Multi-task dynamic control method for vehicle software upgrade based on deep learning

Abstract

The present invention provides a multi-task dynamic control method for vehicle software upgrade based on deep learning, wherein the method includes: obtaining status information of multiple tasks being executed for vehicle software upgrade; formulating a task control plan based on a pre-trained deep learning model according to the status information of each task being executed; and dynamically controlling each task being executed based on the task control plan. The multi-task dynamic control method for vehicle software upgrade based on deep learning of the present invention, based on a pre-trained deep learning model, formulates a task control plan according to the status information of each task being executed for vehicle software upgrade; and dynamically controls each task being executed based on the task control plan. The system can dynamically control tasks according to the actual usage of the car user, thereby improving the experience of the car user and the intelligence of the vehicle system.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle software upgrade task control, and in particular to a multi-task dynamic control method for vehicle software upgrade based on deep learning. Background Art

[0002] At present, with the increase in the use demand of automobile users and the development of intelligent vehicle computers, more and more vehicle software is loaded on the vehicle computer system. In order to keep the vehicle software fresh and improve the user experience, automobile manufacturers need to regularly upgrade the vehicle software on the vehicle computer system. However, when the vehicle software on the vehicle computer system is upgraded at the same time, multiple upgrade tasks will be generated. The vehicle computer software cannot dynamically adjust the tasks according to the actual use of the car user, resulting in poor user experience and insufficient intelligence of the vehicle computer system.

[0003] Therefore, a solution is urgently needed. Summary of the invention

[0004] The present invention provides a multi-task dynamic control method for vehicle software upgrade based on deep learning. Based on a pre-trained deep learning model, a task control plan is formulated according to the status information of each task being executed during the vehicle software upgrade. Based on the task control plan, each task being executed is dynamically controlled. The system can dynamically control the tasks according to the actual usage of the car users, thereby improving the experience of the car users and the intelligence level of the vehicle system.

[0005] The present invention provides a multi-task dynamic control method for vehicle software upgrade based on deep learning, comprising:

[0006] Obtain status information of multiple tasks being executed for vehicle software upgrade;

[0007] Based on the pre-trained deep learning model, formulate task control plans according to the status information of each task being executed;

[0008] Based on the task control scheme, each executing task is dynamically controlled.

[0009] Preferably, the pre-training step of the deep learning model includes:

[0010] Acquire multiple training samples; wherein the training samples include: standard state information of the marked standard task control scheme;

[0011] Based on each training sample, the deep learning model is trained.

[0012] Preferably, based on the task control scheme, after dynamically controlling each task being executed, the method further includes:

[0013] Based on the task screening condition, at least two target execution tasks are screened out from each of the executing tasks;

[0014] Determine whether the target execution task meets the task triggering conditions;

[0015] When the conditions are met, based on the scheme modification template, the task is executed according to the target that meets the task triggering conditions, and the task control scheme is modified;

[0016] Based on the revised task control scheme, each task being executed continues to be dynamically controlled;

[0017] The task screening conditions include one or more of the following combinations:

[0018] There is a task association relationship between the target execution tasks;

[0019] The matching degree between the first feature description factor of the target execution task and the standard feature description factor is greater than or equal to the matching degree threshold;

[0020] The control priority of the local scheme corresponding to the target execution task in the task control scheme is greater than or equal to the priority threshold;

[0021] The task triggering conditions include one or more of the following combinations:

[0022] The task progress of the target execution task is greater than or equal to the progress threshold;

[0023] The user generates a scenario behavior that matches the triggering behavior in the completed scenario of the target execution task;

[0024] The target execution tasks meet the incremental requirements of the vehicle software.

[0025] Preferably, based on the task control scheme, after dynamically controlling each task being executed, the method further includes:

[0026] Analyze the necessary interaction values ​​of the task control scheme;

[0027] When the interaction necessary value is greater than or equal to the necessary value threshold, the solution visualization query information is displayed to the user;

[0028] When the user inputs a solution visualization request based on the solution visualization query information, a solution visualization model of the task control solution is built;

[0029] Displaying a solution visualization model to the user;

[0030] Assist and guide users to make decisions on solution adjustment strategies based on solution visualization models;

[0031] Adjust the task control plan based on the plan adjustment strategy;

[0032] Based on the adjusted task control plan, continue to dynamically control each task in execution.

