Automatic driving task processing system and method

By introducing request reception, control, scheduling and storage modules into the autonomous driving task processing system, dynamically dispatching tasks to task instances or nodes, solving the problems of low task execution efficiency and low resource utilization, and achieving rapid task processing and high reliability Sexual scheduling.

CN118939390APending Publication Date: 2024-11-12CHINA FAW CO LTD +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410829359.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-06-25
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

In the processing of autonomous driving tasks, traditional technology leads to low task execution efficiency, mainly because the loading and startup time of large AI models and Docker images is long, and the task scheduling depends on the cluster state, which is prone to problems such as low resource utilization and scheduling failure.

Method used

An autonomous driving task processing system is designed, including a request reception module, a control module, a scheduling module and a storage module. It generates pending tasks through task templates, dynamically schedules to task instances or target nodes in the task pool, and optimizes the task execution order. Reduce dependence on cluster state, realize task instantiation and efficient utilization of resources.

Benefits of technology

It improves task execution efficiency, shortens the loading and startup time of models and images, improves system resource utilization and robustness, reduces operational costs, and enhances task scheduling reliability and fault tolerance.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118939390A_ABST
    Figure CN118939390A_ABST
Patent Text Reader

Abstract

The invention discloses an automatic driving task processing system and method, and the method comprises the steps: receiving a task processing request through a request receiving module, and transmitting the task processing request to a control module; receiving a task processing request through a control module, generating a to-be-processed task according to the task processing request and a task template, and storing the to-be-processed task in a task queue; determining a target to-be-processed task in the task queue through a scheduling module, and scheduling the target to-be-processed task to a task instance in a task pool, or scheduling the target to-be-processed task to a target node in a processing system of the automatic driving task, so that the target node processes the target to-be-processed task; a task template, a task queue, a task pool, a task instance and state information of a to-be-processed task are stored through a storage module. The system resource utilization rate is improved, the task processing delay is shortened, the model and mirror image loading is accelerated, the dependence on the cluster state is reduced, and the automatic driving task processing efficiency and reliability are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of autonomous driving task processing, and in particular to a system and method for processing autonomous driving tasks. Background Art

[0002] With the rapid development of autonomous driving technology, building an autonomous driving data closed-loop processing system has become an urgent need. This closed-loop system needs to collect and process a large amount of data from multiple sensors (such as lidar, millimeter-wave radar, cameras, etc.) in real time, and perform complex tasks such as cleaning, conversion, storage, labeling, model training, evaluation and simulation on these data.

[0003] For most types of tasks, such as simulation and evaluation, the versions of autonomous driving models for the same batch of tasks are basically the same, but the sensor data, maps and other information used are different. In traditional technologies, each time a task is executed, the relevant image will be pulled up, even if the container used is the same.

[0004] In autonomous driving data processing, it is usually necessary to load large AI models and Docker images. These models and image files are often large in size, and it may take several minutes to load and start. If the corresponding AI model and Docker image are loaded every time a task is executed, the efficiency of task execution will be seriously affected. Summary of the invention

[0005] The present invention provides a system and method for processing an autonomous driving task to achieve processing efficiency and reliability of the autonomous driving task.

[0006] According to one aspect of the present invention, a processing system for an autonomous driving task is provided, comprising: a request receiving module, a control module, a scheduling module and a storage module, wherein the control module, the scheduling module and the storage module are respectively connected to the request receiving module, wherein:

[0007] The request receiving module is used to receive a task processing request and send the task processing request to the control module;

[0008] The control module is used to receive the task processing request, generate a task to be processed according to the task processing request and a task template, and store the task to be processed in a task queue;

[0009] The scheduling module is used to determine the target task to be processed in the task queue, and schedule the target task to be processed to a task instance in the task pool, or to a target node in the processing system of the autonomous driving task, so that the task instance or the target node processes the target task to be processed;

[0010] The storage module is used to store the task template, the task queue, the task pool, the task instance and the status information of the task to be processed.

