An algorithm optimization method for a system chip for automatic driving of an automobile

Abstract

The application discloses a kind of power optimization methods of system chip for automobile automatic driving, and the application relates to power optimization technical field, the optimization method includes: the scene of vehicle under normal driving and the task generated due to environment are transmitted to storage module, and according to task characteristic and the information entered are compared;When vehicle records environment and the characteristics generated by the environment and task in storage module are same when vehicle is normally driven, the task and result in storage module are exported, the advantages of the present application are: by generating instruction module, the environment characteristics in panoramic image are extracted, then the extracted features generate corresponding task instruction, comparison module is matched with the data in storage module by task characteristics and scene characteristics in vehicle driving, and whether the driving system power chip needs to be started is determined, the system chip usage power in automobile automatic driving can be reduced, and the power consumption of system chip is reduced.

Description

Technical Field

[0001] This invention relates to the field of computing power optimization technology, specifically a method for optimizing the computing power of a system chip used in autonomous driving. Background Technology

[0002] Currently, autonomous driving requires the connection of many different sensors. Autonomous vehicles need to process a large amount of data and their algorithms are very complex. Therefore, autonomous vehicles require strong computing power.

[0003] However, during long-term computation, the chips in the autonomous driving system consume excessive computing power because the system chip schedules and calculates multiple sensors in the autonomous vehicle. To address this, we propose a method for optimizing the computing power of the system chip in the autonomous driving system. Summary of the Invention

[0004] The purpose of this invention is to provide a method for optimizing the computing power of a system chip for autonomous driving in automobiles.

[0005] To address the problems mentioned in the background art, the present invention provides the following technical solution: a method for optimizing the computing power of a system chip for autonomous driving in automobiles, the optimization method comprising:

[0006] Step 1: Transmit the scene of the vehicle under normal driving conditions and the tasks generated by the environment to the storage module, and compare them with the entered information based on the characteristics of the tasks.

[0007] Step 2: When the vehicle is driving normally, if the characteristics of the environment and tasks recorded by the vehicle are the same as those in the storage module, then the tasks and results in the storage module are exported.

[0008] Step 3: In autonomous vehicles, multiple processing nodes are set up to collect computing power from environmental data during driving. When the driving scene in the storage module does not match the driving scene features recorded by the vehicle, the driving scene and task are imported into the processing node.

[0009] Step 4: Determine the processing time of the processing nodes for the autonomous driving task based on the preset time of computing power, and reserve 40% of the computing power of the multiple processing nodes.

[0010] Step 5: For tasks that exceed the processing time of the processing node, break them down into multiple subtasks and execute them, and use the reserved computing power of the processing node to perform calculations on the subtasks.

[0011] Step 6: Construct a task processing framework and record tasks that exceed the allocated time.

[0012] Step 7: Establish an AI learning module. When the scene information in the storage module does not match the scene information while the vehicle is driving, the module will retrieve the data from the storage module to match the scene information and classify the driving scene. The classified information will then be stored in the storage module.

[0013] Step 8: Establish a communication connection between the storage module and the cloud, and transmit information between the storage module and the cloud. Then, retrieve and download the corresponding categorized information from the cloud based on the driving scenario.

[0014] As a further aspect of the present invention: in step one, a comparison module is set up, and the comparison module is used to distinguish when the vehicle is driving, based on the road condition information collected in step two, extract the characteristic information in the road scene, and match it with the driving scene features recorded in the storage module.

[0015] As a further aspect of the present invention: in step two, during vehicle operation, road information is recorded based on the panoramic image carried in the vehicle, thereby obtaining environmental feature information.

[0016] As a further aspect of the present invention: in step two, a judgment module is set up, and the judgment module is used to judge whether the vehicle driving scene features entered in the input system match. When they match, the current computing power is terminated.

[0017] As a further aspect of the present invention: step two also includes a command generation module, which generates a task command module based on environmental feature information. If there is a mismatch, the process proceeds to step four.

