

[0008] In summary, the current Android in-vehicle infotainment systems based on Rockchip chips suffer from resource waste, insufficient scenario adaptation, excessive power consumption, and underutilization of hardware performance in terms of computing power allocation. These issues have become key bottlenecks restricting the improvement of the intelligence level and user experience of the in-vehicle infotainment systems. Therefore, it is urgent to propose a dynamic computing power scheduling method adapted to the heterogeneous architecture of Rockchip chips to achieve accurate allocation, scenario adaptation, and efficient utilization of computing power resources, thereby addressing the shortcomings of the existing fixed computing power allocation mode. Summary of the Invention

[0009] This invention provides a method for dynamic scheduling of computing power in an Android vehicle infotainment system based on a Rockchip chip, which solves the problems mentioned in the background section.

[0010] This invention provides the following technical solution: a dynamic scheduling method for computing power in an Android vehicle infotainment system based on a Rockchip chip. This method achieves scenario-based and elastic scheduling of computing power through three core modules: "dynamic modeling," "real-time perception," and "heterogeneous scheduling." Specifically, it includes the following steps:

[0011] S1. The dynamic modeling module establishes a task priority system, dividing all vehicle tasks into four different levels: special, first, second, and third. Based on the scenario triggering and priority adjustment mechanism, it continuously monitors the CAN bus signal and radar detection results, and performs real-time adaptive adjustment of task priorities based on vehicle status signals, providing decision-making basis for scheduling. There are also special cases for parking scenarios and highway scenarios, ensuring that core functions have the highest scheduling authority in parking and emergency scenarios.

[0012] S2. The real-time sensing module combines the real-time computing power sensing method of the Rockchip chip hardware interface. At the hardware level, it uses the computing power statistics interface of the Rockchip chip to collect data on the utilization of each core CPU, NPU / GPU computing power consumption, memory bandwidth, and power consumption in real time, and generates a dynamic computing power map. At the software level, it obtains the application process status through the Android system's ActivityManager, combines it with task priority, calculates the current computing power requirement and the remaining available computing power, and collects data through the software and hardware interfaces to achieve real-time and accurate monitoring of CPU / NPU / GPU computing power and power consumption.

[0013] The S3 heterogeneous scheduling module runs on the CPU cores of the Rockchip chip as an Android kernel-level service or framework-level daemon process. Utilizing a heterogeneous collaborative dynamic scheduling algorithm, it allocates high-priority / first-level tasks to the high-performance cores and NPU / GPU resources of the Rockchip chip, second-level tasks to hybrid cores, and third-level tasks to low-power cores. Furthermore, when a high-priority task starts, it reclaims computing power from low-priority tasks using both software and hardware methods, dynamically expanding the available resources to the target task. When a high-priority task finishes, it releases computing power to the computing power pool for other tasks to utilize, achieving efficient and elastic scaling of computing resources. It also works in conjunction with the Rockchip chip's dynamic voltage and frequency adjustment module to reduce the chip's operating frequency when computing power is sufficient and increase the frequency under high load, balancing performance and power consumption.

[0014] Preferably, the highest level includes emergency vehicle control and collision warning, with a response time of ≤50ms; the first level includes parking image and real-time navigation, with a response time of ≤100ms; the second level includes voice interaction and multimedia playback; and the third level includes background applications and system updates.

[0015] Preferably, the dynamic modeling module determines the final priority of the task based on the scene triggering and priority adjustment mechanism, and the final priority of the task is jointly determined by the basic priority and the scene weight.

[0016] Preferably, the special case of the parking scenario is: when the CAN signal displays "gear = R or D and vehicle speed ≤ 5km / h", the system immediately and forcibly upgrades the priority of the surround-view parking task from level one to level three.

[0017] Preferably, the high-speed scenario special case is as follows: when the CAN signal displays "vehicle speed ≥ 80km / h", the navigation priority remains at level one, and the multimedia task is reduced to level three.

[0018] Preferably, the S2 operation specifically involves:

[0019] I. Hardware-level computing power acquisition:

[0020] CPU load: By reading the CPU load monitoring registers exposed by the Rockchip chip core, the real-time utilization of each core of A76 / A55 can be obtained;

[0021] NPU / GPU Consumption: Obtain the real-time percentage utilization of NPU resources by calling the NPU computing power utilization query API provided by Rockchip;

[0022] Power consumption data: Real-time power consumption data is collected through the chip's PMIC interface;

[0023] II. Software Layer Task Requirement Awareness:

[0024] The PID, status, and historical CPU usage of running processes can be obtained in real time by calling Android Activity Manager.