[0011] According to another aspect of the present invention, a method for processing an autonomous driving task is provided, which is applied to a system for processing an autonomous driving task. The system comprises: a request receiving module, a control module, a scheduling module and a storage module. The control module, the scheduling module and the storage module are respectively connected to the request receiving module. The method comprises:

[0012] receiving a task processing request through the request receiving module, and sending the task processing request to the control module;

[0013] The control module receives the task processing request, generates a task to be processed according to the task processing request and a task template, and stores the task to be processed in a task queue;

[0014] Determine the target task to be processed in the task queue by the scheduling module, and schedule the target task to be processed to a task instance in a task pool, or to a target node in the processing system of the autonomous driving task, so that the task instance or the target node processes the target task to be processed;

[0015] The storage module stores the task template, the task queue, the task pool, the task instance and the status information of the pending tasks.

[0016] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0017] at least one processor; and

[0018] a memory communicatively connected to the at least one processor; wherein,

[0019] The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can perform the processing method of the autonomous driving task described in any embodiment of the present invention.

[0020] According to another aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the method for processing an autonomous driving task described in any embodiment of the present invention when executed.

[0021] The technical solution of the embodiment of the present invention is to receive a task processing request through a request receiving module, and send the task processing request to the control module; receive the task processing request through the control module, generate a task to be processed according to the task processing request and the task template, and store the task to be processed in the task queue; determine the target task to be processed in the task queue through the scheduling module, and schedule the target task to be processed to a task instance in the task pool, or to a target node in the processing system of the autonomous driving task, so that the task instance or the target node processes the target task to be processed. By processing repeatable tasks through task instances in the task pool and processing non-repetitive tasks through nodes, the problem of low efficiency in processing autonomous driving tasks in the prior art is solved, and the loading and startup time of the AI ​​model and container image required to execute the task is reduced, thereby improving the efficiency of task execution.

[0022] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0024] Figure 1 An architecture diagram of a processing system for an autonomous driving task provided in Embodiment 1 of the present invention;

[0025] Figure 2 An architecture diagram of another system for processing an autonomous driving task provided in Embodiment 1 of the present invention;

[0026] Figure 3 A flowchart of a method for processing an autonomous driving task provided in Embodiment 2 of the present invention;

[0027] Figure 4 A flowchart of another method for processing an autonomous driving task provided in Embodiment 2 of the present invention;

[0028] Figure 5 This is a schematic diagram of the structure of an electronic device provided in Embodiment 3 of the present invention. DETAILED DESCRIPTION

[0029] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0030] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0031] In order to further clarify the technical effects of the present invention, before introducing specific embodiments, the problems existing in the prior art are further explained as follows:

[0032] 1. Complexity of task types in data closed loop

[0033] In the closed loop of autonomous driving data, there are many types of tasks that need to be processed, including: sensor data collection, conversion and storage; data labeling and cleaning; model training and evaluation; simulation testing and verification; decision planning and control. These tasks are obviously different, and the requirements for computing resources (such as CPU, GPU, memory, etc.) are also different, which brings great challenges to task scheduling.

[0034] 2. Shortage of computing resources and low utilization

[0035] Due to the complexity of autonomous driving algorithms, the demand for computing resources such as GPUs is increasing. However, these expensive resources are often in limited supply, and how to reasonably plan and efficiently use them has become a prominent issue. The current task scheduling method may lead to problems such as low resource utilization and task delays.

[0036] 3. Complexity of task processing

[0037] There are complex dependencies between autonomous driving data processing tasks, forming a directed acyclic graph (DAG) structure. At the same time, different tasks have different priorities, and high-priority tasks need to be able to preempt resources. This complexity brings great challenges to task scheduling.

[0038] 4. Model and image loading start slowly

[0039] In autonomous driving data processing, it is usually necessary to load large AI models and Docker images. These models and image files are often large in size, and the loading and startup time may take several minutes, which seriously affects the efficiency of task execution.

[0040] 5. The shortcomings of traditional task management

[0041] Currently, the management of autonomous driving data processing tasks mainly relies on cluster management systems such as Kubernetes. However, this approach has a flaw, that is, the task status information is completely dependent on the stability of the cluster. Once the cluster fails, the task status may be wrong, resulting in scheduling failure.

[0042] Embodiment 1

[0043] Figure 1 This is an architecture diagram of a processing system for an autonomous driving task provided in Embodiment 1 of the present invention. The system comprises: a request receiving module, a control module, a scheduling module and a storage module, wherein the control module, the scheduling module and the storage module are respectively connected to the request receiving module.