[0018] As a further aspect of the present invention: in step three, processing nodes are set according to the panoramic image of the vehicle, temporary computing power is allocated from the processing nodes, and tasks that exceed the preset time in step four are imported into the temporary computing power, and budget time is set again for the tasks in the temporary computing power.

[0019] As a further aspect of the present invention: in step five, when the calculation time of a task in the temporary computing power exceeds a preset time, the splitting module is used to split the remaining unfinished task into subtasks and transmit the subtasks to multiple processing nodes.

[0020] As a further aspect of the present invention: Step seven involves using an AI learning module to add scene feature information of the vehicle while driving and performing feature classification to improve the tasks generated by autonomous driving under different factors in the driving scenario.

[0021] As a further aspect of the present invention: step eight involves communication between the storage module and the cloud, and by classifying different driving scenarios and downloading the corresponding driving scenario information from the cloud, the comparability is improved.

[0022] Compared with the prior art, the beneficial effects of the present invention by adopting the above technical solution are as follows:

[0023] 1. This invention extracts environmental features from panoramic images through a command generation module, and then generates corresponding task commands from the extracted features. The comparison module matches the task features and scene features in vehicle driving with the data in the storage module to determine whether the computing power chip of the driving system needs to be activated. This can reduce the computing power used by the system chip in autonomous driving and reduce the computing power consumption of the system chip.

[0024] 2. This invention improves the speed of processing tasks that are difficult to process quickly by using temporary computing power to calculate tasks that are inconvenient to process in a timely manner. When the task calculated by the temporary computing power exceeds the budget time, the splitting module splits the uncalculated task and uses temporary computing power to calculate multiple sub-tasks at the same time, thereby improving the effect of fast processing of large tasks.

[0025] 3. This invention records feature information that exceeds the budget time through an AI learning module, classifies the data, and then transmits the driving scenario data to a storage module. The storage module and the cloud can retrieve data from each other, making it easy to download data from the cloud to the storage module according to the classification. This improves the different features of the vehicle scenario and reduces the computing power of the system chip when the vehicle is driving autonomously. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of the computing power optimization method in an embodiment of the present invention;

[0027] Figure 2 The following is an optimized system flowchart in an embodiment of the present invention. Detailed Implementation

[0028] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings. It should be noted that these descriptions are for the purpose of aiding understanding the present invention, but do not constitute a limitation thereof. Furthermore, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0029] Example 1, please refer to the appendix. Figure 1 - Appendix Figure 2 This invention provides a technical solution: a method for optimizing the computing power of a system chip used in autonomous driving, the optimization method comprising:

[0030] Step 1: Transmit the scene of the vehicle under normal driving conditions and the tasks generated by the environment to the storage module, and compare them with the entered information based on the characteristics of the tasks.

[0031] Step 2: When the vehicle is driving normally, if the characteristics of the environment and tasks recorded by the vehicle are the same as those in the storage module, then the tasks and results in the storage module are exported.

[0032] Step 3: In autonomous vehicles, multiple processing nodes are set up to collect computing power from environmental data during driving. When the driving scene in the storage module does not match the driving scene features recorded by the vehicle, the driving scene and task are imported into the processing node.

[0033] Step 4: Determine the processing time of the processing nodes for the autonomous driving task based on the preset time of computing power, and reserve 40% of the computing power of the multiple processing nodes.

[0034] Step 5: For tasks that exceed the processing time of the processing node, break them down into multiple subtasks and execute them, and use the reserved computing power of the processing node to perform calculations on the subtasks.

[0035] Step 6: Construct a task processing framework and record tasks that exceed the allocated time.

[0036] Step 7: Establish an AI learning module. When the scene information in the storage module does not match the scene information while the vehicle is driving, the module will retrieve the data from the storage module to match the scene information and classify the driving scene. The classified information will then be stored in the storage module.

