On-board heterogeneous computing system and method, and vehicle and storage medium

By combining a neuromorphic computing module and a fault-tolerant scheduling module, and utilizing spintronics technology for dynamic fault-tolerant scheduling and resource management, the problem of insufficient energy efficiency and fault tolerance in traditional in-vehicle computing platforms is solved, and a low-power, high-reliability in-vehicle computing system is realized.

WO2026237566A1PCT designated stage Publication Date: 2026-11-19CHERY AUTOMOBILE CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
PCT/CN2026/086694
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2025-09-04
Filing Date
2026-03-27
Publication Date
2026-11-19

AI Technical Summary

Technical Problem

Traditional in-vehicle computing platforms cannot balance energy efficiency and fault tolerance, and do not utilize low-power technologies such as spintronics to simulate the parallel processing characteristics of the human brain, resulting in resource waste, low efficiency, high power consumption and low reliability of in-vehicle systems.

Method used

A neuromorphic computing module is used to perform convolution operations on a heterogeneous material magnetic synapse array driven by spin orbital moments. Combined with a fault-tolerant scheduling module and an energy efficiency management module, a dynamic fault-tolerant scheduling algorithm is used to monitor the processor status, perform task migration and resource reallocation, and achieve energy efficiency fluctuation suppression.

Benefits of technology

It significantly improves the energy efficiency and reliability of in-vehicle computing systems, reduces power consumption, reduces fault switching latency, and supports high computing power output for L4/L5 level autonomous driving scenarios.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN2026086694_19112026_PF_FP_ABST
    Figure CN2026086694_19112026_PF_FP_ABST
Patent Text Reader

Abstract

The present application relates to an on-board heterogeneous computing system and method, and a vehicle and a storage medium. The system comprises: a neuromorphic computing module 100, which is configured to receive LiDAR point cloud data, and use a spin-orbit torque-driven heterogeneous-material magnetic synapse array to perform a convolution operation on the LiDAR point cloud data; a fault-tolerant scheduling module 200, which is configured to monitor the operating state of a processor, so as to determine a fault situation of the processor on the basis of the operating state, and execute, on the basis of the fault situation, at least one task migration operation on the basis of task priorities; and an energy efficiency management module 300, which is configured to dynamically adjust a power supply voltage and computing power allocation of a target computing resource, so as to release reserved computing power and perform energy efficiency fluctuation suppression when a fault occurs in the processor.
Need to check novelty before this filing date? Find Prior Art

Description

In-vehicle heterogeneous computing system, method, vehicle and storage medium

[0001] This application claims priority to Chinese Patent Application No. 202511260710.2, filed on September 4, 2025, entitled "In-vehicle Heterogeneous Computing System, Method, Vehicle and Storage Medium", the entire contents of which are incorporated herein by reference. Technical Field

[0002] This application relates to the field of vehicle computing technology, and in particular to a vehicle heterogeneous computing system, method, vehicle, and storage medium. Background Technology

[0003] With the rapid development of intelligent connected vehicle technology, Level 4 / Level 5 highly automated driving systems have placed unprecedented demands on onboard computing platforms for high performance, high reliability, and low power consumption. Autonomous driving tasks such as perception, decision-making, and planning require real-time processing of massive amounts of data from multiple heterogeneous sensors, including LiDAR, cameras, millimeter-wave radar, and Vehicle-to-Everything (V2X) communication devices. This poses a severe challenge to the computing power, energy efficiency, and fault tolerance of onboard computing systems.

[0004] In related technologies, mainstream automotive computing platforms generally adopt a heterogeneous architecture with CPU, graphics processing unit (GPU) and dedicated AI accelerator (such as NPU) as the core, such as high-performance chips like NVIDIA Thor and MDC. Summary of the Invention

[0005] This application provides an in-vehicle heterogeneous computing system, method, vehicle, and storage medium to address the problems of traditional architectures failing to balance energy efficiency and fault tolerance, failing to utilize low-power technologies such as spintronics to simulate the parallel processing characteristics of the human brain, and resource waste and low efficiency caused by static fault tolerance strategies (such as task redundancy replication).

[0006] The first aspect of this application provides an in-vehicle heterogeneous computing system, comprising: a neuromorphic computing module for receiving lidar point cloud data and performing convolution operations on the lidar point cloud data using a spin-orbit moment-driven heterogeneous material magnetic synapse array to generate obstacle detection results; a fault-tolerant scheduling module for monitoring the processor's operating status based on the obstacle detection results, determining the processor's fault condition based on the operating status, and executing at least one task migration operation based on the fault condition and task priority; and an energy efficiency management module for dynamically adjusting the power supply voltage and computing power allocation of target computing resources based on the at least one task migration operation, so as to release reserve computing power and suppress energy efficiency fluctuations in the event of a processor failure.

