Running state optimization method and device, electronic equipment and storage medium

Abstract

The application discloses a running state optimization method and device, electronic equipment and a storage medium, and relates to the technical field of intelligent devices. The method comprises the following steps: a terminal device acquires state information in at least one dimension, the state information in the at least one dimension is used for determining the running state of the terminal device, the optimization scene type information is determined according to the state information in the at least one dimension, the optimization strategy information for the terminal device is determined based on the optimization scene type information, and the running state of the terminal device is optimized through the optimization strategy information. By identifying the optimization scene type information corresponding to the state information in at least one dimension of the terminal device and determining the optimization strategy of the terminal device based on the optimization scene type information, the differentiated optimization strategy based on the optimization scene type is formulated, which helps to improve the precision of device state optimization and improve the user experience.

Description

Technical Field

[0001] This application relates to the field of smart device technology, specifically to a method, apparatus, electronic device, and storage medium for optimizing operating status. Background Technology

[0002] With the widespread application of smart terminals, smart home devices, IoT devices, and edge computing devices, these devices continuously generate a large amount of status data during operation, such as processor power consumption, screen power consumption, communication module power consumption, battery level, device temperature, application running status, and user interaction behavior. By identifying and analyzing these device states, energy consumption, performance, and user experience can be optimized and controlled.

[0003] In existing technologies, the device or server typically identifies the current device state, generates corresponding optimization instructions based on the identification results, and controls the device to execute these instructions. For example, when high CPU power consumption is detected, the system can directly generate instructions to reduce the processor frequency, limit background tasks, or close some high-power processes; when high screen power consumption is detected, the system can directly generate instructions to reduce screen brightness, reduce refresh rate, or shorten screen illumination time. This type of solution typically employs a linear processing approach of "state identification—instruction generation—instruction execution."

[0004] However, current equipment status optimization is still relatively crude, and it does not generate differentiated strategies for different equipment statuses, resulting in the need to improve the accuracy of equipment status optimization. Summary of the Invention

[0005] This application provides a method, apparatus, electronic device, and storage medium for optimizing operating status, which can improve the accuracy of device status optimization and enhance user experience.

[0006] In a first aspect, embodiments of this application provide a method for optimizing operating status, applied to a terminal device, the method comprising: The status information of the terminal device in at least one dimension is obtained, and the status information in at least one dimension is used to determine the operating status of the terminal device. Determine the optimization scenario type information based on the state information of at least one dimension; Based on the optimization scenario type information, optimization strategy information for the terminal device is determined; The operating status of the terminal device is optimized using the optimization strategy information.

[0007] Secondly, embodiments of this application also provide an operating state optimization device, applied to a terminal device, the device comprising: The acquisition module is used to acquire the status information of the terminal device in at least one dimension, and the status information in at least one dimension is used to determine the operating status of the terminal device; A type determination module is used to determine the optimized scenario type information based on the state information of at least one dimension. The strategy determination module is used to determine optimization strategy information for the terminal device based on the optimization scenario type information; An optimization module is used to optimize the operating status of the terminal device using the optimization strategy information.

[0008] Optionally, in some embodiments of this application, determining the optimized scene type information based on the state information of the at least one dimension includes: Feature extraction is performed on the state information of at least one dimension to obtain state feature information, which includes at least one of the following: whether the source of power consumption anomaly is clear, whether it involves multi-module collaboration, whether it affects user experience, whether it involves cross-device linkage, and whether the risk of direct optimization is below a threshold. The optimized scenario type information is determined based on the state feature information.

[0009] Optionally, in some embodiments of this application, determining the optimization strategy information for the terminal device based on the optimization scenario type information includes: If the optimized scenario type information includes a direct execution scenario, then the optimized strategy information for the state information of the at least one dimension is generated by the local agent; If the optimized scenario type information includes a cloud collaboration scenario, then a strategy request message for the status information of the at least one dimension is sent to the cloud, and the optimized strategy information for the strategy request message returned by the cloud is received, wherein the optimized strategy information is generated by the cloud through a cloud intelligent agent; If the optimized scenario type information includes a user confirmation scenario, then the optimized strategy information for the state information of the at least one dimension is generated through a local intelligent agent or a cloud intelligent agent.

[0010] Optionally, in some embodiments of this application, the step of generating the optimization strategy information for the state information of the at least one dimension through a local intelligent agent includes: The optimized strategy information for the state information of the at least one dimension is matched from the local policy library by the local agent or the optimized strategy information for the state information of the at least one dimension is generated directly by the local agent. The local policy library is optimized based on the group policy issued by the cloud. The group policy is generated by the cloud based on the knowledge graph construction and analysis of the knowledge summary reported by the cloud agent on at least one terminal device or by federated learning. The knowledge summary includes at least one of the following: policy hash for optimization policy information, score for optimization effect, or status identification information for running status.

