Methods, devices, chips and storage media for virtualizing computing power on construction machinery vehicles

Abstract

This invention relates to the field of automotive embedded technology, and discloses a method, device, chip, and storage medium for virtualizing onboard computing power in engineering machinery. It is applied to an onboard main control chip, which includes a main computing power module and a computing power expansion module. The main computing power module includes a central processing unit and a core computing power unit, while the computing power expansion module includes multiple extended computing power units. The method includes: acquiring computing tasks from microservice software running within the central processing unit; allocating target computing power resources from the core computing power unit or multiple extended computing power units according to a preset computing power allocation queue; and calling the target computing power resources to execute the computing tasks. This invention virtualizes onboard computing power through the main computing power module and the computing power expansion module, enabling unified scheduling and management of computing power resources. It supports elastic expansion of computing power, flexibly meeting the computing power requirements of different levels of intelligent driving and operational scenarios, supporting service-oriented software development, and increasing the flexibility of software and algorithm development.

Description

Technical Field

[0001] This invention relates to the field of vehicle-mounted embedded technology, specifically to a method, device, chip, and storage medium for virtualizing computing power in engineering machinery vehicles. Background Technology

[0002] With the sweeping wave of digitalization and intelligentization, improving production safety, increasing operational efficiency, and achieving unmanned or minimally manned operation of construction machinery in mining, port, and industrial park settings have become core demands for enterprises and customers in recent years. Therefore, onboard computing platforms specifically designed for construction machinery are the primary solution for future intelligent driving and operation. Currently, intelligent equipment in the construction machinery field suffers from problems such as insufficient chips, limited computing power, lack of software, and difficulty in use. To address this, some technologies use a main computing module paired with multiple extended computing modules to form a system-on-a-chip (SoC) capable of expanding computing power. Current methods primarily involve activating only the extended computing module needed for a given task and refraining from activating it when not in use. However, in many scenarios, the architectures of these extended computing modules may differ, requiring AI algorithms to consider communication issues between these heterogeneous modules when using extended computing resources. This results in an exclusive mode for accessing extended resources, calling only one module at a time without dynamic adjustment, which is inconvenient. Summary of the Invention

[0003] In view of this, the present invention provides a method, device, chip and storage medium for virtualizing onboard computing power of construction machinery, in order to solve the problem of inflexible expansion of computing power for autonomous driving tasks of construction machinery.

[0004] In a first aspect, the present invention provides a method for virtualizing computing power on an engineering machinery vehicle, applied to an on-board main control chip. The on-board main control chip includes a main computing power module and a computing power expansion module. The main computing power module includes a central processing unit and a core computing power unit, and the computing power expansion module includes multiple expansion computing power units. The method includes:

[0005] Retrieve computing tasks from microservice software running within the central processing unit;

[0006] The target computing power resources are allocated from the core computing power unit or multiple extended computing power units according to the preset computing power allocation queue;

[0007] Call upon the target computing resources to execute computing tasks.

[0008] The vehicle-mounted computing power virtualization method for engineering machinery provided in this invention acquires the computing tasks of microservice software through an on-board main control chip including a main computing power module and a computing power expansion module. It allocates target computing power resources from the core computing power unit of the main computing power module or multiple extended computing power units of the computing power expansion module according to a preset computing power allocation queue, and then calls upon the target computing power resources to execute the computing tasks. This invention virtualizes on-board computing power through the main computing power module and the computing power expansion module, and uniformly schedules and manages computing power resources according to the computing tasks of the microservice software. This supports elastic expansion of computing power, flexibly meets the computing power requirements of different levels of intelligent driving and operational scenarios, supports service-oriented software development, and increases the flexibility of software and algorithm development.

[0009] In one optional implementation, after calling the target computing resources to perform the computing task, the method further includes: different microservice software communicating the computing results through middleware communication software deployed within the central processing unit.

[0010] This invention deploys middleware communication software to enable communication between different microservice software, supports service-oriented software development, and can combine different software service functions to achieve more complex intelligent driving and operation scenarios.

[0011] In one optional implementation, the process of determining the preset computing power allocation queue includes: sequentially setting computing power numbers for the core computing power unit and multiple extended computing power units, with the core computing power unit having the smallest computing power number and the computing power numbers of the multiple extended computing power units increasing sequentially; and sequentially writing the computing power numbers into the first-in-first-out queue in ascending order to determine the preset computing power allocation queue.

[0012] This invention assigns computing power numbers to each computing power unit, enabling computing power allocation based on the first-in-first-out principle in the computing power allocation queue according to the computing power number. This allows users to perform tasks without needing to worry about whether the target computing power resources they call come from the main computing power module or the computing power extension module operating system. In other words, it eliminates the need to focus on the allocation of underlying computing resources, simplifying the software development process and increasing the flexibility of software and algorithm development and deployment.