[0007] Optionally, in one embodiment of this application, the neuromorphic computing module includes: a magnetization reversal unit, configured to stabilize the magnetization direction through antiferromagnetic order and drive magnetization reversal based on the magnetization direction using the spin orbital moment to generate magnetization state switching data; a weight adjustment unit, configured to adjust the weights based on the magnetization state switching data according to a preset synaptic weight formula to generate final weights; and a convolution operation unit, configured to perform convolution operations based on the final weights using the heterogeneous material magnetic synaptic array to generate the obstacle detection result.

[0008] Optionally, in one embodiment of this application, the fault-tolerant scheduling module includes: a fault detection unit, configured to monitor the temperature and cyclic redundancy check (CRC) error rate of the processor using a dynamic fault-tolerant scheduling algorithm, and determine the fault condition of the processor when the temperature exceeds a preset temperature threshold or the CRC error rate exceeds a preset error threshold; and a task migration unit, configured to perform the task migration operation based on the fault condition and the task priority.

[0009] Optionally, in one embodiment of this application, the energy efficiency management module includes: an increase unit, configured to increase the number of processor cores from a preset value to a reserve value, and increase the actual voltage value to a target voltage value; and a resource reallocation unit, configured to release the reserve computing power in the event of a processor failure based on the reserve value and the target voltage value, and control energy efficiency fluctuations through an adaptive voltage regulation algorithm to perform a resource reallocation operation.

[0010] Optionally, in one embodiment of this application, the energy efficiency management module further includes: a voltage regulation unit, used to dynamically adjust the power supply voltage of the computing resources using the adaptive voltage regulation algorithm; and a computing power scheduling unit, used to coordinate the computing power allocation of each computing node based on the task priority and the real-time distribution of the target computing resources.

[0011] A second aspect of this application provides an in-vehicle heterogeneous computing method, comprising the following steps: receiving lidar point cloud data and performing convolution operations on the lidar point cloud data using a spin-orbit moment-driven heterogeneous material magnetic synapse array to generate obstacle detection results; based on the obstacle detection results, monitoring the operating status of the processor to determine the processor's fault condition according to the operating status, and based on the fault condition, performing at least one task migration operation according to task priority; and based on the at least one task migration operation, dynamically adjusting the power supply voltage and computing power allocation of the target computing resources to release reserve computing power and suppress energy efficiency fluctuations in the event of a processor failure.

[0012] Optionally, in one embodiment of this application, the step of receiving lidar point cloud data and performing convolution operations on the lidar point cloud data using a spin-orbit moment-driven heterogeneous material magnetic synapse array to generate obstacle detection results includes: stabilizing the magnetization direction through antiferromagnetic ordering, and based on the magnetization direction, driving magnetization reversal using the spin-orbit moment to generate magnetization state switching data; adjusting the weights according to a preset synaptic weight formula based on the magnetization state switching data to generate final weights; and performing convolution operations using the heterogeneous material magnetic synapse array based on the final weights to generate the obstacle detection results.

[0013] Optionally, in one embodiment of this application, monitoring the processor's operating status to determine the processor's fault condition based on the operating status, and performing at least one task migration operation based on the fault condition and task priority, includes: monitoring the processor's temperature and cyclic redundancy check (CRC) error rate using a dynamic fault-tolerant scheduling algorithm, and determining the processor's fault condition if the temperature exceeds a preset temperature threshold or the CRC error rate exceeds a preset error threshold; and performing the task migration operation based on the fault condition and task priority.

[0014] Optionally, in one embodiment of this application, the step of dynamically adjusting the power supply voltage and computing power allocation of the target computing resources based on the at least one task migration operation to release reserve computing power and suppress energy efficiency fluctuations in the event of a processor failure includes: increasing the number of processor cores from a preset value to a reserve value and increasing the actual voltage value to a target voltage value; based on the reserve value and the target voltage value, releasing the reserve computing power in the event of a processor failure, and controlling energy efficiency fluctuations through an adaptive voltage adjustment algorithm to perform a resource reallocation operation.