[0037] Step 8: Establish a communication connection between the storage module and the cloud, and transmit information between the storage module and the cloud. Then, retrieve and download the corresponding categorized information from the cloud based on the driving scenario.

[0038] Please see the appendix Figure 1 In step one, a comparison module is set up. This module is used to distinguish the road conditions collected in step two when the vehicle is driving. It extracts the characteristic information in the road scene and matches it with the driving scene characteristics recorded in the storage module. In step two, the vehicle records road information based on the panoramic image carried in the vehicle to obtain environmental characteristic information. In step two, a judgment module is set up to determine whether the driving scene characteristics recorded in the system match. If they match, the current computing power ends. Step two also sets up a command generation module to generate task commands based on environmental characteristic information. If they do not match, the process proceeds to step four.

[0039] In this embodiment, the formula is used:

[0040]

[0041] Where S represents the task instruction generated by the vehicle while it is in motion, A represents the various features in the task instruction, and (1) and (0) represent whether there are corresponding instruction features in the storage module.

[0042] During use, the command generation module extracts environmental features from the panoramic image, and then generates corresponding task commands from the extracted features. The comparison module matches the task features and scene features in the vehicle driving with the data in the storage module to determine whether the computing power chip for the driving system needs to be activated. This can reduce the computing power used by the system chip in the autonomous driving of the car and reduce the computing power consumption of the system chip.

[0043] Example 2, please refer to the appendix. Figure 1 - Appendix Figure 2 This invention provides a technical solution: a computing power optimization method for autonomous driving of a vehicle, the optimization method comprising:

[0044] Step 1: Transmit the scene of the vehicle under normal driving conditions and the tasks generated by the environment to the storage module, and compare them with the entered information based on the characteristics of the tasks.

[0045] Step 2: When the vehicle is driving normally, if the characteristics of the environment and tasks recorded by the vehicle are the same as those in the storage module, then the tasks and results in the storage module are exported.

[0046] Step 3: In autonomous vehicles, multiple processing nodes are set up to collect computing power from environmental data during driving. When the driving scene in the storage module does not match the driving scene features recorded by the vehicle, the driving scene and task are imported into the processing node.

[0047] Step 4: Determine the processing time of the processing nodes for the autonomous driving task based on the preset time of computing power, and reserve 40% of the computing power of the multiple processing nodes.

[0048] Step 5: For tasks that exceed the processing time of the processing node, break them down into multiple subtasks and execute them, and use the reserved computing power of the processing node to perform calculations on the subtasks.

[0049] Step 6: Construct a task processing framework and record tasks that exceed the allocated time.

[0050] Step 7: Establish an AI learning module. When the scene information in the storage module does not match the scene information while the vehicle is driving, the module will retrieve the data from the storage module to match the scene information and classify the driving scene. The classified information will then be stored in the storage module.

[0051] Step 8: Establish a communication connection between the storage module and the cloud, and transmit information between the storage module and the cloud. Then, retrieve and download the corresponding categorized information from the cloud based on the driving scenario.

[0052] Please see the appendix Figure 1 Step two also includes a command generation module, which generates task commands based on environmental feature information. If there is a mismatch, the process proceeds to step four. In step three, processing nodes are set based on the vehicle panoramic image, and temporary computing power is allocated from the processing nodes. Tasks that exceed the preset time in step four are imported into the temporary computing power, and a budget time is set again for the tasks in the temporary computing power. In step five, a splitting module is used to break down the remaining unfinished tasks into sub-tasks when the calculation time of a task in the temporary computing power exceeds the preset time. The sub-tasks are then transmitted to multiple processing nodes.

[0053] In this embodiment, the formula is used:

[0054]

[0055] Where Q represents the subtasks split by the splitting module, W represents the tasks generated by the generation instruction module, E represents the portion of the task that has been calculated, and N represents the number of processing nodes.