[0025] Computing power requirement quantification: Calculate computing power requirements based on the final priority of the task;

[0026] III. Assessment of Remaining Computing Power:

[0027] The remaining available computing power is calculated in real time. The remaining available computing power is equal to the total computing power of the system minus the total computing power required by all tasks.

[0028] Preferably, the core resource-oriented allocation strategy of the heterogeneous collaborative dynamic scheduling algorithm is a priority allocation principle and a low power consumption constraint principle. The implementation method is to bind tasks of different priorities to a specified CPU core cluster by modifying the cgroup or taskset configuration of the Linux / Android kernel.

[0029] Preferably, the triggering condition for the heterogeneous scheduling module to achieve elastic scaling of computing resources is as follows: when a special-level / first-level task is started and the remaining available computing power is less than the task's computing power requirement, recycling is triggered; the recycling strategy adopted is: recycling computing power from low to high priority; soft recycling is: pausing the CPU usage of third-level background applications and reducing their thread scheduling weight; hard recycling is: reducing the **Nice value** of the second-level multimedia playback thread and forcibly migrating it to the A55 core.

[0030] Preferably, the specific trigger signal for the heterogeneous scheduling module linkage mechanism is:

[0031] The low-power mode is triggered when the average utilization of all A76 cores is below 30% for five consecutive scheduling cycles and there are no high-level tasks. The system reduces the operating frequency of the A76 cores through the DVFS interface, reducing the normal operating current from ≥1A to ≤800mA.

[0032] The high-performance mode is triggered when a special task starts, the system immediately boosts the corresponding core frequency to the maximum through the DVFS interface to ensure a response time of ≤50ms.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] 1. Improved response speed of core functions: The response latency of special / first-level tasks (such as parking image and emergency navigation) has been reduced from ≥300ms to ≤100ms, and the stuttering rate has been reduced by more than 95%;

[0035] 2. Optimize computing resource utilization: Reduce computing waste by 60% in low-load scenarios and improve computing power allocation accuracy by 80% in high-load scenarios;

[0036] 3. Reduce vehicle infotainment system power consumption: Normal operating current is reduced from ≥1A to ≤800mA, standby power consumption is reduced by 30%, and vehicle infotainment system battery life is extended;

[0037] 4. Enhanced system stability: Avoids system crashes caused by multi-task computing power contention, increasing the vehicle system's continuous trouble-free operation time to over 1000 hours;

[0038] 5. Compatible with multiple hardware platforms: Compatible with all Rockchip automotive-grade chips, requiring no reconstruction of underlying code and resulting in low adaptation costs. Attached Figure Description

[0039] Figure 1 This is a flowchart of a dynamic computing power scheduling method for an Android vehicle infotainment system based on a Rockchip chip, as proposed in this invention.

[0040] Figure 2 Dynamically model task priorities and create a scenario triggering flowchart;

[0041] Figure 3 Flowchart for real-time computing power perception and evaluation;

[0042] Figure 4 Here is a flowchart of the heterogeneous collaborative dynamic scheduling algorithm;

[0043] Figure 5 This refers to the control process for the linkage between computing power scheduling and DVFS. Detailed Implementation

[0044] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0045] This invention provides one embodiment: Please refer to Figures 1-5 A dynamic computing power scheduling method for an Android vehicle infotainment system based on Rockchip chips is proposed. This method achieves scenario-based and elastic scheduling of computing power through three core modules: "dynamic modeling," "real-time perception," and "heterogeneous scheduling." Specifically, it includes the following steps:

[0046] S1. The dynamic modeling module establishes a task priority system, dividing all vehicle tasks into four different levels: Special Level (P4), Level 1 (P3), Level 2 (P2), and Level 3 (P1). Based on the scenario triggering and priority adjustment mechanism, it continuously monitors CAN bus signals and radar detection results, and performs real-time adaptive adjustment of task priorities based on vehicle status signals, providing a basis for scheduling decisions. There are also special cases for parking scenarios and highway scenarios, ensuring that core functions have the highest scheduling authority in parking, emergency, and other scenarios.

[0047] The three levels are as follows: Level 1 includes emergency vehicle control and collision warning, with a response time of ≤50ms; Level 2 includes parking image and real-time navigation, with a response time of ≤100ms; Level 3 includes voice interaction and multimedia playback; and Level 4 includes background applications and system updates.

[0048] The principle of the scene triggering and priority adjustment mechanism is as follows: the system continuously monitors CAN bus signals (gear position, vehicle speed) and radar detection results, and calculates the scene weight W. scene .

[0049] Priority function: Final task priority P find Based on the basic priority P base and scene weight W scene Joint decision: P find =P base +ΔP•W scene ΔP represents the increment or change in priority.