[0015] Optionally, in one embodiment of this application, the step of dynamically adjusting the power supply voltage and computing power allocation of the target computing resources based on the at least one task migration operation, so as to release reserve computing power and suppress energy efficiency fluctuations in the event of a processor failure, further includes: dynamically adjusting the power supply voltage of the computing resources using the adaptive voltage adjustment algorithm; and coordinating the computing power allocation of each computing node based on the task priority and the real-time distribution of the target computing resources.

[0016] A third aspect of this application provides a vehicle, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the in-vehicle heterogeneous computing method as described in the above embodiments.

[0017] A fourth aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described vehicle-mounted heterogeneous computing method.

[0018] This application's embodiments can be based on a neuromorphic computing hardware architecture, stabilizing the magnetization direction through antiferromagnetic ordering without an external magnetic field. Magnetization reversal is driven by spin orbital moments to switch energy, and neuromorphic synapse design is implemented. Based on the Dynamic Fault-Tolerant Scheduling (FTDR) algorithm, during fault detection, processor temperature and CRC error rate are monitored, task priority is calculated, and task migration is completed, achieving dynamic fault-tolerant switching under sudden failures. Based on resource reallocation, reserve computing power is released, energy efficiency fluctuations are suppressed, and energy efficiency management is adopted to improve the energy efficiency ratio of heterogeneous computing. Therefore, it solves the problems of traditional architectures failing to balance energy efficiency and fault tolerance, failing to utilize low-power technologies such as spintronics to simulate the parallel processing characteristics of the human brain, and the resource waste and low efficiency caused by static fault-tolerant strategies (such as task redundancy replication), as well as the contradiction between high power consumption (>800W) and low reliability (switching latency >200ms) of automotive systems.

[0019] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0020] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0021] Figure 1 is a schematic diagram of a vehicle-mounted heterogeneous computing system according to an embodiment of this application;

[0022] Figure 2 is an architecture diagram of an in-vehicle heterogeneous computing system according to an embodiment of this application;

[0023] Figure 3 is a schematic diagram of the neural synaptic structure of an in-vehicle heterogeneous computing system according to an embodiment of this application;

[0024] Figure 4 is a schematic diagram of the dynamic fault-tolerant process of an in-vehicle heterogeneous computing system according to an embodiment of this application;

[0025] Figure 5 is a flowchart illustrating an in-vehicle heterogeneous computing method according to an embodiment of this application;

[0026] Figure 6 is a structural schematic diagram of a vehicle provided according to an embodiment of this application. Detailed Implementation

[0027] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0028] Traditional CPU / GPU architectures have peak power consumption exceeding 800W (such as NVIDIA Thor chips), which is insufficient to meet the low power consumption requirements of automotive applications. Hardware failure switching latency exceeds 200ms, failing to meet the ASIL-D safety standard. Furthermore, the lack of dynamic adjustment of computing power allocation based on real-time risks results in insufficient resource utilization, which urgently needs improvement.

[0029] The following description, with reference to the accompanying drawings, describes an in-vehicle heterogeneous computing system, method, vehicle, and storage medium according to embodiments of this application. To address the issues mentioned in the background technology, such as the inability of traditional architectures to balance energy efficiency and fault tolerance, the failure to utilize low-power technologies like spintronics to simulate the parallel processing characteristics of the human brain, the waste of resources and low efficiency caused by static fault-tolerance strategies (such as task redundancy replication), and the contradiction between high power consumption (>800W) and low reliability (switching latency >200ms) in automotive systems, this application provides an automotive heterogeneous computing method. This method utilizes a neuromorphic computing hardware architecture, stabilizing magnetization direction through antiferromagnetic ordering without an external magnetic field. It drives magnetization reversal via spin orbital moments to switch energy and performs brain-like synapse design. Based on the Fault-Tolerant Dynamic Rescheduling (FTDR) algorithm, during fault detection, it monitors processor temperature and Cyclic Redundancy Check (CRC) error rate, calculates task priorities, and completes task migration, achieving dynamic fault-tolerance switching under sudden failures. Furthermore, it releases reserve computing power through resource reallocation, suppresses energy efficiency fluctuations, and employs energy efficiency management to improve the energy efficiency ratio of heterogeneous computing. This solves the problems of traditional architectures being unable to balance energy efficiency and fault tolerance, failing to utilize low-power technologies such as spintronics to simulate the parallel processing characteristics of the human brain, resource waste and low efficiency caused by static fault tolerance strategies (such as task redundancy replication), and the contradiction between high power consumption (>800W) and low reliability (switching latency >200ms) of automotive systems.