[0033] Preferably, the interaction necessary values ​​of the analytical task control scheme include:

[0034] Determine whether the solution type of the resolution task control solution exists in the standard solution type library;

[0035] When the answer is yes, the interaction necessary value is counted as the preset threshold; otherwise, the interaction necessary value corresponding to the second characteristic description factor of the task control scheme is determined from the necessary value library;

[0036] Among them, the preset threshold is greater than or equal to the necessary value threshold.

[0037] Preferably, the scheme visualization model for building the task control scheme includes:

[0038] Based on the visualization processing template, the task control scheme is visualized to obtain the basic visualization model;

[0039] Obtain multimodal data of users;

[0040] Generate a template based on the visual field conditions and generate visual field range conditions based on multimodal data;

[0041] Based on the field of view condition, multiple target field of view ranges are determined from the basic visualization model;

[0042] Creating a quick perspective including at least one target visual range; the quick perspective meets the visualization condition of the target visual range in the quick perspective;

[0043] Based on each quick perspective, generate a quick perspective list;

[0044] The quick view list is set in the basic visualization model to obtain the solution visualization model.

[0045] Preferably, the method of assisting and guiding the user to decide on a solution adjustment strategy based on the solution visualization model includes:

[0046] When a user selects a quick perspective in the quick perspective list when viewing a solution visualization model, the quick perspective selected by the user is used as a target quick perspective;

[0047] Based on the template for generating the adjustment strategy to be selected, the adjustment strategy to be selected is generated according to the visual field content of the target visual field range within the target quick visual field;

[0048] Displaying the adjustment strategies to be selected to the user;

[0049] When the user selects a candidate adjustment strategy, the candidate adjustment strategy selected by the user is used as the solution adjustment strategy.

[0050] The present invention provides a multi-task dynamic control system for vehicle software upgrade based on deep learning, including:

[0051] An acquisition module, used to acquire status information of multiple tasks being executed during vehicle software upgrade;

[0052] The formulation module is used to formulate task control plans based on the pre-trained deep learning model and the status information of each task being executed;

[0053] The control module is used to dynamically control each executing task based on the task control plan.

[0054] The pre-training steps for deep learning models include:

[0055] Acquire multiple training samples; wherein the training samples include: standard state information of the marked standard task control scheme;

[0056] Based on each training sample, the deep learning model is trained.

[0057] The control module dynamically controls each task in execution based on the task control scheme, and also includes:

[0058] Modify the module to include:

[0059] Based on the task screening condition, at least two target execution tasks are screened out from each of the executing tasks;

[0060] Determine whether the target execution task meets the task triggering conditions;

[0061] When the conditions are met, based on the scheme modification template, the task is executed according to the target that meets the task triggering conditions, and the task control scheme is modified;

[0062] Based on the revised task control scheme, each task being executed continues to be dynamically controlled;

[0063] The task screening conditions include one or more of the following combinations:

[0064] There is a task association relationship between the target execution tasks;

[0065] The matching degree between the first feature description factor of the target execution task and the standard feature description factor is greater than or equal to the matching degree threshold;

[0066] The control priority of the local scheme corresponding to the target execution task in the task control scheme is greater than or equal to the priority threshold;

[0067] The task triggering conditions include one or more of the following combinations:

[0068] The task progress of the target execution task is greater than or equal to the progress threshold;

[0069] The user generates a scenario behavior that matches the triggering behavior in the completed scenario of the target execution task;

[0070] The target execution tasks meet the incremental requirements of the vehicle software.

[0071] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description and the accompanying drawings.

[0072] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0074] Figure 1 A flowchart of a multi-task dynamic control method for vehicle software upgrade based on deep learning in an embodiment of the present invention;

[0075] Figure 2 Schematic diagram of a multi-task dynamic control system for vehicle software upgrade based on deep learning in an embodiment of the present invention. DETAILED DESCRIPTION

[0076] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0077] The present invention provides a multi-task dynamic control method for vehicle software upgrade based on deep learning, such as Figure 1 As shown, including:

[0078] S1. Obtaining status information of multiple tasks being executed for vehicle software upgrade;

[0079] S2. Based on the pre-trained deep learning model, formulate a task control plan according to the status information of each task being executed;

[0080] S3. Based on the task control scheme, each task in execution is dynamically controlled.