[0013] In one optional implementation, the target computing power resource is allocated from the core computing power unit or multiple extended computing power units according to a preset computing power allocation queue, including: obtaining the computing power number from the preset computing power allocation queue according to the first-in-first-out principle; determining the core computing power unit or extended computing power unit that is not performing computing tasks according to the computing power number, and using it as the target computing power resource.

[0014] This invention can retrieve the computing power number of an idle computing power unit through a computing power allocation queue, thereby executing computing tasks.

[0015] In one optional implementation, after calling the target computing resources to execute the computing task, the method further includes: putting the computing power number corresponding to the core computing power unit or extended computing power unit that executed the computing task back into the preset computing power allocation queue.

[0016] This invention improves the utilization efficiency of computing resources by returning the corresponding computing power number to the computing power allocation queue after the computing power resources have completed their computing tasks.

[0017] In one alternative implementation, the method further includes: if the preset computing power allocation queue is empty, pausing the response until any computing power number is returned to the preset computing power allocation queue.

[0018] This invention pauses the response of computing tasks when there are no idle computing resources, which enables computing tasks to proceed in an orderly manner and prevents computing chaos.

[0019] In one optional implementation, the process of calling target computing resources to perform computing tasks further includes: real-time monitoring of the computing power utilization rate of the core computing power unit and multiple extended computing power units; if the computing power utilization rate is less than a preset threshold, deleting the computing power numbers corresponding to a preset number of computing power resources in the multiple extended computing power units from the preset computing power allocation queue according to the computing power utilization rate; if the computing power utilization rate is greater than the preset threshold, adding the computing power numbers corresponding to the offline computing power resources to the preset computing power allocation queue according to the computing power utilization rate.

[0020] This invention monitors the utilization rate of computing resources, enabling the deployment of an appropriate number of computing resources based on the level of intelligent driving and operational scenarios, thereby preventing the waste of computing resources and improving resource utilization.

[0021] Secondly, this invention provides an onboard computing power virtualization device for engineering machinery, applied to an onboard main control chip. The onboard main control chip includes a main computing power module and a computing power expansion module. The main computing power module includes a central processing unit and a core computing power unit. The computing power expansion module includes multiple expansion computing power units. The device includes:

[0022] The task acquisition module is used to acquire computing tasks running on microservice software within the central processing unit;

[0023] The computing power analysis module is used to allocate target computing power resources from the core computing power unit or multiple extended computing power units according to the preset computing power allocation queue;

[0024] The computing power invocation module is used to invoke target computing resources to execute computing tasks.

[0025] The vehicle-mounted computing power virtualization device for engineering machinery provided in this invention acquires computing tasks from microservice software through an on-board main control chip comprising a main computing power module and a computing power expansion module. It allocates target computing power resources from the core computing power unit of the main computing power module or multiple extended computing power units of the computing power expansion module according to a preset computing power allocation queue, and then invokes the target computing power resources to execute the computing tasks. This invention virtualizes on-board computing power through the main computing power module and the computing power expansion module, supporting elastic expansion of computing power. It can flexibly meet the computing power requirements of different levels of intelligent driving and operational scenarios. Furthermore, by allocating computing power resources according to microservice software, it supports service-oriented software development, increasing the flexibility of software and algorithm development.

[0026] Thirdly, the present invention provides an in-vehicle main control chip, characterized in that it comprises: a general-purpose chip and multiple computing power expansion chips. The general-purpose chip includes a central processing unit (CPU) and a core computing power chip. The CPU is used to perform computing power resource scheduling and logical calculations of the vehicle. The core computing power chip is used to perform inference calculations of the vehicle. The computing power expansion chips are used to expand the computing power of the core computing power chip. The CPU establishes a communication connection with the core computing power chip, and each computing power expansion chip establishes a communication connection with the core computing power chip. The CPU, the core computing power chip, and each computing power expansion chip are packaged into an in-vehicle main control chip using chip packaging technology. The CPU stores computer instructions, and the CPU executes the computer instructions to perform the methods of the first aspect and any one of the first aspects.

[0027] Fourthly, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to perform the method described in the first aspect or any corresponding embodiment thereof. Attached Figure Description

[0028] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0029] Figure 1 This is a flowchart illustrating a method for virtualizing onboard computing power in engineering machinery according to an embodiment of the present invention.

[0030] Figure 2 This is a schematic diagram of the vehicle-mounted main control chip packaging for the vehicle-mounted computing power virtualization method for engineering machinery according to an embodiment of the present invention;

[0031] Figure 3This is a schematic diagram of the operating system of an intelligent connected robot for engineering machinery according to the virtualization method of on-board computing power for engineering machinery according to an embodiment of the present invention.