[0044] It should be noted that the embodiment of the present invention can build a processing system for autonomous driving tasks based on the cloud native Kubernetes architecture. Among them, the request receiving module can be the Apiserver (API server) of the Kubernetes cluster, the control module can be the Controller-Manager (controller manager) of the Kubernetes cluster, the scheduling module can be the Scheduler (scheduler) of the Kubernetes cluster, and the storage module can be the Etcd (database service) of the Kubernetes cluster.

[0045] Among them, Apiserver is the interactive entrance of the entire service. All external services and user-executed commands are entered through Apiserver; Api server is mainly responsible for interacting with the database. Controller-Manager mainly controls resource instantiation and related processing logic, including task queues, task pools, task instances, etc. Scheduler contains a variety of scheduling strategies and is responsible for scheduling task instances to the most suitable nodes. Etcd is mainly used to store information such as task templates, task queues, task pools, task instances, task status, etc., and is also used to select the master of the scheduler and control manager. Figure 2 An architectural diagram of another system for processing autonomous driving tasks provided in Embodiment 1 of the present invention.

[0046] The request receiving module is used to receive a task processing request and send the task processing request to the control module. The request receiving module can provide a module for the internal or external interaction interface of the processing system of the autonomous driving task system. The external request can be transmitted to the processing system of the autonomous driving task through the request receiving module. The control module and the scheduling module can also interact based on the request receiving module.

[0047] In an optional embodiment, when the request receiving module receives a task processing request sent by a third-party application, the request receiving module may send the task processing request to the control module so that the task processing request is processed by the control module.

[0048] The third-party application can be any application that generates a task processing request in the vehicle's autonomous driving. The autonomous driving task processing system of this embodiment processes the pending tasks corresponding to the task requests, obtains the corresponding processing results, and returns the processing results to the third-party application, providing a basis for the autonomous driving control decision of the third-party application.

[0049] The control module is used to receive the task processing request, generate a task to be processed according to the task processing request and a task template, and store the task to be processed in a task queue.

[0050] Specifically, the control module can parse the received task processing request and extract the necessary information required for the task. Then, based on the parsed task information and the preset task template, the control module will generate specific tasks to be processed. The tasks to be processed will be stored in the task queue by the control module, waiting for subsequent scheduling and execution. The task queue is a buffer in the system for storing tasks to be processed.

[0051] The scheduling module is used to determine the target task to be processed in the task queue, and schedule the target task to be processed to a task instance in a task pool, or to a target node in the processing system of the autonomous driving task, so that the task instance or the target node processes the target task to be processed.

[0052] The target pending task may be a task that needs to be processed currently selected by the scheduling module from the task queue. The scheduling module schedules the target pending task to a task instance in the task pool or to a target node for processing the target pending task, and the target node may be a service node of a Kubernetes cluster.

[0053] The storage module is used to store the task template, the task queue, the task pool, the task instance and the status information of the task to be processed.

[0054] In the embodiment of the present invention, optionally, the request receiving module includes:

[0055] The authentication submodule is used to authenticate the request sender when receiving a communication request from the request sender.

[0056] The authorization and control submodule is used to set the access resources and access rights of the sender, so as to receive the task processing request sent by the sender after the access rights and the access resources are set.

[0057] Specifically, the request sender can be an application or component that sends a task processing request. The authentication submodule and the authorization and control submodule can authenticate the request sender before establishing communication with the request sender, and further set the access resources and access rights of the request sender if the authentication is passed. That is, based on the two submodules, mechanisms such as authentication, authorization, access control, API registration and discovery can be provided, and REST operation requests from internal components and external users can be processed, task status can be maintained, and task-related information status can be stored in etcd.

[0058] In the embodiment of the present invention, the control module includes: a task type determination submodule, a task template association submodule and a task addition submodule, wherein:

[0059] The task type determination submodule is used to determine the task type corresponding to the task processing request, wherein the task type includes at least one of a simulation task, an evaluation task and a data processing task;

[0060] A task template association submodule, used to determine the association template corresponding to the task processing request according to the task type, and process the task processing request based on the association template to obtain the task to be processed;

[0061] The task adding submodule is used to add the to-be-processed task to the task queue corresponding to the task type.