[0081] The pre-training steps for deep learning models include:

[0082] Acquire multiple training samples; wherein the training samples include: standard state information of the marked standard task control scheme;

[0083] Based on each training sample, the deep learning model is trained.

[0084] The multiple tasks being executed refer to the tasks of upgrading multiple vehicle software on the vehicle system; the status information includes: the type of upgraded software being executed, the occupied upgrade resources, the current usage of the upgraded software by the car user, etc.; the deep learning model is pre-trained, and the deep learning model can formulate a task control plan based on the status information of each task being executed, for example: if the status information is that the type of upgraded software being executed includes educational game software and car navigation software, and the current car user is using the educational game software and not using the car navigation software, then the formulated task control plan is to only control the vehicle system to upgrade the educational game software; based on the task control plan, each task being executed is dynamically controlled; when pre-training the deep learning model, the standard status information of the marked standard task control plan is used, for example: if the standard status information is that the type of upgraded software being executed includes educational game software and car navigation software, and the current car user is using the educational game software and not using the car navigation software, then the marked standard task control plan is to only control the vehicle system to upgrade the educational game software; when the deep learning model is trained with this training sample, the deep learning model can learn how to formulate a task control plan.

[0085] This application is based on a pre-trained deep learning model. According to the status information of each task being executed during the vehicle software upgrade, a task control plan is formulated. Based on the task control plan, each task being executed is dynamically controlled. The system can dynamically control tasks according to the actual usage of the car users, thereby improving the experience of the car users and the intelligence level of the car system.

[0086] In one embodiment, after dynamically regulating each of the executing tasks based on the task regulation scheme, the method further includes:

[0087] Based on the task screening condition, at least two target execution tasks are screened out from each of the executing tasks;

[0088] Determine whether the target execution task meets the task triggering conditions;

[0089] When the conditions are met, based on the scheme modification template, the task is executed according to the target that meets the task triggering conditions, and the task control scheme is modified;

[0090] Based on the revised task control scheme, each task being executed continues to be dynamically controlled;

[0091] The task screening conditions include one or more of the following combinations:

[0092] There is a task association relationship between the target execution tasks;

[0093] The matching degree between the first feature description factor of the target execution task and the standard feature description factor is greater than or equal to the matching degree threshold;

[0094] The control priority of the local scheme corresponding to the target execution task in the task control scheme is greater than or equal to the priority threshold;

[0095] The task triggering conditions include one or more of the following combinations:

[0096] The task progress of the target execution task is greater than or equal to the progress threshold;

[0097] The user generates a scenario behavior that matches the triggering behavior in the completed scenario of the target execution task;

[0098] The target execution tasks meet the incremental requirements of the vehicle software.