[0032] Figure 4 This is a schematic diagram of microservice software communication in the on-board computing power virtualization method for engineering machinery according to an embodiment of the present invention;

[0033] Figure 5 This is a schematic diagram illustrating the runtime engine computing power resource call of the on-board computing power virtualization method for engineering machinery according to an embodiment of the present invention;

[0034] Figure 6 This is a flowchart illustrating another method for virtualizing onboard computing power of engineering machinery according to an embodiment of the present invention;

[0035] Figure 7 This is a schematic diagram of the computing power allocation queue for another method of virtualizing computing power on engineering machinery according to an embodiment of the present invention;

[0036] Figure 8 This is a flowchart illustrating another method for virtualizing onboard computing power of engineering machinery according to an embodiment of the present invention;

[0037] Figure 9 This is a schematic diagram of the computing resource monitoring process of another method for virtualizing computing power on engineering machinery according to an embodiment of the present invention.

[0038] Figure 10 This is a schematic diagram of the structure of an onboard computing power virtualization device for engineering machinery according to an embodiment of the present invention. Detailed Implementation

[0039] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, 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.

[0040] According to an embodiment of the present invention, an embodiment of a method for virtualizing computing power on construction machinery is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0041] This embodiment provides a method for virtualizing the computing power of engineering machinery vehicles, which can be used in the aforementioned vehicle-mounted main control chip. The vehicle-mounted main control chip includes a main computing power module and a computing power expansion module. The main computing power module includes a central processing unit and a core computing power unit, and the computing power expansion module includes multiple expansion computing power units. Figure 1 This is a flowchart of a method for virtualizing onboard computing power in engineering machinery according to an embodiment of the present invention, such as... Figure 1 As shown, the process includes the following steps:

[0042] Step S101: Obtain the computing tasks of the microservice software running in the central processing unit.

[0043] Step S102: Allocate target computing resources from the core computing unit or multiple extended computing units according to the preset computing power allocation queue.

[0044] Step S103: Call the target computing resources to execute the computing task.

[0045] Specifically, in this embodiment of the invention, the vehicle-mounted main control chip adopts a "1+1+N" chipplet heterogeneous splicing technology to integrate multiple chips with different functions into one piece. The first "1" represents one CPU-Chiplet, which carries the main software of the vehicle-mounted computing platform, deploys the operating system of the intelligent connected robot for engineering machinery, and realizes virtualized access to the computing resources within the computing platform and efficient middleware communication technology. The second "1" represents a Hub-Chiplet, and "N" represents multiple selectable NPU-Chiplets (N = 0, 1, 2...), realizing elastic scaling of computing power. For example... Figure 2As shown, the General Chiplet, composed of CPU-Chiplet and Hub-Chiplet, is the main computing power module. The CPU-Chiplet is the central processing unit, the Hub-Chiplet is the core computing power unit, and the Extend Chiplet, composed of multiple NPU-Chiplet, is the computing power expansion module. Each NPU-Chiplet is an extended computing power unit. This invention further utilizes System-In-Package (SIP) technology to encapsulate various chips in the main computing module and computing power expansion module into an onboard main control chip. This enables a single hardware circuit board to meet the computing power requirements for intelligent driving and operations in various scenarios from L2 to L4. Specifically, a single GeneralChiplet can at least meet the computing power requirements for intelligent driving and operations in L2 scenarios, with AI computing power located within the Hub-Chiplet. The General Chiplet, combined with Extend Chiplet, can meet the computing power requirements for intelligent driving and operations in L4 scenarios. For example, expanding the General Chiplet with two NPU-Chiplets can meet the requirements for L2+ scenarios, and expanding it with six NPU-Chiplets can meet the requirements for L3 scenarios. These are merely examples and are not intended to limit the scope of the invention. Meanwhile, in this embodiment of the invention, the SIP (System In Package) package of the chiplet chip achieves high-speed UCIe interface interconnection between the CPU-Chiplet and NPU-Chiplet through the HUB-Chiplet. UCIe interconnection technology features high energy efficiency, high bandwidth density, low end-to-end latency, and robustness. Multiple NPU-Chiplets serve as coprocessing units, reducing communication latency between chiplets to the microsecond level. Deploying an intelligent connected robot operating system on the CPU-Chiplet provides a unified software communication interface, reducing software development difficulty. 2) Furthermore, the compatible design ensures that the size and specifications of the SIP (System In Package) are not affected regardless of how the Extend Chiplet is expanded. This successfully normalizes the structure and hardware of the onboard computing platform in L2-L4 level application scenarios, significantly reducing R&D and production costs.