[0062] Specifically, in order to ensure that the task is processed correctly and efficiently, the task type judgment submodule is responsible for receiving the task processing request and judging the task type corresponding to the request. The task type can be a simulation task, an evaluation task, a data processing task, or any combination of these three types. For example, by parsing the task type information in the request, it is judged whether the task type belongs to the predefined task type, and then the judgment result is passed to the task template association submodule.

[0063] The task template association submodule receives the task type passed by the task type judgment submodule, selects the corresponding template from the predefined task template library according to the task type, and uses the selected template to process the task processing request, including but not limited to parameter verification, data formatting, task configuration, etc. After processing, it generates pending tasks and passes them to the task addition submodule. The task addition submodule is responsible for adding pending tasks to the task queue corresponding to the task type so that subsequent nodes or task instances can process them.

[0064] In the embodiment of the present invention, the scheduling module includes: a target to-be-processed task determination module, which is used to filter out the target to-be-processed tasks from the task queue through a screening and filtering algorithm.

[0065] Specifically, the target pending task determination module can accurately select the target pending tasks that need to be processed first or meet specific conditions from the task queue through a screening and filtering algorithm.

[0066] The target pending task determination module is responsible for monitoring the task queue and identifying the target pending tasks from the queue according to the preset screening and filtering algorithm. The screening conditions can be based on the priority, urgency, resource requirements, dependencies, etc. of the pending tasks. The target pending task determination module can ensure that the target pending tasks can be efficiently determined when a large number of tasks arrive at the same time. The screening and filtering algorithm can also be adjusted or new screening conditions can be added to adapt to different autonomous driving task processing requirements.

[0067] For example, tasks can be sorted according to their priorities (such as high, medium, and low) and the highest priority task can be selected. Alternatively, tasks can be filtered based on their urgency. Alternatively, tasks can be selected based on the current system resources (such as CPU, memory, storage, etc.) and the resources required for the task, and the task that best suits the current resource situation can be selected. Alternatively, the dependencies between tasks can be considered to ensure that the tasks are processed in the correct order.

[0068] In an embodiment of the present invention, the scheduling module includes: a first scheduling submodule, which is used to schedule the target to-be-processed task to a target task instance according to the task pool state corresponding to the target to-be-processed task when the target to-be-processed task is a repeatable task;

[0069] The second scheduling submodule is used to schedule the target to-be-processed task to the target node through a target scheduling algorithm when the target to-be-processed task is a non-repeatedly executed task.

[0070] It should also be noted that for most task types, such as simulation and evaluation, the versions of the autonomous driving models for the same batch of tasks are basically the same, but the sensor data, maps and other information used are different. Each time a task is executed, the relevant image will be pulled up, even if the container used is the same. Because the model is too large, it takes a lot of time to pull and start the container. Therefore, before the target pending task is processed, it can be determined whether the target pending task is a repetitive task.

[0071] If the task is repeatable, the first scheduling submodule can further check the status of the task pool related to the task, such as the load, idle status, resource utilization rate, etc. of the task instance. Then, based on the status information of the task pool, the first scheduling submodule selects one or more suitable task instances as target task instances, and then schedules the target to-be-processed task to the selected target task instance for processing.

[0072] It can be understood that since the task is repeatable, it has been instantiated and the task instance is stored in the task pool when this type of task is first scheduled. Therefore, when encountering a repeated task, the target pending task can be directly scheduled to the target task instance corresponding to the pending task.

[0073] If the target pending task is a non-repeated task, the second scheduling submodule will schedule the task to the best target node for processing through the target scheduling algorithm. The second scheduling submodule will evaluate the resource status of each node in the system, such as CPU, memory, and storage. Further based on the evaluation results, the second scheduling submodule applies a target scheduling algorithm, such as a load balancing algorithm, a priority scheduling algorithm, etc., to determine the best target node and schedule the target pending task to the selected target node for processing.

[0074] In an embodiment of the present invention, preferably, the first scheduling submodule includes an idle instance determination unit, which is used to determine whether there is an idle task instance corresponding to the target task to be processed in the task pool. If so, the idle task instance is used as the target task instance, and the target task to be processed is scheduled to the target task instance.