[0050] A special case for parking scenarios is: when the CAN signal displays "gear = R or D and vehicle speed ≤ 5km / h", the system immediately and forcibly upgrades the priority of the surround-view parking task from Level 1 to Level 2.

[0051] The special case for high-speed scenarios is: when the CAN signal shows "vehicle speed ≥ 80km / h", the navigation priority remains at level one, and the multimedia task is reduced to level three.

[0052] By setting up the dynamic modeling module, dynamic modeling and hierarchical management of task priorities are achieved, enabling the dynamic scheduling of computing power in the Android vehicle system based on the Rockchip chip to support scenario-based adaptive adjustment, ensuring the absolute priority of critical safety tasks and improving the response speed of core functions.

[0053] S2. The real-time sensing module combines the real-time computing power sensing method of the Rockchip chip hardware interface. At the hardware level, it utilizes the computing power statistics interface of the Rockchip chip (such as the CPU load monitoring register of RK3588 and the NPU computing power usage query API) to collect data on the utilization of each core CPU, NPU / GPU computing power consumption, memory bandwidth, and power consumption in real time, generating a dynamic computing power graph. At the software level, it obtains the application process status through the Android system's ActivityManager, combines it with task priority, calculates the current computing power demand and the remaining available computing power, and collects data through the hardware and software interfaces to achieve real-time and accurate monitoring of CPU / NPU / GPU computing power and power consumption.

[0054] The specific steps are as follows:

[0055] I. Hardware-level computing power acquisition (Rockchip interface):

[0056] CPU load: The real-time utilization of each core of A76 / A55 can be obtained by reading the CPU load monitoring registers exposed by the Rockchip chip core (such as the Per-Core Load Register of RK3588).

[0057] NPU / GPU Consumption: Obtain the real-time percentage utilization of NPU resources by calling the NPU computing power utilization query API provided by Rockchip (such as RKNPU SDK);

[0058] Power consumption data: Real-time power consumption data is collected through the chip's PMIC (Power Management IC) interface;

[0059] II. Software Layer Task Requirement Awareness (Android Manager):

[0060] The PID, status (foreground / background), and historical CPU usage of running processes can be obtained in real time by calling Android Activity Manager.

[0061] Computing power requirement quantification: based on the final priority P of the task find Quantitative computing power requirements D task For example, the D-class parking image system. task It is configured with 3 A76 cores + 60% NPU computing power;

[0062] III. Assessment of Remaining Computing Power:

[0063] The remaining available computing power A is calculated in real time. The remaining available computing power A equals the total system computing power minus the total computing power required by all tasks. The formula is: A = (Total Compute) - ∑D task .

[0064] Through the real-time sensing module, real-time perception of computing power and demand assessment are realized, enabling accurate monitoring of CPU / NPU / GPU computing power and power consumption.

[0065] The S3 heterogeneous scheduling module runs on the CPU cores of the Rockchip chip as an Android kernel-level service or framework-level daemon process. Utilizing a heterogeneous collaborative dynamic scheduling algorithm, it allocates high-priority / first-level tasks to the high-performance cores and NPU / GPU resources of the Rockchip chip, second-level tasks to hybrid cores, and third-level tasks to low-power cores. When a high-priority task starts, it reclaims computing power from low-priority tasks using both software and hardware methods, dynamically expanding the available resources to the target task. When a high-priority task ends, it releases computing power to the computing power pool for other tasks to use, achieving efficient and elastic scaling of computing resources. Furthermore, it works in conjunction with the Rockchip chip's DVFS (Dynamic Voltage and Frequency Scaling) module to reduce the chip's operating frequency when computing power is sufficient and increase the frequency under high load, balancing performance and power consumption.

[0066] The core resource allocation strategy of the heterogeneous collaborative dynamic scheduling algorithm is the priority allocation principle and the low power consumption restriction principle. Priority allocation principle: high-priority / first-level tasks take priority in using the **high-performance core (A76)** and NPU / GPU resources of the Rockchip chip. Low power consumption restriction principle: third-level tasks are restricted to running on the **low-power core (A55)**, and the total computing power occupied is ≤10%. The implementation method is to bind tasks of different priorities to the specified CPU core cluster by modifying the cgroup or taskset configuration of the Linux / Android kernel.

[0067] The triggering condition for the heterogeneous scheduling module to achieve elastic scaling of computing resources is as follows: when a special-level / first-level task is started and the remaining available computing power is less than the task's computing power requirement, recycling is triggered; the recycling strategy adopted is: recycling computing power from low to high priority; soft recycling is: pausing the CPU usage of third-level (P1) background applications and reducing their thread scheduling weight; hard recycling is: reducing the **Nice value** of the second-level (P2) multimedia playback thread and forcibly migrating it to the A55 core.