[0099] When the executing task meets the task screening conditions, it means that the executing task may need to be regulated first as the target executing task; the target executing task is screened out based on the task screening conditions; when the target executing task meets the task triggering conditions, it means that the target executing task needs to be regulated first; the scheme correction template is to correct the task regulation scheme so that the target executing task that meets the task triggering conditions is regulated first; based on the corrected task regulation scheme, each executing task is continuously dynamically regulated; in the task screening conditions, the task association relationship can be of the same task type, for example: two target executing tasks are to upgrade game software of the same puzzle game type level, but the car user can only use one game software at the same time. Therefore, it may be necessary to regulate the two target execution tasks to determine which game software should be upgraded first, so as to minimize the excessive resource occupation caused by simultaneous upgrades; secondly, the first feature description factor is obtained by feature description processing of the target execution task, and the matching degree threshold can be, for example, 75%. When the matching degree between the first feature description factor and the standard feature description factor is greater than or equal to the matching degree threshold, it means that the target execution task may need to be regulated first. For example, the vehicle software upgrade task constructed into the standard feature description factor is the vehicle application software that the user has used for the longest time in the recent period of time. , indicating that the user prefers to use the vehicle application software, and the upgrade task of the vehicle application software needs to be regulated first; the priority threshold can be, for example, 3; finally, it is also possible to determine whether the target execution task needs to be regulated first based on whether the regulation priority of the local solution corresponding to the target execution task in the task regulation scheme is greater than or equal to the priority threshold. The regulation priority of the local solution represents the degree to which the target execution task may need to be regulated first, which can be set in advance by technical personnel according to actual needs; in the task triggering condition, the task progress of the target execution task is the progress of the system executing the target execution task; the progress threshold can be, for example, 45%; when the target execution task If the task progress is greater than or equal to the progress threshold, the execution priority can be increased to complete the task as soon as possible, provided that the task screening conditions are met; secondly, the completed scenario of the target execution task refers to the scenario in which the vehicle software upgraded when the target execution task is executed is located in the vehicle system, and the triggering behavior is the behavior that the user hopes the target execution task will be completed as soon as possible, for example: frequently checking the upgrade progress of the vehicle software upgraded when the target execution task is executed. Therefore, when the user generates a scenario behavior that matches the triggering behavior in the completed scenario of the target execution task, it means that the execution priority can be increased to complete the task as soon as possible, provided that the task screening conditions are met;The incremental demand for vehicle software refers to the new demand for the use of vehicle software predicted by the vehicle system based on the vehicle's environment. For example, when a vehicle enters an area that it has never entered before, the user may need to use the navigation software. The incremental demand is to use the navigation software. The target execution task is to upgrade the navigation software, which meets the incremental demand. Therefore, when the target execution task meets the incremental demand for vehicle software, it means that the execution priority can be increased and the task execution can be completed as soon as possible under the premise of meeting the task screening conditions. ;

[0100] The deep learning model can dynamically control each task being executed. However, during the dynamic control, as the vehicle's use environment changes, some special situations may arise that require priority control. The embodiments of the present invention can perfectly deal with such situations. The introduction of task screening conditions can preliminarily determine the target execution tasks that may need to be regulated first, and the introduction of task triggering conditions can further determine whether the target execution tasks really need to be regulated first, which greatly improves the comprehensiveness of the control and indirectly improves the user experience.

[0101] In one embodiment, after dynamically regulating each of the executing tasks based on the task regulation scheme, the method further includes:

[0102] Analyze the necessary interaction values ​​of the task control scheme;

[0103] When the interaction necessary value is greater than or equal to the necessary value threshold, the solution visualization query information is displayed to the user;

[0104] When the user inputs a solution visualization request based on the solution visualization query information, a solution visualization model of the task control solution is built;

[0105] Displaying a solution visualization model to the user;

[0106] Assist and guide users to make decisions on solution adjustment strategies based on solution visualization models;

[0107] Adjust the task control plan based on the plan adjustment strategy;

[0108] Based on the adjusted task control plan, continue to dynamically control each task in execution.

[0109] The necessary value for interaction represents the degree of necessity for visually displaying the task control scheme to the user and assisting the user in making a decision on the scheme adjustment strategy; the necessary value threshold can be, for example, 7; when the necessary value for interaction is greater than or equal to the necessary value threshold, the scheme visualization inquiry information is displayed to the user, and the scheme visualization inquiry information is information asking the user whether the user needs to visualize the task control scheme and make a decision on the scheme adjustment strategy; when the user inputs a scheme visualization request based on the scheme visualization inquiry information, a scheme visualization model of the task control scheme is built; the user is assisted in guiding the user to make a decision on the scheme adjustment strategy based on the scheme visualization model; the scheme adjustment strategy is a strategy for adjusting the task control scheme decided by the user according to his actual needs; based on the scheme adjustment strategy, the task control scheme is adjusted; based on the adjusted task control scheme, each task being executed continues to be dynamically regulated. Generally, users may have opinions on the deep learning model for task control. In view of this special situation, the embodiment of the present invention sets a channel for users to make decisions on the scheme adjustment strategy, which is more humane.

[0110] In one embodiment, the interaction necessary values ​​of the analysis task control scheme include:

[0111] Determine whether the solution type of the resolution task control solution exists in the standard solution type library;

[0112] When the answer is yes, the interaction necessary value is counted as the preset threshold; otherwise, the interaction necessary value corresponding to the second characteristic description factor of the task control scheme is determined from the necessary value library;

[0113] Among them, the preset threshold is greater than or equal to the necessary value threshold.