[0075] Among them, the idle instance determination unit is responsible for checking the status of task instances in the task pool, mainly looking for idle task instances that match the target pending task type, resource requirements, etc. Once such an idle instance is found, it will perform a scheduling operation and assign the target pending task to the instance.

[0076] Specifically, the idle instance determination unit receives the target pending task from the scheduling system. The idle instance determination unit can traverse all task instances in the task pool and check their current status, including whether they are idle, resource occupancy, supported task types, etc. Then, based on the type of the target pending task, resource requirements, etc., the idle instance determination unit will search for a matching idle task instance in the task pool. If a matching idle task instance is found, the unit will designate it as the target task instance. Finally, the target pending task is scheduled to the target task instance, that is, the instance is notified to start processing the task.

[0077] On the basis of the above technical solution, the first scheduling submodule also includes a task pool capacity determination unit, which is used to determine the capacity of the task pool when there is no idle task instance corresponding to the target task to be processed in the task pool; if the capacity of the task pool does not reach the capacity threshold, the target task to be processed is forwarded to the second scheduling submodule; the second scheduling submodule is also used to schedule the target task to be processed to the target node through a target scheduling algorithm.

[0078] It can be understood that when there is no idle task instance corresponding to the target pending task in the task pool, the task pool capacity will first check the current capacity of the task pool. If the capacity of the task pool does not reach the preset capacity threshold, it means that the repeatable target pending task may be executed for the first time, and the target pending task can be forwarded to the second scheduling submodule for further processing. After receiving the task, the second scheduling submodule will schedule the target pending task to a suitable target node.

[0079] In a preferred embodiment, the storage module includes a task instance storage sub-module, which is used to store the task instance of the target task to be processed in the task pool after the target task to be processed is instantiated when the target task to be processed is a repeatable task and the target task to be processed is scheduled for the first time.

[0080] When the target pending task is a repeatable task and it is scheduled for the first time, the task instance storage submodule will store its task instance in the task pool after the target pending task is instantiated, so that the next time the repeatable task is executed, it can be directly scheduled to the task instance in the task pool.

[0081] The present invention proposes solutions to a series of problems existing in the prior art in the processing of autonomous driving tasks. The specific problems and corresponding solutions are as follows: 1. The prior art lacks a targeted scheduling method: The prior art fails to fully consider the special needs of autonomous driving data processing, such as the real-time nature, priority and resource requirements of tasks, resulting in low system resource utilization and delayed processing of key tasks. The method proposed in the present invention dynamically determines the order of task execution by comprehensively considering factors such as task type, priority, resource requirements and dependencies, effectively improving system resource utilization and the processing speed of key tasks. 2. Unable to dynamically adapt to complex and changeable working environments: When facing the complex and changeable working environment of autonomous driving, the prior art is difficult to adjust the order of task execution in time, affecting the real-time response capability of the system. The present invention dynamically adjusts the order of task execution through an intelligent algorithm to improve the robustness and real-time response capability of the system. 3. Slow loading and starting of models and images: In the prior art, the loading and starting time of large AI models and Docker images is long, affecting the efficiency of task execution. The present invention accelerates the loading and starting of models and images and reduces the task startup overhead by optimizing the task scheduling strategy. 4. Insufficients of traditional task management: The existing technology is too dependent on the cluster status. Once the cluster fails, the task status may be wrong, resulting in scheduling failure. The present invention reduces the dependence on the cluster status and improves the reliability and fault tolerance of task scheduling.

[0082] The technical effects directly brought about by the method of the embodiment of the present invention include: 1. Improving system resource utilization: Through intelligent scheduling algorithms, the utilization of system resources is effectively improved and operating costs are reduced. 2. Shortening task processing delays: Optimizing the order of task execution, ensuring rapid processing of key tasks, and improving the real-time and responsiveness of the autonomous driving system. 3. Improving system robustness: Being able to dynamically adapt to complex and changing working environments and improve system stability and security. 4. Accelerating model and image loading: By optimizing scheduling strategies, the loading and startup time of models and images is significantly reduced, and task execution efficiency is improved. 5. Improving the reliability of task scheduling: Reducing dependence on cluster status and improving the reliability and fault tolerance of task scheduling. 6. Good scalability: Being able to cope with the growth of future data volume and computing power requirements, and ensuring the sustainable development of autonomous driving technology.