[0046] In some alternative implementations, such as Figure 3As shown, this embodiment of the invention deploys an intelligent connected robot operating system for construction machinery on the CPU-Chiplet, used to manage various hardware resources including the General Chiplet and Extend Chiplet. This operating system is based on the Linux system kernel and has a customized operating system interface. Users can choose whether to use a graphical interface according to their own development habits; this is merely an example and not a limitation. The intelligent connected robot operating system for construction machinery in this embodiment of the invention provides intermediate communication software, runtime engine software, a customized operating system interface, a Linux Kernel, and a file system.

[0047] In some optional implementations, the intermediate communication software provided in these embodiments supports service-oriented software and algorithm development. The unified standard input / output interface provided by the middleware communication software ensures high-speed, low-latency communication. Typically, even when sharing large amounts of data among various microservice software, microsecond-level latency can be achieved. Figure 4 The microservice software examples shown, such as camera access, perception algorithm-1, perception algorithm-2, and LiDAR access, are merely illustrative and not intended to limit the scope. Each microservice can be developed independently, but all utilize the unified standard input / output interface provided by the middleware communication software. For instance, the "camera access" microservice uses the unified standard output interface provided by the middleware communication software, while the "perception algorithm-2" microservice also uses the unified standard input interface. This allows the "camera access" microservice and the "perception algorithm-2" microservice to share data at the software level. Typically, image data from the camera sensor is preprocessed within the "camera access" microservice, and then shared with the "perception algorithm-2" microservice for specific perception and recognition algorithm processing. Since each Microsoft service software requires computational resources for its calculations, this embodiment of the invention obtains the computational tasks of different Microsoft service software and allocates computational resources accordingly. By leveraging the middleware communication software provided by the intelligent connected robot operating system for construction machinery, service-oriented software and algorithm development is achieved. This reduces the complexity of software and algorithm development for intelligent connected equipment in intelligent driving / operation scenarios. Each microservice software is like a building block; different microservice software can be combined to form a complete application software based on the intelligent driving / operation scenario of the construction machinery. Furthermore, the code reusability is very high; a single software package can be applied to the functional requirements of different types of intelligent connected equipment in different intelligent driving / operation scenarios (simply by recombining the various microservice software with different topologies, which will not be elaborated further here).

[0048] In some optional implementations, the runtime engine software provided in this embodiment of the invention can virtualize the AI ​​computing power of the Hub-Chiplet in the General Chiplet and the AI ​​computing power of the NXP-Chiplet in the Extend Chiplet. 2) Users can call the software interface provided by the runtime engine software to achieve indiscriminate use of the underlying AI computing power, that is, when users call the runtime engine software to perform AI inference calculations, they do not need to care whether the AI ​​computing power of the Hub-Chiplet or the AI ​​computing power of the NPU-Chiplet is used. The intelligent connected robot operating system for engineering machinery in this embodiment of the invention relies on many modular microservice software to realize the intelligent driving / operation function in a specific scenario. These microservice software are responsible for realizing functions such as environmental perception, precise positioning, path planning, and motion control. See attached... Figure 5 As shown, a specific microservice software utilizes the computing resources on the Hub-Chiplet and NPU-Chiplet through the runtime engine in the operating system of the intelligent connected robot for construction machinery: The main thread of the microservice software runs on the CPU-Chiplet. This main thread may create several auxiliary threads during its execution. Both the main thread and auxiliary threads perform AI inference calculations according to the needs of the actual algorithm. Because the CPU-Chiplet is primarily good at resource scheduling and logical computation, all operations involving AI inference calculations are not executed using the CPU-Chiplet, but rather through the runtime engine, which calls upon the AI ​​computing resources on the Hub-Chiplet and NPU-Chiplet. Furthermore, the Hub-Chiplet and NPU-Chiplet primarily contain AI computing resources. The AI ​​computing resources on the Hub-Chiplet can meet the needs of L2-level intelligent driving / operation scenarios for intelligent connected equipment for construction machinery. Combined with the NPU-Chiplet computing resources in the Hub-Chiplet and ExtendChiplet, it can meet the needs of L4-level intelligent driving / operation scenarios for intelligent connected equipment for construction machinery.

[0049] In one optional implementation, the underlying runtime engine software of this embodiment uses a pre-set computing power allocation queue to manage and maintain the AI ​​computing power on Hub-Chiplet and NPU-Chiplet, matching target computing power resources for the computing tasks of Microsoft service software, and then calling the target computing power resources to execute the computing tasks through the runtime engine software.