[0114] The standard solution type library contains standard solution types that represent the task control solutions that must be adjusted by the user. For example, when adjusting the upgrade task of the application software recently used by the user, the user needs to intervene to determine whether it is urgent to continue using it. When the solution type of the analysis task control solution exists in the standard solution type library, it means that the task control solution must be adjusted by the user, and the interaction necessary value is calculated as a preset threshold greater than or equal to the necessary value threshold. The necessary value library contains the interaction necessary values ​​corresponding to the second characteristic description factors of different task control solutions. The technical staff can set the corresponding interaction necessary values ​​in advance according to the degree to which different task control solutions require user adjustments.

[0115] In one embodiment, the scheme visualization model for building a task control scheme includes:

[0116] Based on the visualization processing template, the task control scheme is visualized to obtain a basic visualization model; the visualization processing template may be a table that displays the task control schemes one by one at the same time, and during the visualization processing, the task control schemes may be filled into the table;

[0117] Obtain multimodal data of users; multimodal data includes: age, gender, occupation, preference for using vehicle software, etc.;

[0118] Based on the visual field condition generation template, the visual field range condition is generated according to the multimodal data; the visual field condition generation template is, for example, if the multimodal data is that the preference for using vehicle software is to use puzzle game software, then the visual field range condition is generated to include the upgrade control scheme related to the puzzle game software within the determined target visual field;

[0119] Based on the field of view condition, a plurality of target field of view ranges are determined from the basic visualization model; the target field of view ranges meet the field of view condition;

[0120] A quick perspective including at least one target visual range is created; the quick perspective meets the visualization condition of the target visual range in the quick perspective; the visualization condition is a condition representing that each content in the target visual range can be clearly seen in the quick perspective;

[0121] Based on each quick perspective, generate a quick perspective list;

[0122] The quick view list is set in the basic visualization model to obtain the solution visualization model.

[0123] In one embodiment, the method of assisting and guiding the user to decide on a solution adjustment strategy based on the solution visualization model includes:

[0124] When a user selects a quick perspective in the quick perspective list when viewing a solution visualization model, the quick perspective selected by the user is used as a target quick perspective;

[0125] Based on the template for generating the candidate adjustment strategy, the candidate adjustment strategy is generated according to the field of view content of the target field of view within the target quick field of view; for example, if the field of view content is an upgrade control scheme related to puzzle game software, the generated candidate adjustment strategy is to adjust the upgrade priority of the upgrade task, etc.;

[0126] Displaying the adjustment strategies to be selected to the user;

[0127] When the user selects a candidate adjustment strategy, the candidate adjustment strategy selected by the user is used as the solution adjustment strategy.

[0128] The present invention provides a multi-task dynamic control system for vehicle software upgrade based on deep learning, such as Figure 2 As shown, including:

[0129] An acquisition module 1 is used to acquire status information of multiple tasks being executed during the vehicle software upgrade;

[0130] Formulating module 2, which is used to formulate task control plans based on the pre-trained deep learning model and the status information of each task being executed;

[0131] The control module 3 is used to dynamically control each task being executed based on the task control scheme.

[0132] The pre-training steps for deep learning models include:

[0133] Acquire multiple training samples; wherein the training samples include: standard state information of the marked standard task control scheme;

[0134] Based on each training sample, the deep learning model is trained.

[0135] After dynamically regulating each task being executed based on the task regulation scheme, the regulation module 3 further includes:

[0136] Modify the module to include:

[0137] Based on the task screening condition, at least two target execution tasks are screened out from each of the executing tasks;

[0138] Determine whether the target execution task meets the task triggering conditions;

[0139] When the conditions are met, based on the scheme modification template, the task is executed according to the target that meets the task triggering conditions, and the task control scheme is modified;

[0140] Based on the revised task control scheme, each task being executed continues to be dynamically controlled;

[0141] The task screening conditions include one or more of the following combinations:

[0142] There is a task association relationship between the target execution tasks;

[0143] The matching degree between the first feature description factor of the target execution task and the standard feature description factor is greater than or equal to the matching degree threshold;

[0144] The control priority of the local scheme corresponding to the target execution task in the task control scheme is greater than or equal to the priority threshold;

[0145] The task triggering conditions include one or more of the following combinations:

[0146] The task progress of the target execution task is greater than or equal to the progress threshold;

[0147] The user generates a scenario behavior that matches the triggering behavior in the completed scenario of the target execution task;

[0148] The target execution tasks meet the incremental requirements of the vehicle software.