[0030] Specifically, Figure 1 is a schematic diagram of the structure of an in-vehicle heterogeneous computing system provided in an embodiment of this application.

[0031] As shown in Figure 1, the vehicle-mounted heterogeneous computing system 10 includes: a neuromorphic computing module 100, a fault-tolerant scheduling module 200, and an energy efficiency management module 300.

[0032] Specifically, the neuromorphic computing module 100 is used to receive lidar point cloud data and perform convolution operations on the lidar point cloud data using a heterogeneous material magnetic synapse array driven by spin orbital moments to generate obstacle detection results.

[0033] It is understood that the neuromorphic computing module 100 in the embodiments of this application can be a neuromorphic computing hardware architecture.

[0034] In actual implementation, the neuromorphic computing module 100 in this embodiment can perform neuromorphic computing task processing, including:

[0035] 1. Data input: LiDAR point cloud data is transmitted to the neuromorphic computing layer through a Time-Sensitive Networking (TSN) network;

[0036] 2. Parallel processing: Spin electron synapse array performs convolution operations (processing speed: 1.2 T OPS);

[0037] S-type neuron activation function simulates human brain decision-making (delay <1μs);

[0038] 3. Output decision: Generate obstacle detection results and transmit them to the autonomous driving controller.

[0039] The embodiments of this application can use a heterogeneous material magnetic synapse array driven by spin orbital moment for parallel computing, which improves the overall energy efficiency ratio of the system and simulates the decision-making mechanism of the human brain to enhance the robustness of the model in complex scenarios.

[0040] Optionally, in one embodiment of this application, the neuromorphic computing module 100 includes: a magnetization reversal unit, used to stabilize the magnetization direction by antiferromagnetic order and drive magnetization reversal based on the magnetization direction using spin orbital moments to generate magnetization state switching data; a weight adjustment unit, used to adjust the weights based on the magnetization state switching data according to a preset synaptic weight formula to generate final weights; and a convolution operation unit, used to perform convolution operations based on the final weights using a heterogeneous material magnetic synaptic array to generate obstacle detection results.

[0041] In this application, a multilayer stack of W (50nm) / Pt (2nm) / Co (1.5nm) / NiO (3nm) / Pt (2nm) is used. The magnetization direction is stabilized by the antiferromagnetic order (NiO layer). No external magnetic field is required. The magnetization is driven by the spin-orbit torque (SOT), and the switching energy is 0.1pJ / bit (1pJ / bit in the traditional scheme).

[0042] In neuromorphic synapse design, the synaptic weight formula is: ΔW=α·V_input·V_output

[0043] Where α = 0.05, and the input / output voltage range is 0-1V.

[0044] The temperature range it can withstand is -40℃ to 150℃.

[0045] The embodiments of this application can stabilize the magnetization direction through antiferromagnetic order, realize magnetization reversal without the need for an external magnetic field, significantly reduce power consumption, the magnetic synapse array supports in-memory computing architecture, the convolution operation is completed in the array in situ, avoiding data transfer bottlenecks, and the heterogeneous material structure has excellent thermal stability and vibration resistance, adapting to the harsh automotive environment.

[0046] The fault-tolerant scheduling module 200 is used to monitor the processor's operating status based on obstacle detection results, determine the processor's fault status according to the operating status, and perform at least one task migration operation according to the task priority based on the fault status.

[0047] It is understood that the operating states in the embodiments of this application include, but are not limited to, processor operating temperature, CRC check error rate, number of memory error checking and correcting (ECC) errors, and task response latency.

[0048] In actual execution, the fault-tolerant scheduling module 200 in this embodiment can monitor the processor's operating status to determine the processor's fault status based on the operating status, and perform at least one task migration operation based on the fault status and task priority.

[0049] The embodiments of this application can achieve millisecond-level detection and judgment of hardware faults, avoid system failure due to single point of failure, and significantly improve the functional safety of the vehicle computing system.

[0050] Optionally, in one embodiment of this application, the fault-tolerant scheduling module 200 includes: a fault detection unit, used to monitor the processor's temperature and cyclic redundancy check (CRC) error rate using a dynamic fault-tolerant scheduling algorithm, and to determine the processor's fault condition when the temperature exceeds a preset temperature threshold or the CRC error rate exceeds a preset error threshold; and a task migration unit, used to perform task migration operations based on the fault condition and according to task priority.

[0051] It is understood that the preset temperature threshold in the embodiments of this application can be a threshold of 105°C, and the preset error threshold can be a threshold value.