[0056] When in use, temporary computing power is used to calculate tasks that are not convenient to process in a timely manner, thereby improving the speed of processing more difficult tasks. When the task calculated by the temporary computing power exceeds the budget time, the splitting module splits the uncalculated task and uses temporary computing power to calculate multiple sub-tasks at the same time, thereby improving the effect of fast processing of large tasks.

[0057] Example 3, please refer to the appendix. Figure 1 - Appendix Figure 2 This invention provides a technical solution: a method for optimizing the computing power of a system chip used in autonomous driving, the optimization method comprising:

[0058] Step 1: Transmit the scene of the vehicle under normal driving conditions and the tasks generated by the environment to the storage module, and compare them with the entered information based on the characteristics of the tasks.

[0059] Step 2: When the vehicle is driving normally, if the characteristics of the environment and tasks recorded by the vehicle are the same as those in the storage module, then the tasks and results in the storage module are exported.

[0060] Step 3: In autonomous vehicles, multiple processing nodes are set up to collect computing power from environmental data during driving. When the driving scene in the storage module does not match the driving scene features recorded by the vehicle, the driving scene and task are imported into the processing node.

[0061] Step 4: Determine the processing time of the processing nodes for the autonomous driving task based on the preset time of computing power, and reserve 40% of the computing power of the multiple processing nodes.

[0062] Step 5: For tasks that exceed the processing time of the processing node, break them down into multiple subtasks and execute them, and use the reserved computing power of the processing node to perform calculations on the subtasks.

[0063] Step 6: Construct a task processing framework and record tasks that exceed the allocated time.

[0064] Step 7: Establish an AI learning module. When the scene information in the storage module does not match the scene information while the vehicle is driving, the module will retrieve the data from the storage module to match the scene information and classify the driving scene. The classified information will then be stored in the storage module.

[0065] Step 8: Establish a communication connection between the storage module and the cloud, and transmit information between the storage module and the cloud. Then, retrieve and download the corresponding categorized information from the cloud based on the driving scenario.

[0066] Please see the appendix Figure 2 Step seven involves using the AI ​​learning module to add scene feature information of the vehicle while driving and to classify the features, thereby improving the tasks generated by autonomous driving under different factors in the driving scenario. Step eight involves exchanging information with the cloud through the storage module, classifying different driving scenarios, and downloading the corresponding classified driving scenario information from the cloud to improve the comparability.

[0067] During use, the AI ​​learning module records the feature information of exceeding the budget time, classifies the data, and then transmits the driving scenario data to the storage module. The storage module and the cloud can retrieve data from each other, making it easy to download data from the cloud to the storage module according to the classification, thereby improving the different features of the vehicle scenario and reducing the computing power of the system chip when the vehicle is driving autonomously.

[0068] Working principle:

[0069] First, the driving scene under normal driving conditions and the tasks generated by the environment are transmitted to the storage module. The task characteristics are compared with the recorded information. When the characteristics of the vehicle's recorded environment and the environment and tasks in the storage module are the same during normal driving, the tasks and results in the storage module are exported. Second, multiple processing nodes are set up to acquire the computing power generated by the autonomous vehicle's environmental data collection during driving. When the driving scene in the storage module does not match the driving scene characteristics recorded by the vehicle, the driving scene and tasks are imported into the processing nodes. The processing time for the autonomous driving tasks is determined by the preset time of the multiple processing nodes based on the computing power, and a percentage is reserved for each processing node. The system utilizes 40% of the computing power to process tasks that exceed the processing time of the processing nodes. This is achieved by breaking down the tasks into multiple sub-tasks and utilizing the reserved computing power of the processing nodes to perform calculations on these sub-tasks. Finally, by constructing a task processing framework and recording tasks exceeding the reserved time, an AI learning module is established. When the scene information stored in the storage module does not match the vehicle's driving conditions, the system retrieves data from the storage module to match the scene information and categorizes the driving scene. This categorized information is then stored in the storage module, and a communication connection is established between the storage module and the cloud for mutual information transmission. Finally, the system retrieves and downloads the corresponding categorized information from the cloud based on the driving scene. This completes the entire workflow.