[0050] The vehicle-mounted computing power virtualization method for engineering machinery provided in this invention acquires the computing tasks of microservice software through an on-board main control chip including a main computing power module and a computing power expansion module. It allocates target computing power resources from the core computing power unit of the main computing power module or multiple extended computing power units of the computing power expansion module according to a preset computing power allocation queue, and then calls upon the target computing power resources to execute the computing tasks. This invention virtualizes on-board computing power through the main computing power module and the computing power expansion module, and uniformly schedules and manages computing power resources according to the computing tasks of the microservice software. This supports elastic expansion of computing power, flexibly meets the computing power requirements of different levels of intelligent driving and operational scenarios, supports service-oriented software development, and increases the flexibility of software and algorithm development.

[0051] This embodiment provides a method for virtualizing the computing power of engineering machinery vehicles, which can be used in the aforementioned vehicle-mounted main control chip. The vehicle-mounted main control chip includes a main computing power module and a computing power expansion module. The main computing power module includes a central processing unit and a core computing power unit, and the computing power expansion module includes multiple expansion computing power units. Figure 6 This is a flowchart of a method for virtualizing onboard computing power in engineering machinery according to an embodiment of the present invention, such as... Figure 6 As shown, the process includes the following steps:

[0052] Step S601: Obtain the computing tasks of the microservice software running within the central processing unit. For details, please refer to [link to relevant documentation]. Figure 1 Step S101 of the illustrated embodiment will not be described again here.

[0053] Step S602: Allocate target computing resources from the core computing unit or multiple extended computing units according to the preset computing power allocation queue.

[0054] Specifically, step S602 includes:

[0055] Step S6021: Set computing power numbers for the core computing power unit and multiple extended computing power units in sequence. The core computing power unit has the smallest computing power number, and the computing power numbers of the multiple extended computing power units are sequentially increased. Write the computing power numbers into the first-in-first-out queue in ascending order to determine the preset computing power allocation queue.

[0056] Specifically, in embodiments of the present invention, such as Figure 7 As shown, the number of the AI ​​computing hardware unit on the Hub-Chiplet is set to "0", and the number of the AI ​​computing hardware units on the N NPU-Chiplet is set to "1", "2", ..., "N-1", "N-1". During the runtime engine initialization process, the numbers of each AI computing hardware unit on the Hub-Chiplet and NPU-Chiplet are written into a first-in-first-out (FIFO) queue in ascending order to obtain the computing power allocation queue.

[0057] Step S6022: Obtain the computing power number from the preset computing power allocation queue according to the first-in-first-out principle; determine the core computing power unit or extended computing power unit that is not performing computing tasks based on the computing power number, and use it as the target computing power resource.

[0058] Specifically, in this embodiment of the invention, when a microservice in the intelligent driving / operation application software of the intelligent connected robot operating system for engineering machinery calls the AI ​​inference interface provided by the runtime engine in a specific scenario, the runtime engine follows... Figure 7 The usage method shown retrieves the number of the first AI computing hardware unit added to the computing power allocation queue and maps it to the actual hardware resources to begin the AI ​​inference process. Specifically, in actual operation, when allocating computing resources, one computing unit can be allocated to one microservice software, or one computing unit can be allocated to multiple microservice software, depending on the computing power resources of the computing unit and the computing power requirements of the microservice software.

[0059] Step S6023: Return the computing power number corresponding to the core computing power unit or extended computing power unit that performs the computing task to the preset computing power allocation queue.

[0060] Specifically, in this embodiment of the invention, after the local AI inference calculation is completed, the runtime engine then... Figure 7 The return method shown puts the number of the AI ​​computing power hardware unit that has been used up back into the computing power allocation queue.

[0061] Step S6024: If the preset computing power allocation queue is empty, pause the response until any computing power number is put back into the preset computing power allocation queue.

[0062] Specifically, in this embodiment of the invention, when a microservice software is using a low-level AI computing hardware unit, and simultaneously another microservice software in an intelligent driving / task application calls the AI ​​inference interface provided by the runtime engine, the runtime engine continues to retrieve the number of an AI computing hardware unit from the computing power allocation queue and assigns it to an actual hardware resource to begin the AI ​​inference process. If all low-level AI computing units are currently performing AI inference, and a new microservice software calls the AI ​​inference interface provided by the runtime engine, then the computing power allocation queue is empty. It must wait until an AI inference process finishes and returns the number of the occupied AI computing hardware unit to the computing power allocation queue before it can continue responding.

[0063] Step S603: Utilize the target computing resources to execute the computation task. For details, please refer to [link to relevant documentation]. Figure 1Step S103 of the illustrated embodiment will not be described again here.