[0149] After dynamically regulating each task being executed based on the task regulation scheme, the regulation module 3 further includes:

[0150] Adjustment modules to include:

[0151] Analyze the necessary interaction values ​​of the task control scheme;

[0152] When the interaction necessary value is greater than or equal to the necessary value threshold, the solution visualization query information is displayed to the user;

[0153] When the user inputs a solution visualization request based on the solution visualization query information, a solution visualization model of the task control solution is built;

[0154] Displaying a solution visualization model to the user;

[0155] Assist and guide users to make decisions on solution adjustment strategies based on solution visualization models;

[0156] Adjust the task control plan based on the plan adjustment strategy;

[0157] Based on the adjusted task control plan, continue to dynamically control each task in execution.

[0158] The adjustment module analyzes the interaction necessary values ​​of the task control scheme, including:

[0159] Determine whether the solution type of the resolution task control solution exists in the standard solution type library;

[0160] When the answer is yes, the interaction necessary value is counted as the preset threshold; otherwise, the interaction necessary value corresponding to the second characteristic description factor of the task control scheme is determined from the necessary value library;

[0161] Among them, the preset threshold is greater than or equal to the necessary value threshold.

[0162] The adjustment module builds a solution visualization model of the task control solution, including:

[0163] Based on the visualization processing template, the task control scheme is visualized to obtain the basic visualization model;

[0164] Obtain multimodal data of users;

[0165] Generate a template based on the visual field conditions and generate visual field range conditions based on multimodal data;

[0166] Based on the field of view condition, multiple target field of view ranges are determined from the basic visualization model;

[0167] Creating a quick perspective including at least one target visual range; the quick perspective meets the visualization condition of the target visual range in the quick perspective;

[0168] Based on each quick perspective, generate a quick perspective list;

[0169] The quick view list is set in the basic visualization model to obtain the solution visualization model.

[0170] The adjustment module assists and guides the user to decide on a solution adjustment strategy based on the solution visualization model, including:

[0171] When a user selects a quick perspective in the quick perspective list when viewing a solution visualization model, the quick perspective selected by the user is used as a target quick perspective;

[0172] Based on the template for generating the adjustment strategy to be selected, the adjustment strategy to be selected is generated according to the visual field content of the target visual field range within the target quick visual field;

[0173] Displaying the adjustment strategies to be selected to the user;

[0174] When the user selects a candidate adjustment strategy, the candidate adjustment strategy selected by the user is used as the solution adjustment strategy.

[0175] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A multi-task dynamic control method for vehicle software upgrade based on deep learning, characterized in that: include: Obtain status information of multiple tasks being executed for vehicle software upgrade; Based on the pre-trained deep learning model, formulate task control plans according to the status information of each task being executed; Based on the task control scheme, each task in execution is dynamically controlled; Based on the task screening condition, at least two target execution tasks are screened out from each of the executing tasks; Determine whether the target execution task meets the task triggering conditions; When the conditions are met, based on the scheme modification template, the task is executed according to the target that meets the task triggering conditions, and the task control scheme is modified; Based on the revised task control scheme, each task being executed continues to be dynamically controlled; The task screening conditions include one or more of the following combinations: There is a task association relationship between the target execution tasks; The matching degree between the first feature description factor of the target execution task and the standard feature description factor is greater than or equal to the matching degree threshold; The control priority of the local scheme corresponding to the target execution task in the task control scheme is greater than or equal to the priority threshold; The task triggering conditions include one or more of the following combinations: The task progress of the target execution task is greater than or equal to the progress threshold; The user generates a scenario behavior that matches the triggering behavior in the completed scenario of the target execution task; The target execution tasks meet the incremental requirements of the vehicle software.

2. The multi-task dynamic control method for vehicle software upgrade based on deep learning as claimed in claim 1, characterized in that: The pre-training steps for deep learning models include: Acquire multiple training samples; wherein the training samples include: standard state information of the marked standard task control scheme; Based on each training sample, the deep learning model is trained.