[0068] The specific trigger signal for the heterogeneous scheduling module linkage mechanism is:

[0069] The low-power mode is triggered when the average utilization of all A76 cores is below 30% for five consecutive scheduling cycles and there are no high-level tasks. The system reduces the operating frequency of the A76 cores through the DVFS interface, reducing the normal operating current from ≥1A to ≤800mA.

[0070] The high-performance mode is triggered when a special task starts, the system immediately boosts the corresponding core frequency to the maximum through the DVFS interface to ensure a response time of ≤50ms.

[0071] By employing a heterogeneous scheduling module and utilizing cgroup / taskset core binding technology, high-priority tasks are allocated to A76 / NPU resources, while low-priority tasks undergo software / hardware computing power reclamation. This achieves efficient and elastic scaling of computing resources and optimizes their utilization. Furthermore, it enables linkage between the computing power scheduling layer and the Rockchip chip's underlying power consumption control module, dynamically adjusting the chip's operating frequency based on real-time load to balance performance and power consumption. This reduces the vehicle's system power consumption while preventing system crashes caused by multi-task computing power contention, extending the vehicle's continuous fault-free operation time to over 1000 hours and improving system stability.

[0072] Example:

[0073] The scheduling process of an Android vehicle infotainment system equipped with a Rockchip RK3588 chip after applying the method of this invention is as follows:

[0074] 1. Normal driving scenario (vehicle speed ≥ 30km / h): Navigation (Level 1) is allocated 2 A76 cores (accounting for 50%) and 20% of NPU computing power; music playback (Level 2) is allocated 1 A55 core (accounting for 20%); background applications (Level 3) use the remaining low-power computing power, and the total system power consumption is ≤ 800mA;

[0075] 2. Parking scenario (reverse gear engaged, vehicle speed ≤ 5km / h): 360° surround view parking (upgraded to premium level) automatically reclaims computing power from music playback and background applications, allocates 3 A76 cores (accounting for 75%) and 60% of NPU computing power, with processing latency ≤ 100ms and smooth video playback;

[0076] 3. Emergency Scenario (Collision Warning Triggered): Collision warning (Special Level) preempts 90% of computing power, suspends non-essential tasks, and has a response time of ≤50ms to ensure timely warning.

[0077] In summary, the dynamic scheduling method for computing power of Android vehicle infotainment systems based on Rockchip chips has the following beneficial effects: 1. Improved response speed of core functions: The response latency of special / first-level tasks (such as parking image and emergency navigation) is reduced from ≥300ms to ≤100ms, and the stuttering rate is reduced by more than 95%;

[0078] 2. Optimize computing resource utilization: Reduce computing waste by 60% in low-load scenarios and improve computing power allocation accuracy by 80% in high-load scenarios;

[0079] 3. Reduce vehicle infotainment system power consumption: Normal operating current is reduced from ≥1A to ≤800mA, standby power consumption is reduced by 30%, and vehicle infotainment system battery life is extended;

[0080] 4. Enhanced system stability: Avoids system crashes caused by multi-task computing power contention, increasing the vehicle system's continuous trouble-free operation time to over 1000 hours;

[0081] 5. Compatible with multiple hardware platforms: Compatible with all Rockchip automotive-grade chips, requiring no reconstruction of underlying code and resulting in low adaptation costs.