[0064] The vehicle-mounted computing power virtualization method for engineering machinery provided in this invention acquires the computing tasks of microservice software through an on-board main control chip including a main computing power module and a computing power expansion module. It allocates target computing power resources from the core computing power unit of the main computing power module or multiple extended computing power units of the computing power expansion module according to a preset computing power allocation queue, and then calls upon the target computing power resources to execute the computing tasks. This invention virtualizes on-board computing power through the main computing power module and the computing power expansion module, and uniformly schedules and manages computing power resources according to the computing tasks of the microservice software. This supports elastic expansion of computing power, flexibly meets the computing power requirements of different levels of intelligent driving and operational scenarios, supports service-oriented software development, and increases the flexibility of software and algorithm development.

[0065] This embodiment provides a method for virtualizing the computing power of engineering machinery vehicles, which can be used in the aforementioned vehicle-mounted main control chip. The vehicle-mounted main control chip includes a main computing power module and a computing power expansion module. The main computing power module includes a central processing unit and a core computing power unit, and the computing power expansion module includes multiple expansion computing power units. Figure 8 This is a flowchart of a method for virtualizing onboard computing power in engineering machinery according to an embodiment of the present invention, such as... Figure 8 As shown, the process includes the following steps:

[0066] Step S801: Obtain the computing tasks of the microservice software running within the central processing unit. For details, please refer to [link to relevant documentation]. Figure 1 Step S101 of the illustrated embodiment will not be described again here.

[0067] Step S802: Allocate target computing power resources from the core computing power unit or multiple extended computing power units according to the preset computing power allocation queue. For details, please refer to [link to relevant documentation]. Figure 1 Step S101 of the illustrated embodiment will not be described again here.

[0068] Step S803: Utilize the target computing resources to execute the computation task. For details, please refer to [link to relevant documentation]. Figure 1 Step S103 of the illustrated embodiment will not be described again here.

[0069] Step S804: Monitor the computing power utilization rate of the core computing power unit and multiple extended computing power units in real time.

[0070] Specifically, in this embodiment of the invention, if the computing power utilization rate is less than a preset threshold, such as 75%, then the computing power numbers corresponding to a preset number of computing power resources in multiple extended computing power units are deleted from the preset computing power allocation queue according to the computing power utilization rate; if the computing power utilization rate is greater than the preset threshold, such as 100%, then the computing power numbers corresponding to the offline computing power resources are added to the preset computing power allocation queue according to the computing power utilization rate. The specific process is as follows: Figure 9As shown, the onboard main control chip of the intelligent connected construction machinery equipment performs the initialization process of intelligent driving / operation functions in specific scenarios. The runtime engine in the intelligent connected construction machinery robot operating system will bring all AI computing resources on the Hub-Chiplet and NPU-Chiplet online, that is, put the corresponding AI computing hardware unit numbers into the computing power allocation queue for management and maintenance. When the intelligent driving / operation function starts running, the runtime engine software will monitor and statistically analyze the consumption of the underlying AI computing power of the onboard computing platform by each microservice software that constitutes the intelligent driving / operation application software in real time. Based on the statistically analyzed AI computing power utilization rate and other indicators, it will determine whether there is a waste of hardware resources. For example, if four AI computing hardware units are online, but the total resource utilization rate of each AI computing hardware unit is less than 50%, then two of the four online AI computing hardware units can be taken offline. This is just an example and is not a limitation.

[0071] In some optional implementations, embodiments of the present invention provide two methods for calculating computing power utilization. The first method involves determining the number of microservice software currently running and summing up the individual AI computing power consumed by each microservice software to obtain the resource utilization rate of the total computing power. For example, if a total of 4 AI computing hardware units are online, assuming each AI computing hardware unit consumes 8-TOPs, then the total online AI computing power is 4×8=32TOPs. Further assuming that 5 microservice software are currently running, with the first consuming 2TOPs, the second consuming 3TOPs, the third consuming 4TOPs, the fourth consuming 5TOPs, and the fifth consuming 6TOPs, then the total AI computing power consumption is: 2+3+4+5+6=20TOPs. Therefore, the computing power utilization rate can be calculated as: 20 / 32=2.5 / 4, meaning that only 2.5 of the 4 online AI computing hardware units are used, leaving (4-2.5)=1.5 AI computing hardware units with surplus computing power. However, in this case, the number of offline units cannot be 1.5; it can only be 1 offline. The second method: Calculate the resource utilization rate of each of the N online AI computing hardware units individually, and then calculate the total AI computing power consumed. This is the cumulative product of the computing power of each individual AI computing hardware unit and its own resource utilization rate. For example, if four 8-TOPS AI computing hardware units are online, and the first unit consumes 50%, the second 25%, the third 25%, and the fourth 30%, then the total AI computing power consumed is: 8 × 50% + 8 × 25% + 8 × 25% + 8 × 30% = 10.4TOPS.