[0083] Embodiment 2

[0084] Figure 3 This is a flowchart of a method for processing an autonomous driving task provided in the second embodiment of the present invention. This embodiment is applicable to the case of processing an autonomous driving task. The method can be executed by a processing system for an autonomous driving task. The system can be implemented in the form of hardware and / or software. The system device can be configured in a vehicle or in the cloud. As shown in the figure, the method includes:

[0085] S210: Receive a task processing request through the request receiving module, and send the task processing request to the control module.

[0086] Specifically, the request receiving module can provide a module for the internal or external interaction interface of the processing system of the autonomous driving task system. External requests can be transmitted to the processing system of the autonomous driving task through the request receiving module, and the control module and the scheduling module can also interact based on the request receiving module.

[0087] S220, receiving the task processing request through the control module, generating a task to be processed according to the task processing request and a task template, and storing the task to be processed in a task queue.

[0088] Specifically, the control module can parse the received task processing request and extract the necessary information required for the task. Then, based on the parsed task information and the preset task template, the control module will generate specific tasks to be processed. The tasks to be processed will be stored in the task queue by the control module, waiting for subsequent scheduling and execution. The task queue is a buffer in the system for storing tasks to be processed.

[0089] S230. Determine the target task to be processed in the task queue through the scheduling module, and schedule the target task to be processed to a task instance in the task pool, or to a target node in the processing system of the autonomous driving task, so that the task instance or the target node processes the target task to be processed.

[0090] Specifically, the target pending task can be a task that needs to be processed currently selected by the scheduling module from the task queue. The scheduling module schedules the target pending task to a task instance in the task pool or to a target node to process the target pending task, and the target node can be a service node of a Kubernetes cluster.

[0091] In a preferred embodiment, determining the target task to be processed in the task queue by the scheduling module, and scheduling the target task to be processed to a task instance in a task pool, or to a target node in the processing system of the autonomous driving task, includes:

[0092] When the target pending task is a repeatable task, the target pending task is scheduled to the target task instance through the first scheduling submodule according to the task pool status corresponding to the target pending task; when the target pending task is a non-repeated task, the target pending task is scheduled to the target node based on the second scheduling submodule and the target scheduling algorithm.

[0093] Specifically, for most task types, such as simulation and evaluation, the versions of the autonomous driving models for the same batch of tasks are basically the same, but the sensor data, maps and other information used are different. Each time a task is executed, the relevant image will be pulled up, even if the container used is the same. Because the model is too large, it takes a lot of time to pull and start the container. Therefore, before the target pending task is processed, it can be determined whether the target pending task is a repetitive task.

[0094] If the task is repeatable, the first scheduling submodule can further check the status of the task pool related to the task, such as the load, idle status, resource utilization rate, etc. of the task instance. Then, based on the status information of the task pool, the first scheduling submodule selects one or more suitable task instances as target task instances, and then schedules the target to-be-processed task to the selected target task instance for processing.

[0095] It is understandable that, since the task is repeatable, it has been instantiated and the task instance is stored in the task pool when the task of this type is first scheduled. Therefore, when a repeated task is encountered, the target task to be processed can be directly scheduled to the target task instance corresponding to the task to be processed. S240, storing the task template, the task queue, the task pool, the task instance and the status information of the task to be processed by the storage module.

[0096] In a preferred embodiment, the Figure 4 The process of processing pending tasks, Figure 4 A flowchart of another method for processing an autonomous driving task provided in Embodiment 2 of the present invention.

[0097] In the embodiment of the present invention, queues can be divided according to task types, and one queue can contain tasks of the same type to be scheduled. Different queues can be configured with relative weights and restrictions of computing resources to ensure reasonable allocation of resources and improve resource utilization. When the service is started, a queue will be started for each type of task according to the configuration, waiting for tasks to enter.