[0052] The FTDR scheduling formula is: Task priority P = α * (criticality) + β * (remaining time)

[0053] Where α = 0.7 and β = 0.3, the weights are dynamically adjusted.

[0054] For example, the fault detection unit in this embodiment uses the Dynamic Fault-Tolerant Scheduling (FTDR) algorithm for fault detection: it monitors the processor temperature and CRC error rate, and determines the processor's fault condition when the temperature exceeds a threshold of 105°C or the CRC error rate exceeds a threshold. The task migration unit in this embodiment can perform task priority calculation.

[0055] P = 0.7 × Criterion coefficient + 0.3 × (1 - Remaining time / Total time)

[0056] Response time < 0.1ms (triggered by hardware interrupt)

[0057] The dynamic fault-tolerant switching in this embodiment includes:

[0058] 1. Fault detection:

[0059] The node temperature of the Neural Processing Unit (NPU) exceeded the threshold (>105℃);

[0060] 2. Task migration:

[0061] The FTDR algorithm migrates the Simultaneous Localization and Mapping (SLAM) task to a backup FPGA node (taking 0.1ms).

[0062] Unleash nuclear computing power (NPU cores increased from 8 to 32);

[0063] 3. Energy efficiency regulation:

[0064] Reducing the power supply voltage to the faulty node to 0.5V reduces power consumption by 80%.

[0065] The embodiments of this application can achieve millisecond-level detection and judgment of hardware faults by real-time monitoring of multi-dimensional operating status parameters such as processor temperature and CRC error rate, avoiding system failure due to single point of failure and significantly improving the functional safety of the vehicle computing system.

[0066] The energy efficiency management module 300 is used to dynamically adjust the power supply voltage and computing power allocation of the target computing resources based on at least one task migration operation, so as to release reserve computing power and suppress energy efficiency fluctuations in the event of a processor failure.

[0067] Specifically, the energy efficiency management module 300 in this embodiment can dynamically adjust the power supply voltage and computing power allocation of the target computing resources. In the event of a processor failure, it releases reserve computing power and suppresses energy efficiency fluctuations, enabling the system to maintain high computing power output even under single-point failures. This ensures the continuous execution capability of critical autonomous driving tasks, actively suppresses energy efficiency fluctuations caused by sudden load changes during task migration, and avoids localized overheating through dynamic computing power allocation, thereby improving resource utilization and thermal management capabilities. This application is applicable to L4 / L5 level autonomous driving scenarios, achieving low-power parallel computing through spintronic devices and combining a dynamic fault-tolerant mechanism to ensure system reliability.

[0068] Furthermore, the basis for implementing energy efficiency standards may include:

[0069] 1. Definition of Energy Efficiency Ratio: Energy Efficiency Ratio (TOPS / W) = Computing Power (Tera Operations Per Second, TOPS) / Power Consumption (W);

[0070] 2. Test conditions

[0071] Ambient temperature 25℃±2℃, input voltage 12V±0.1V;

[0072] 3. Test Dataset

[0073] KITTI Autonomous Driving Scenario Library (including urban roads, highways, and nighttime scenarios).

[0074] Optionally, in one embodiment of this application, the energy efficiency management module 300 includes: an increase unit, used to increase the number of processor cores from a preset value to a reserve value, and increase the actual voltage value to a target voltage value; and a resource reallocation unit, used to release reserve computing power in the event of a processor failure based on the reserve value and the target voltage value, and control energy efficiency fluctuations through an adaptive voltage regulation algorithm to perform resource reallocation operations.

[0075] It is understood that the preset value in the embodiments of this application can be 8, the spare value can be 24, and the target voltage value can be 12V.

[0076] In actual implementation, the embodiments of this application can perform resource reallocation, release reserve computing power (NPU core count from 8→24, voltage increased to 12V), and suppress energy efficiency fluctuations by >95% (achieved through an adaptive voltage adjustment algorithm).

[0077] The embodiments of this application can work in deep collaboration with fault tolerance mechanisms to form a closed loop of "fault detection → task migration → computing power expansion → energy efficiency control", thereby comprehensively improving the reliability, real-time performance and energy efficiency stability of the vehicle-mounted heterogeneous computing system.

[0078] Optionally, in one embodiment of this application, the energy efficiency management module 300 further includes: a voltage regulation unit, used to dynamically adjust the power supply voltage of computing resources using an adaptive voltage regulation algorithm; and a computing power scheduling unit, used to coordinate the computing power allocation of each computing node based on task priority and the real-time distribution of target computing resources.