[0082] The contents not described in detail in this specification are prior art known to those skilled in the art. Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1.A Rockchip chip-based Android vehicle machine system computing power dynamic scheduling method, characterized by: It includes three modules of "dynamic modeling", "real-time perception", and "heterogeneous scheduling"; specifically including the following steps: S1, the dynamic modeling module establishes a task priority system, divides all tasks of the vehicle into four different levels of special, first, second, and third levels, continuously monitors CAN bus signals and radar detection results based on scene triggering and priority adjustment mechanism, and adjusts the task priority in real time based on the vehicle state signal, provides decision basis for scheduling, and exists parking scene special case and high-speed scene special case to ensure the highest scheduling right of core functions in parking and emergency scenes; S2, the real-time perception module combines the computing power of the Rockchip chip hardware interface real-time perception method, uses the computing power statistical interface of the Rockchip chip at the hardware level to collect real-time CPU utilization, NPU / GPU computing power consumption, memory bandwidth, and power consumption data, and generates a dynamic atlas of computing power; at the software level, the application process state is obtained through the ActivityManager of the Android system, combined with the task priority, the current computing power demand and the remaining available computing power are calculated, and the data is collected through the software and hardware interface to realize real-time and accurate monitoring of CPU / NPU / GPU computing power and power consumption; S3, the heterogeneous scheduling module runs on the CPU core of the Rockchip chip in the form of an Android kernel-level service or a framework layer daemon process, uses a heterogeneous collaborative dynamic scheduling algorithm to direct assign special / first-level tasks to high-performance cores and NPU / GPU resources of the Rockchip chip, second-level tasks to mixed cores, and third-level tasks to low-power cores, and when high-priority tasks are started, soft / hard computing power is recovered for low-priority tasks to dynamically expand to target tasks; when high-priority tasks end, the computing power is released to the computing power pool for other tasks to call, realizing efficient and flexible expansion of computing power resources; and it is linked with the dynamic voltage frequency adjustment module of the Rockchip chip to reduce the chip operating frequency when the computing power is sufficient, and to increase the frequency when the load is high, balancing performance and power consumption. 2.The Rockchip chip-based Android vehicle machine system computing power dynamic scheduling method according to claim 1, characterized in that: The special level includes emergency vehicle control and collision warning, which requires a response of ≤50ms; the first level includes parking video and real-time navigation, which requires a response of ≤100ms; the second level includes voice interaction and multimedia playback; and the third level includes background applications and system updates. 3.The Rockchip chip-based Android vehicle machine system computing power dynamic scheduling method of claim 1, wherein: The dynamic modeling module determines the final priority of the task based on the scene triggering and priority adjustment mechanism, and the final priority of the task is determined by the basic priority and the scene weight. 4.The Rockchip chip-based Android vehicle machine system computing power dynamic scheduling method of claim 1, wherein: The parking scene special case is: when the CAN signal shows "gear = R or D and vehicle speed ≤ 5km / h", the system immediately promotes the surround parking task priority from the first level to the special level. 5.The Rockchip chip-based Android vehicle machine system computing power dynamic scheduling method of claim 1, wherein: The high-speed scene special case is: when the CAN signal shows "vehicle speed ≥ 80km / h", the navigation priority remains at the first level, and the multimedia task is reduced to the third level. 6.The Rockchip chip-based Android vehicle machine system computing power dynamic scheduling method of claim 1, wherein: The specific operation of S2 is as follows: I. Hardware layer computing power collection: CPU load: by reading the CPU load monitoring register exposed by the Rockchip chip kernel, the real-time utilization rate of each core of A76 / A55 is obtained; NPU / GPU consumption: by calling the NPU computing power occupation query API provided by Rockchip, the real-time percentage occupation of NPU resources is obtained; Power consumption data: through the PMIC interface of the chip, real-time power consumption data is collected; II. Software layer task demand perception: By calling the Android Activity Manager, the PID, state and historical CPU usage of running processes are obtained in real time; Computing power demand quantification: according to the final priority of the task, the computing power demand is quantitatively calculated; III. Remaining computing power evaluation: Real-time calculation of remaining available computing power, which is equal to the total computing power of the system minus the sum of computing power required by all tasks. 7.The Rockchip chip-based Android vehicle machine system computing power dynamic scheduling method of claim 1, wherein: The core resource directional allocation strategy of the heterogeneous collaborative dynamic scheduling algorithm is the priority allocation principle and the low power consumption limitation principle, and the implementation is: by modifying the cgroup or taskset configuration of the Linux / Android kernel, tasks of different priorities are bound to specified CPU core clusters. 8.The Rockchip chip-based Android vehicle machine system computing power dynamic scheduling method of claim 1, wherein: The triggering condition for the heterogeneous scheduling module to realize computing power resource elastic scaling is: when a special / first-level task is started and the remaining available computing power is less than the task computing power demand, triggering recovery; The recovery strategy adopted is: according to the priority from low to high, the computing power is recovered; Soft recovery is: suspend the CPU occupation of the third-level background application, and reduce its thread scheduling weight; Hard recovery is: reduce the thread Nice value of the second-level multimedia player, and force it to migrate to the A55 core. 9.The Rockchip chip-based Android vehicle machine system computing power dynamic scheduling method of claim 1, wherein: The specific triggering signal of the linkage mechanism of the heterogeneous scheduling module is: Low power consumption mode triggering condition: when the average utilization rate of all A76 cores is less than 30% for 5 consecutive scheduling periods and there is no special task, the system reduces the operating frequency of the A76 core through the DVFS interface, and reduces the normal working current from ≥1A to ≤800mA; High performance mode triggering condition: when a special task is started, the system immediately raises the corresponding core frequency to the highest through the DVFS interface, ensuring that the response time is ≤50ms.

Images