[0098] The task pool of the present invention manages "resident" task instances. For most task types, such as simulation, evaluation, etc., the versions of the autonomous driving models for the same batch of tasks are basically the same, except that the sensor data, maps and other information used are different. In traditional k8s jobs, each task will be pulled up with a related image, even if the container used is the same. Because the model is too large, it takes a lot of time to pull and start the container, but the corresponding hardware resources have been allocated since the task was instantiated. This approach greatly reduces the efficiency of task execution and wastes the utilization of hardware resources. Therefore, for this type of resident task instance, a task pool is set. When such tasks are executed, they only need to load the required data and can be executed repeatedly, which greatly reduces the execution time of the task and improves the utilization of resources. The tasks in the task pool will be created with a delay, and the task pool will be instantiated during the first scheduling. When there is no such type of task in the queue, the instance in the task pool will be released, and the relevant hardware resources will be released in time.

[0099] The tasks to be processed in the present invention refer to specific instantiated tasks, which can be divided into simulation tasks, evaluation tasks, data cleaning tasks, data conversion tasks, etc. according to their purpose. According to their repeatability, they can be divided into one-time tasks and permanent tasks.

[0100] The present invention proposes solutions to a series of problems existing in the prior art in the processing of autonomous driving tasks. The specific problems and corresponding solutions are as follows: 1. The prior art lacks a targeted scheduling method: The prior art fails to fully consider the special needs of autonomous driving data processing, such as the real-time nature, priority and resource requirements of tasks, resulting in low system resource utilization and delayed processing of key tasks. The method proposed in the present invention dynamically determines the order of task execution by comprehensively considering factors such as task type, priority, resource requirements and dependencies, effectively improving system resource utilization and the processing speed of key tasks. 2. Unable to dynamically adapt to complex and changeable working environments: When facing the complex and changeable working environment of autonomous driving, the prior art is difficult to adjust the order of task execution in time, affecting the real-time response capability of the system. The present invention dynamically adjusts the order of task execution through an intelligent algorithm to improve the robustness and real-time response capability of the system. 3. Slow loading and starting of models and images: In the prior art, the loading and starting time of large AI models and Docker images is long, affecting the efficiency of task execution. The present invention accelerates the loading and starting of models and images and reduces the task startup overhead by optimizing the task scheduling strategy. 4. Insufficients of traditional task management: The existing technology is too dependent on the cluster status. Once the cluster fails, the task status may be wrong, resulting in scheduling failure. The present invention reduces the dependence on the cluster status and improves the reliability and fault tolerance of task scheduling.

[0101] The technical effects directly brought about by the method of the embodiment of the present invention include: 1. Improving system resource utilization: Through intelligent scheduling algorithms, the utilization of system resources is effectively improved and operating costs are reduced. 2. Shortening task processing delays: Optimizing the order of task execution, ensuring rapid processing of key tasks, and improving the real-time and responsiveness of the autonomous driving system. 3. Improving system robustness: Being able to dynamically adapt to complex and changing working environments and improve system stability and security. 4. Accelerating model and image loading: By optimizing scheduling strategies, the loading and startup time of models and images is significantly reduced, and task execution efficiency is improved. 5. Improving the reliability of task scheduling: Reducing dependence on cluster status and improving the reliability and fault tolerance of task scheduling. 6. Good scalability: Being able to cope with the growth of future data volume and computing power requirements, and ensuring the sustainable development of autonomous driving technology.

[0102] Embodiment 3

[0103] Figure 5 A schematic diagram of the structure of an electronic device provided for Embodiment 3 of the present invention. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0104] like Figure 5 As shown, the electronic device 10 includes at least one processor 11, and a memory connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 to the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0105] A number of components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0106] The processor 11 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 executes the various methods and processes described above, such as a processing method for an autonomous driving task.

[0107] In some embodiments, a method for processing an autonomous driving task may be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the method for processing an autonomous driving task described above may be performed. Alternatively, in other embodiments, the processor 11 may be configured to execute a method for processing an autonomous driving task in any other appropriate manner (e.g., by means of firmware).

[0108] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0109] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0110] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in combination with an instruction execution system, device or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0111] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0112] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0113] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0114] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0115] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A processing system for an autonomous driving task, characterized in that: include: A request receiving module, a control module, a scheduling module and a storage module, wherein the control module, the scheduling module and the storage module are respectively connected to the request receiving module, wherein: The request receiving module is used to receive a task processing request and send the task processing request to the control module; The control module is used to receive the task processing request, generate a task to be processed according to the task processing request and a task template, and store the task to be processed in a task queue; The scheduling module is used to determine the target task to be processed in the task queue, and schedule the target task to be processed to a task instance in the task pool, or to a target node in the processing system of the autonomous driving task, so that the task instance or the target node processes the target task to be processed; The storage module is used to store the task template, the task queue, the task pool, the task instance and the status information of the task to be processed.