[0079] In actual implementation, the embodiments of this application can dynamically adjust the power supply voltage of the target computing resources based on the processor's running status and task load changes using an adaptive voltage adjustment algorithm. Based on task priority and the real-time distribution status of the target computing resources, the computing power allocation between each computing node can be coordinated. The task priority is calculated by weighting the task criticality coefficient and the remaining execution time.

[0080] This application embodiment can comprehensively consider task priority and resource distribution to achieve system allocation, improve system practicality and resource utilization, achieve breakthrough in energy efficiency, predict 50 TOPS / W (INT8 dense computing power), which is about 127% higher than the existing solution, improve fault tolerance performance, with a fault switching latency of 0.1ms, support for continuous 3-node fault recovery, and have compatibility, supporting 16 types of sensors (including LiDAR and V2X).

[0081] Specifically, the working principle of the vehicle-mounted heterogeneous computing system in this application embodiment can be described in detail with reference to Figures 2 to 4, using a specific embodiment as an example.

[0082] Figure 2 shows the system architecture, which consists of three layers: neuromorphic computing layer (magnetic material array), dynamic scheduling layer (FTDR algorithm), and energy efficiency management layer (voltage regulation module).

[0083] Furthermore, Figure 3 is a schematic diagram of the neural synapse structure, which can show the cross-section of the W / Pt / Co / NiO / Pt heteromaterial and mark the spin electron transport path.

[0084] Furthermore, Figure 4 is a dynamic fault tolerance flowchart, which includes three stages: fault detection, task migration, and resource reallocation.

[0085] It should be noted that, in terms of hardware replacement, quantum computing modules can replace the neuromorphic layer (requires a low-temperature environment, reducing latency to 0.05ms), and silicon carbide (SiC) power supply can replace gallium nitride (efficiency improvement of 5%, cost increase of 20%). In terms of algorithm replacement, a distributed fault-tolerant strategy based on federated learning can be adopted (enhanced privacy protection, response time increased to 0.3ms).

[0086] The in-vehicle heterogeneous computing system proposed in this application can be based on a neuromorphic computing hardware architecture. It stabilizes the magnetization direction through antiferromagnetic ordering, eliminating the need for an external magnetic field. Magnetization reversal is driven by spin orbital moments to switch energy, and neuromorphic synapse design is implemented. Based on the Dynamic Fault-Tolerant Scheduling (FTDR) algorithm, during fault detection, processor temperature and CRC error rate are monitored, task priority is calculated, and task migration is completed, achieving dynamic fault-tolerant switching under sudden failures. Based on resource reallocation, reserve computing power is released, energy efficiency fluctuations are suppressed, and energy efficiency management is adopted to improve the energy efficiency ratio of heterogeneous computing. This solves the problems of traditional architectures failing to balance energy efficiency and fault tolerance, failing to utilize low-power technologies such as spintronics to simulate the parallel processing characteristics of the human brain, and the resource waste and low efficiency caused by static fault-tolerant strategies (such as task redundancy replication). It also addresses the contradiction between high power consumption (>800W) and low reliability (switching latency >200ms) in in-vehicle systems.

[0087] Next, the vehicle-mounted heterogeneous computing method proposed according to the embodiments of this application is described with reference to the accompanying drawings.

[0088] Figure 5 is a flowchart illustrating the vehicle-mounted heterogeneous computing method according to an embodiment of this application.

[0089] As shown in Figure 5, the vehicle-mounted heterogeneous computing method includes the following steps:

[0090] In step S501, lidar point cloud data is received, and a heterogeneous material magnetic synapse array driven by spin orbit moment is used to perform convolution operation on the lidar point cloud data to generate obstacle detection results.

[0091] In step S502, based on the obstacle detection results, the processor's operating status is monitored to determine the processor's fault condition according to the operating status, and based on the fault condition, at least one task migration operation is executed according to the task priority.

[0092] In step S503, based on at least one task migration operation, the power supply voltage and computing power allocation of the target computing resources are dynamically adjusted to release reserve computing power and suppress energy efficiency fluctuations in the event of a processor failure.

[0093] Optionally, in one embodiment of this application, receiving lidar point cloud data and performing convolution operations on the lidar point cloud data using a spin-orbit-driven heterogeneous material magnetic synapse array to generate obstacle detection results includes: stabilizing the magnetization direction through antiferromagnetic ordering, and based on the magnetization direction, driving magnetization reversal using a spin-orbit-driven array to generate magnetization state switching data; adjusting the weights according to a preset synaptic weight formula based on the magnetization state switching data to generate final weights; and performing convolution operations using the heterogeneous material magnetic synapse array based on the final weights to generate obstacle detection results.