[0072] In some optional implementations, if the current intelligent connected construction machinery equipment performs intelligent driving / operation operations with low resource utilization of the online AI computing hardware units, it is necessary to take some AI computing hardware units offline. That is, temporarily remove the corresponding numbers of these AI computing hardware units from the aforementioned computing power allocation queue, and also cut off the power supply of these AI computing hardware units to achieve low-power processing and save energy consumption.

[0073] In some optional implementations, if the resource utilization rate of the AI ​​computing hardware units already online on the onboard computing platform of the current intelligent connected construction machinery equipment is not low, it is also necessary to determine whether to re-bring the temporarily offline AI computing hardware units online using another threshold, such as 100%. If the total load rate of the already online AI computing hardware units in the current scenario of intelligent driving / operation of the intelligent connected construction machinery equipment is higher than 100%, then the number of NPU-Chiplets in the Extend Chiplet to be re-bringed is determined according to the total load rate and other indicators. That is, the corresponding AI computing hardware unit is first powered on and initialized individually, and then its corresponding number is added to the aforementioned computing power allocation queue. For example, if four AI computing hardware units are currently online, and the total AI resource load rate is close to 110% over a continuous period, then one temporarily offline AI computing unit needs to be re-bringed online. This is just an example and not a limitation.

[0074] In some optional implementations, after the runtime engine software on the intelligent connected robot operating system of the construction machinery dynamically takes the AI ​​computing hardware units at the bottom layer of the vehicle computing platform offline or online, the currently online AI computing hardware units are managed and maintained by the runtime engine, and are constantly responding to the various microservice AI inference computing needs of the intelligent driving / operation application software of the intelligent connected equipment of the construction machinery in various changing scenario environments.

[0075] This invention virtualizes onboard computing power through a main computing power module and a computing power expansion module, and uniformly schedules and manages computing resources according to the computing tasks of microservice software. It can support elastic expansion of computing power, flexibly meet the computing power requirements of different levels of intelligent driving and operation scenarios, support service-oriented software development, and increase the flexibility of software and algorithm development.

[0076] This embodiment also provides an on-board computing power virtualization device for engineering machinery. This device is used to implement the above embodiments and preferred embodiments, and details already described will not be repeated. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0077] This embodiment provides a vehicle-mounted computing power virtualization device for engineering machinery, such as... Figure 10 As shown, this is applied to an automotive main control chip. The automotive main control chip includes a main computing power module and a computing power expansion module. The main computing power module includes a central processing unit and a core computing power unit. The computing power expansion module includes multiple expansion computing power units. The device includes:

[0078] The task acquisition module 1001 is used to acquire computing tasks running on microservice software within the central processing unit.

[0079] The computing power analysis module 1002 is used to allocate target computing power resources from the core computing power unit or multiple extended computing power units according to the preset computing power allocation queue;

[0080] The computing power invocation module 1003 is used to invoke target computing power resources to execute computing tasks.

[0081] In some alternative implementations, it also includes:

[0082] The software communication module is used for different microservice software to communicate computation results through middleware communication software deployed within the central processing unit.

[0083] In some alternative implementations, the computing power analysis module 1002 includes:

[0084] The queue setting unit is used to sequentially set computing power numbers for the core computing power unit and multiple extended computing power units. The core computing power unit has the smallest computing power number, and the computing power numbers of multiple extended computing power units are sequentially increased. The computing power numbers are written into the first-in-first-out queue in ascending order to determine the preset computing power allocation queue.

[0085] The computing power allocation unit is used to obtain computing power numbers from the preset computing power allocation queue according to the first-in-first-out principle; and to determine the core computing power units or extended computing power units that are not performing computing tasks based on the computing power numbers, and to use them as target computing power resources.

[0086] The computing power return unit is used to return the computing power number corresponding to the core computing power unit or extended computing power unit that performs the computing task to the preset computing power allocation queue.

[0087] The computing power waiting unit is used to pause the response if the preset computing power allocation queue is empty, until any computing power number is put back into the preset computing power allocation queue.

[0088] In some alternative implementations, it also includes:

[0089] The computing power resource adjustment module is used to monitor the computing power utilization rate of the core computing power unit and multiple extended computing power units in real time. If the computing power utilization rate is less than the preset threshold, the computing power number corresponding to a preset number of computing power resources in multiple extended computing power units will be deleted from the preset computing power allocation queue according to the computing power utilization rate. If the computing power utilization rate is greater than the preset threshold, the computing power number corresponding to the offline computing power resources will be added to the preset computing power allocation queue according to the computing power utilization rate.

[0090] Further functional descriptions of the above modules and units are the same as those in the corresponding embodiments described above, and will not be repeated here.

[0091] The on-board computing virtualization device for construction machinery in this embodiment is presented in the form of functional units. Here, a unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that execute one or more software or fixed programs, and / or other devices that can provide the above functions.