2. The system according to claim 1, characterized in that The request receiving module includes: an authentication submodule and an authorization and control submodule; wherein, The authentication submodule is used to authenticate the request sender when receiving a communication request from the request sender; The authorization and control submodule is used to set the access resources and access rights of the sender, so as to receive the task processing request sent by the sender after the access rights and the access resources are set.

3. The system according to claim 1, characterized in that The control module includes: a task type determination submodule, a task template association submodule and a task addition submodule, wherein: The task type determination submodule is used to determine the task type corresponding to the task processing request, wherein the task type includes at least one of a simulation task, an evaluation task and a data processing task; A task template association submodule, used to determine the association template corresponding to the task processing request according to the task type, and process the task processing request based on the association template to obtain the task to be processed; The task adding submodule is used to add the to-be-processed task to the task queue corresponding to the task type.

4. The system according to claim 1, characterized in that The scheduling module includes: a target task to be processed determination module, wherein: The target to-be-processed task determination module is used to filter out the target to-be-processed tasks from the task queue through a screening and filtering algorithm.

5. The system according to claim 1, characterized in that The scheduling module includes: a first scheduling submodule and a second scheduling submodule, wherein: The first scheduling submodule is used for scheduling the target to-be-processed task to a target task instance according to a task pool state corresponding to the target to-be-processed task when the target to-be-processed task is a repeatable task; The second scheduling submodule is used to schedule the target to-be-processed task to the target node through a target scheduling algorithm when the target to-be-processed task is a non-repeatedly executed task.

6. The system according to claim 5, characterized in that The first scheduling submodule includes an idle instance determination unit, wherein: The idle instance determination unit is used to determine whether there is an idle task instance corresponding to the target task to be processed in the task pool. If so, the idle task instance is used as the target task instance and the target task to be processed is scheduled to the target task instance.

7. The system according to claim 6, characterized in that The first scheduling submodule further includes a task pool capacity determination unit, wherein: The task pool capacity determination unit is used to determine the capacity of the task pool when there is no idle task instance corresponding to the target to-be-processed task in the task pool, and forward the target to-be-processed task to the second scheduling submodule if the capacity of the task pool does not reach the capacity threshold; The second scheduling submodule is also used to schedule the target to-be-processed task to the target node through a target scheduling algorithm.

8. The system according to claim 7, characterized in that The storage module includes a task instance storage submodule, wherein: The task instance storage submodule is used to store the task instance of the target task to be processed in the task pool after the instantiation of the target task to be processed is completed when the target task to be processed is a repeatable task and the target task to be processed is scheduled for the first time.

9. A method for processing an autonomous driving task, characterized in that: A processing system applied to an autonomous driving task, the system comprising: a request receiving module, a control module, a scheduling module and a storage module, the control module, the scheduling module and the storage module are respectively connected to the request receiving module, and the method comprises: receiving a task processing request through the request receiving module, and sending the task processing request to the control module; The control module receives the task processing request, generates a task to be processed according to the task processing request and a task template, and stores the task to be processed in a task queue; Determine the target task to be processed in the task queue by the scheduling module, and schedule the target task to be processed to a task instance in a task pool, or to a target node in the processing system of the autonomous driving task, so that the task instance or the target node processes the target task to be processed; The storage module stores the task template, the task queue, the task pool, the task instance and the status information of the pending tasks.

10. The method according to claim 9, characterized in that The step of determining the target task to be processed in the task queue by the scheduling module, and scheduling the target task to be processed to a task instance in a task pool, or to a target node in the processing system of the autonomous driving task, includes: When the target to-be-processed task is a repeatable task, the target to-be-processed task is scheduled to the target task instance through the first scheduling submodule according to the task pool state corresponding to the target to-be-processed task; When the target to-be-processed task is a non-repeatedly executed task, the target to-be-processed task is scheduled to the target node based on the second scheduling submodule and the target scheduling algorithm.