[0094] Optionally, in one embodiment of this application, the processor's operating status is monitored to determine the processor's fault condition based on the operating status, and at least one task migration operation is performed based on the fault condition and task priority. This includes: monitoring the processor's temperature and cyclic redundancy check (CRC) error rate using a dynamic fault-tolerant scheduling algorithm, and determining the processor's fault condition if the temperature exceeds a preset temperature threshold or the CRC error rate exceeds a preset error threshold; and performing a task migration operation based on the fault condition and task priority.

[0095] Optionally, in one embodiment of this application, the power supply voltage and computing power allocation of the target computing resources are dynamically adjusted based on at least one task migration operation to release reserve computing power and suppress energy efficiency fluctuations in the event of a processor failure. This includes: increasing the number of processor cores from a preset value to a reserve value and increasing the actual voltage value to a target voltage value; based on the reserve value and the target voltage value, releasing reserve computing power in the event of a processor failure, and controlling energy efficiency fluctuations through an adaptive voltage adjustment algorithm to perform resource reallocation operations.

[0096] Optionally, in one embodiment of this application, the power supply voltage and computing power allocation of the target computing resources are dynamically adjusted based on at least one task migration operation to release reserve computing power and suppress energy efficiency fluctuations in the event of a processor failure. The method further includes: dynamically adjusting the power supply voltage of the computing resources using an adaptive voltage adjustment algorithm; and coordinating the computing power allocation of each computing node based on task priority and the real-time distribution of the target computing resources.

[0097] It should be noted that the foregoing explanation of the embodiment of the vehicle-mounted heterogeneous computing system also applies to the vehicle-mounted heterogeneous computing method of this embodiment, and will not be repeated here.

[0098] The in-vehicle heterogeneous computing method proposed in this application can be based on a neuromorphic computing hardware architecture. It stabilizes the magnetization direction through antiferromagnetic ordering, eliminating the need for an external magnetic field. Magnetization reversal is driven by spin orbital moments to switch energy, and neuromorphic synapse design is implemented. Based on the Dynamic Fault-Tolerant Scheduling (FTDR) algorithm, during fault detection, processor temperature and CRC error rate are monitored to calculate task priorities and complete task migration, achieving dynamic fault-tolerant switching under sudden failures. Resource reallocation releases reserve computing power, suppresses energy efficiency fluctuations, and employs energy efficiency management to improve the energy efficiency ratio of heterogeneous computing. This solves the problems of traditional architectures failing to balance energy efficiency and fault tolerance, failing to utilize low-power technologies such as spintronics to simulate the parallel processing characteristics of the human brain, and the resource waste and inefficiency caused by static fault-tolerance strategies (such as task redundancy replication). It also addresses the contradiction between high power consumption (>800W) and low reliability (switching latency >200ms) in in-vehicle systems.

[0099] Figure 6 is a structural schematic diagram of a vehicle provided in an embodiment of this application. The vehicle may include:

[0100] The memory 601, the processor 602, and the computer program stored on the memory 601 and capable of running on the processor 602.

[0101] When the processor 602 executes the program, it implements the vehicle-mounted heterogeneous computing method provided in the above embodiments.

[0102] Furthermore, the vehicle also includes:

[0103] Communication interface 603 is used for communication between memory 601 and processor 602.

[0104] The memory 601 is used to store computer programs that can run on the processor 602.

[0105] The memory 601 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0106] If the memory 601, processor 602, and communication interface 603 are implemented independently, they can be interconnected via a bus to communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of representation, only one thick line is used in Figure 6, but this does not indicate that there is only one bus or one type of bus.

[0107] Optionally, in a specific implementation, if the memory 601, processor 602, and communication interface 603 are integrated on a single chip, then the memory 601, processor 602, and communication interface 603 can communicate with each other through an internal interface.

[0108] The processor 602 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of this application.

[0109] This embodiment also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the above-described vehicle-mounted heterogeneous computing method.

[0110] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0111] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0112] Any process or method described in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or N executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0113] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) with one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), fiber optic devices, and compact disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, since the program can be obtained electronically by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0114] It should be understood that the various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, the N steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0115] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0116] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0117] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of this application.