3. The multi-task dynamic control method for vehicle software upgrade based on deep learning as claimed in claim 1, characterized in that: Based on the task control scheme, after dynamically controlling each task in execution, it also includes: Analyze the necessary interaction values ​​of the task control scheme; When the interaction necessary value is greater than or equal to the necessary value threshold, the solution visualization query information is displayed to the user; When the user inputs a solution visualization request based on the solution visualization query information, a solution visualization model of the task control solution is built; Displaying a solution visualization model to the user; Assist and guide users to make decisions on solution adjustment strategies based on solution visualization models; Adjust the task control plan based on the plan adjustment strategy; Based on the adjusted task control plan, continue to dynamically control each task in execution.

4. The multi-task dynamic control method for vehicle software upgrade based on deep learning as claimed in claim 3, characterized in that: The necessary interactive values ​​of the analysis task control scheme include: Determine whether the solution type of the resolution task control solution exists in the standard solution type library; When the answer is yes, the interaction necessary value is counted as the preset threshold; otherwise, the interaction necessary value corresponding to the second characteristic description factor of the task control scheme is determined from the necessary value library; Among them, the preset threshold is greater than or equal to the necessary value threshold.

5. The multi-task dynamic control method for vehicle software upgrade based on deep learning as claimed in claim 3, characterized in that: The scheme visualization model for building a task control scheme includes: Based on the visualization processing template, the task control scheme is visualized to obtain the basic visualization model; Obtain multimodal data of users; Generate a template based on the visual field conditions and generate visual field range conditions based on multimodal data; Based on the field of view condition, multiple target field of view ranges are determined from the basic visualization model; Creating a quick perspective including at least one target visual range; the quick perspective meets the visualization condition of the target visual range in the quick perspective; Based on each quick perspective, generate a quick perspective list; The quick view list is set in the basic visualization model to obtain the solution visualization model.

6. The multi-task dynamic control method for vehicle software upgrade based on deep learning as claimed in claim 5, characterized in that: The auxiliary guidance of the user to decide on a solution adjustment strategy based on the solution visualization model includes: When a user selects a quick perspective in the quick perspective list when viewing a solution visualization model, the quick perspective selected by the user is used as a target quick perspective; Based on the template for generating the adjustment strategy to be selected, the adjustment strategy to be selected is generated according to the visual field content of the target visual field range within the target quick visual field; Displaying the selected adjustment strategies to the user; When the user selects a candidate adjustment strategy, the candidate adjustment strategy selected by the user is used as the solution adjustment strategy.

7. A multi-task dynamic control system for vehicle software upgrade based on deep learning, characterized in that: include: An acquisition module, used to acquire status information of multiple tasks being executed during vehicle software upgrade; The formulation module is used to formulate task control plans based on the pre-trained deep learning model and the status information of each task being executed; The control module is used to dynamically control each task in execution based on the task control scheme; Modify the module to include: Based on the task screening condition, at least two target execution tasks are screened out from each of the executing tasks; Determine whether the target execution task meets the task triggering conditions; When the conditions are met, based on the scheme modification template, the task is executed according to the target that meets the task triggering conditions, and the task control scheme is modified; Based on the revised task control scheme, each task being executed continues to be dynamically controlled; The task screening conditions include one or more of the following combinations: There is a task association relationship between the target execution tasks; The matching degree between the first feature description factor of the target execution task and the standard feature description factor is greater than or equal to the matching degree threshold; The control priority of the local scheme corresponding to the target execution task in the task control scheme is greater than or equal to the priority threshold; The task triggering conditions include one or more of the following combinations: The task progress of the target execution task is greater than or equal to the progress threshold; The user generates a scenario behavior that matches the triggering behavior in the completed scenario of the target execution task; The target execution tasks meet the incremental requirements of the vehicle software.

8. The multi-task dynamic control system for vehicle software upgrade based on deep learning as claimed in claim 7, characterized in that: The pre-training steps for deep learning models include: Acquire multiple training samples; wherein the training samples include: standard state information of the marked standard task control scheme; Based on each training sample, the deep learning model is trained.

Images