[0092] This invention also provides an in-vehicle main control chip, which has the above-mentioned... Figure 10 The image shows a vehicle-mounted computing virtualization device for engineering machinery.

[0093] According to embodiments of the present invention, such as Figure 2 As shown, an in-vehicle main control chip is provided, including a general-purpose chip and multiple computing power expansion chips. The general-purpose chip includes a central processing unit (CPU) and a core computing power chip. The CPU is used to perform computing power resource scheduling and logical calculations for the vehicle. The core computing power chip is used to perform inference calculations for the vehicle. The computing power expansion chips are used to expand the computing power of the core computing power chip. The CPU establishes a communication connection with the core computing power chip, and each computing power expansion chip also establishes a communication connection with the core computing power chip. The CPU, the core computing power chip, and each computing power expansion chip are packaged into an in-vehicle main control chip using chip packaging technology. The CPU stores computer instructions and executes the computer instructions.

[0094] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code, which, when accessed and executed by the computer, processor, or hardware, implements the methods shown in the above embodiments.

[0095] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method for virtualizing onboard computing power in engineering machinery, characterized in that, The method is applied to an automotive main control chip, wherein the automotive main control chip includes a main computing power module and a computing power expansion module, the main computing power module includes a central processing unit and a core computing power unit, and the computing power expansion module includes multiple expansion computing power units. Obtain the computing tasks of the microservice software running within the central processing unit; Target computing power resources are allocated from the core computing power unit or the plurality of extended computing power units according to the preset computing power allocation queue; The target computing resources are invoked to execute the computing task; The process of determining the preset computing power allocation queue includes: The core computing power unit and the plurality of extended computing power units are sequentially assigned computing power numbers, wherein the core computing power unit is assigned the smallest computing power number, and the computing power numbers of the plurality of extended computing power units are sequentially increased. The computing power numbers are written into the first-in-first-out queue in ascending order to determine the preset computing power allocation queue; The allocation of target computing power resources from the core computing power unit or the plurality of extended computing power units according to the preset computing power allocation queue includes: The computing power number is obtained from the preset computing power allocation queue according to the first-in-first-out principle; The core computing power unit or the extended computing power unit that is not performing a computing task is identified based on the computing power number and is used as the target computing power resource. After invoking the target computing resources to execute the computing task, the method further includes: The computing power number corresponding to the core computing power unit or the extended computing power unit that performs the computing task is returned to the preset computing power allocation queue.

2. The method according to claim 1, characterized in that, After invoking the target computing resources to execute the computing task, the process also includes: Different microservice software communicate their computation results through middleware communication software deployed within the central processing unit.

3. The method according to claim 1, characterized in that, The method further includes: If the preset computing power allocation queue is empty, the response is paused until any of the computing power numbers is returned to the preset computing power allocation queue.

4. The method according to claim 1, characterized in that, The process of calling upon the target computing resources to execute the computing task also includes: Real-time monitoring of the computing power utilization rate of the core computing power unit and the multiple extended computing power units; If the computing power utilization rate is less than a preset threshold, then the computing power number corresponding to a preset number of computing power resources in the plurality of extended computing power units will be deleted from the preset computing power allocation queue according to the computing power utilization rate. If the computing power utilization rate is greater than a preset threshold, then the computing power number corresponding to the offline computing power resources is added to the preset computing power allocation queue according to the computing power utilization rate.

5. A vehicle-mounted computing power virtualization device for engineering machinery, used to implement the method described in any one of claims 1-4, characterized in that, The device is applied to an automotive main control chip, which includes a main computing power module and a computing power expansion module. The main computing power module includes a central processing unit and a core computing power unit. The computing power expansion module includes multiple expansion computing power units. The device includes: The task acquisition module is used to acquire computing tasks of microservice software running in the central processing unit; The computing power analysis module is used to allocate target computing power resources from the core computing power unit or the multiple extended computing power units according to a preset computing power allocation queue; The computing power invocation module is used to invoke the target computing power resources to execute the computing task.

6. A vehicle-mounted main control chip, characterized in that, include: The system comprises a general-purpose chip and multiple computing power expansion chips. The general-purpose chip includes a central processing unit (CPU) and a core computing power chip. The CPU is used to perform computing power resource scheduling and logical calculations for the vehicle. The core computing power chip is used to perform inference calculations for the vehicle. The computing power expansion chips are used to expand the computing power of the core computing power chip. The CPU establishes a communication connection with the core computing power chip, and each of the computing power expansion chips establishes a communication connection with the core computing power chip. The CPU, the core computing power chip, and each of the computing power expansion chips are packaged into the vehicle main control chip using chip packaging technology. The CPU stores computer instructions, and the CPU executes the computer instructions to perform the method of any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing a computer to perform the method of any one of claims 1 to 4.

Images