Claims

1. A vehicle-mounted heterogeneous computing system, wherein, include: The neuromorphic computing module is used to receive lidar point cloud data and perform convolution operations on the lidar point cloud data using a heterogeneous material magnetic synapse array driven by spin orbital moments to generate obstacle detection results. The fault-tolerant scheduling module is used to monitor the processor's operating status based on the obstacle detection results, determine the processor's fault status based on the operating status, and perform at least one task migration operation based on the fault status and task priority. An energy efficiency management module is used to dynamically adjust the power supply voltage and computing power allocation of the target computing resources based on the at least one task migration operation, so as to release reserve computing power and suppress energy efficiency fluctuations in the event of a processor failure.

2. The vehicle-mounted heterogeneous computing system according to claim 1, wherein, The neuromorphic computing module includes: A magnetization reversal unit is used to stabilize the magnetization direction through antiferromagnetic order, and based on the magnetization direction, drive the magnetization reversal using the spin orbital moment to generate magnetization state switching data. The weight adjustment unit is used to adjust the weights based on the magnetization state switching data and according to a preset synaptic weight formula to generate the final weights. A convolution operation unit is used to perform convolution operations based on the final weights using the heterogeneous material magnetic synapse array to generate the obstacle detection results.

3. The vehicle-mounted heterogeneous computing system according to claim 1, wherein, The fault-tolerant scheduling module includes: The fault detection unit is used to monitor the temperature and cyclic redundancy check (CRC) error rate of the processor using a dynamic fault-tolerant scheduling algorithm, and to determine the fault condition of the processor when the temperature exceeds a preset temperature threshold or the CRC error rate exceeds a preset error threshold. The task migration unit is used to perform the task migration operation based on the fault condition and the task priority.

4. The vehicle-mounted heterogeneous computing system according to claim 1, wherein, The energy efficiency management module includes: An addition unit is used to increase the number of processor cores from a preset value to a spare value, and to increase the actual voltage value to a target voltage value; The resource reallocation unit is used to release the reserve computing power in the event of a processor failure, based on the reserve value and the target voltage value, and to control energy efficiency fluctuations through an adaptive voltage regulation algorithm to perform the resource reallocation operation.

5. The vehicle-mounted heterogeneous computing system according to claim 4, wherein, The energy efficiency management module also includes: A voltage regulation unit is used to dynamically adjust the power supply voltage of the computing resources using the adaptive voltage regulation algorithm. The computing power scheduling unit is used to coordinate the allocation of computing power among various computing nodes based on the task priority and the real-time distribution of the target computing resources.

6. A vehicle-mounted heterogeneous computing method, wherein, Includes the following steps: The system receives lidar point cloud data and performs convolution operations on the lidar point cloud data using a heterogeneous material magnetic synapse array driven by spin orbital moments to generate obstacle detection results. Based on the obstacle detection results, the processor's operating status is monitored to determine the processor's fault condition according to the operating status, and based on the fault condition, at least one task migration operation is executed according to the task priority. Based on the at least one task migration operation, the power supply voltage and computing power allocation of the target computing resources are dynamically adjusted to release reserve computing power and suppress energy efficiency fluctuations in the event of a processor failure.

7. The vehicle-mounted heterogeneous computing method according to claim 6, wherein, The process of receiving lidar point cloud data and performing convolution operations on the lidar point cloud data using a spin-orbit-driven heterogeneous material magnetic synapse array to generate obstacle detection results includes: The magnetization direction is stabilized by antiferromagnetic order, and based on the magnetization direction, the magnetization is reversed by using the spin orbital moment to generate magnetization state switching data. Based on the magnetization state switching data, the weights are adjusted according to a preset synaptic weight formula to generate the final weights; Based on the final weights, convolution operations are performed using the heterogeneous material magnetic synapse array to generate the obstacle detection results.

8. The vehicle-mounted heterogeneous computing method according to claim 6, wherein, The monitoring processor's operating status is used to determine the processor's fault condition based on the operating status, and based on the fault condition, at least one task migration operation is performed according to task priority, including: The processor's temperature and cyclic redundancy check (CRC) error rate are monitored using a dynamic fault-tolerant scheduling algorithm. If the temperature exceeds a preset temperature threshold or the CRC error rate exceeds a preset error threshold, the processor's fault condition is determined. Based on the aforementioned fault condition, the task migration operation is performed according to the task priority.

9. A vehicle, wherein, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, the processor executing the program to implement the in-vehicle heterogeneous computing method as described in any one of claims 5-8.

10. A computer-readable storage medium having a computer program stored thereon, wherein, The program is executed by the processor to implement the in-vehicle heterogeneous computing method as described in any one of claims 5-8.