[0070] Those skilled in the art will understand that the modules or steps of the present invention described above can be implemented using general-purpose computer devices. Optionally, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage device for execution by a computer device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. The present invention is not limited to any particular combination of hardware and software.

[0071] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention can have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0072] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A method for optimizing the computing power of a system chip for autonomous driving in automobiles, characterized in that: The optimization method includes: Step 1: Transmit the scene of the vehicle under normal driving conditions and the tasks generated by the environment to the storage module, and compare them with the entered information based on the characteristics of the tasks. Step 2: When the vehicle is driving normally, if the characteristics of the environment and tasks recorded by the vehicle are the same as those in the storage module, then the tasks and results in the storage module are exported. Step 3: In autonomous vehicles, multiple processing nodes are set up to collect computing power from environmental data during driving. When the driving scene in the storage module does not match the driving scene features recorded by the vehicle, the driving scene and task are imported into the processing node. Step 4: Determine the processing time of the processing nodes for the autonomous driving task based on the preset time of computing power, and reserve 40% of the computing power of the multiple processing nodes. Step 5: For tasks that exceed the processing time of the processing node, break them down into multiple subtasks and execute them, and use the reserved computing power of the processing node to perform calculations on the subtasks. Step 6: Construct a task processing framework and record tasks that exceed the allocated time. Step 7: Establish an AI learning module. When the scene information in the storage module does not match the scene information while the vehicle is driving, the module will retrieve the data from the storage module to match the scene information and classify the driving scene. The classified information will then be stored in the storage module. Step 8: Establish a communication connection between the storage module and the cloud, and transmit information between the storage module and the cloud. Then, retrieve and download the corresponding categorized information from the cloud based on the driving scenario.

2. The computing power optimization method for a system chip for autonomous driving in automobiles according to claim 1, characterized in that: In step one, a comparison module is set up. The comparison module is used to distinguish when the vehicle is driving. Based on the road condition information collected in step two, the characteristic information in the road scene is extracted and matched with the driving scene characteristics recorded in the storage module.

3. The computing power optimization method for a system chip for autonomous driving in automobiles according to claim 1, characterized in that: In step two, during vehicle operation, road information is recorded based on the panoramic images carried in the vehicle, thereby obtaining environmental feature information.

4. The computing power optimization method for a system chip for autonomous driving in automobiles according to claim 3, characterized in that: In step two, a judgment module is set up, which is used to determine whether the vehicle driving scene features entered in the input system match. If they match, the current computing power is terminated.

5. The computing power optimization method for a system chip for autonomous driving in automobiles according to claim 4, characterized in that: Step two also includes a command generation module, which generates task commands based on environmental feature information. If there is a mismatch, the process proceeds to step four.

6. The computing power optimization method for a system chip for autonomous driving in automobiles according to claim 1, characterized in that: In step three, processing nodes are set according to the vehicle panoramic image, temporary computing power is allocated from the processing nodes, and tasks that exceed the preset time in step four are imported into the temporary computing power, and budget time is set again for the tasks in the temporary computing power.

7. The computing power optimization method for a system chip for autonomous driving in automobiles according to claim 6, characterized in that: In step five, when the calculation time of a task in the temporary computing power exceeds a preset time, the splitting module is used to split the remaining unfinished task into subtasks and transmit the subtasks to multiple processing nodes.

8. The computing power optimization method for a system chip for autonomous driving in automobiles according to claim 1, characterized in that: Step seven involves using an AI learning module to add scene feature information about the vehicle while it is in motion, and then classifying these features to improve the tasks generated by the autonomous driving system under different factors in the driving scenario.

9. The computing power optimization method for a system chip for autonomous driving in automobiles according to claim 1, characterized in that: Step eight involves communication between the storage module and the cloud, and by classifying different driving scenarios and downloading the corresponding driving scenario information from the cloud, the comparability is improved.

Images