[0011] Optionally, in some embodiments of this application, the specific steps of optimizing the local policy library based on the group policy distributed from the cloud include: Receive the group policy sent from the cloud; The first fusion weight corresponding to the group strategy, the second fusion weight corresponding to the local strategy library, and the third fusion weight corresponding to user personalization are determined based on the strategy influencing factors. The strategy influencing factors include at least one of confidence level, historical effect, data coverage, scene familiarity, user status, device health status, and network status. Based on the first fusion weight, the second fusion weight, and the third fusion weight, the group policy and the local policy in the local policy library are fused to obtain the fused policy. The local policy library is updated based on the merged policy.

[0012] Optionally, in some embodiments of this application, the updating of the local policy library further includes: Determine at least one candidate strategy information corresponding to the running status; The reward estimate for each candidate strategy is designed based on a reward function, which is based on at least one of power saving, performance loss, or negative user feedback. The optimization strategy information corresponding to the running state is determined based on the exploration probability, utilization probability, and reward estimation corresponding to each candidate strategy information. The local policy library is updated based on the operating status and the optimization strategy information.

[0013] Optionally, in some embodiments of this application, the knowledge graph construction and analysis includes constructing a knowledge graph and generating cloud-based policy information, wherein several cloud-based policy information constitute the group policy; The knowledge graph includes scene nodes, strategy nodes, device nodes, and effect nodes. Each node includes its own attributes and relationship descriptions, and edges between nodes are established through the relationship descriptions. The generated cloud-based strategy information includes: Based on the strategy effect, sample size, and risk screening strategy nodes indicated by the strategy nodes, cloud-based strategy information is obtained.

[0014] Optionally, in some embodiments of this application, the group strategy generated based on federated learning specifically includes: A knowledge graph is constructed based on the knowledge summary. The knowledge graph includes scene nodes, strategy nodes, device nodes, and effect nodes. Each node includes its own attributes and relationship descriptions. Edges between nodes are established through the relationship descriptions. Based on the strategy node, at least one associated effect node is determined; Calculate at least one statistical indicator based on the effect node, the statistical indicator including at least one of average power saving rate, average performance loss rate, or sample device coverage; The strategy fitness is obtained by fusing the statistical indicators of the at least one dimension, and the strategy fitness is used to determine the degree of acceptance of the strategy corresponding to the strategy node in the knowledge graph. The target strategy is identified according to the fitness of the strategy, and the target strategy is used as the parent to undergo crossover mutation to obtain the offspring strategy. The offspring strategy was verified; The group strategy is constructed based on the target strategy and the validated offspring strategy.

[0015] Optionally, in some embodiments of this application, optimizing the operating state of the terminal device through the optimization strategy information includes: If the optimization strategy information involves user experience, then a graphical interface is generated and displayed based on the optimization strategy information, and the user's selection operation on the graphical interface is received, and the strategy action corresponding to the selection operation is executed. If the optimization strategy information involves sufficient verification or the execution risk is below a threshold, then the strategy action corresponding to the optimization strategy information is executed directly.

[0016] Thirdly, embodiments of this application also provide an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it implements the steps in the above-described operation state optimization method.

[0017] Fourthly, embodiments of this application also provide a storage medium, including a computer-readable storage medium on which a computer program is stored, and which, when executed by a processor, implements the steps in the above-described operation state optimization method.

[0018] Fifthly, embodiments of this application also provide a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the methods provided in the various optional implementations described in embodiments of this application.

[0019] In summary, the terminal device in this application embodiment acquires state information in at least one dimension, the state information in at least one dimension is used to determine the operating state of the terminal device, optimize scenario type information is determined based on the state information in at least one dimension, optimize strategy information for the terminal device is determined based on the optimized scenario type information, and optimize the operating state of the terminal device through the optimization strategy information.

[0020] In this embodiment, by identifying the optimization scenario type information corresponding to at least one dimension of the terminal device's state information, and controlling the determination of the terminal device's optimization strategy based on the optimization scenario type information, the formulation of differentiated optimization strategies based on the optimization scenario type is realized, which helps to improve the accuracy of device state optimization and enhance user experience. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a schematic diagram illustrating a scenario where a terminal device, as provided in an embodiment of this application, executes the operation state optimization method; Figure 2 This is a flowchart illustrating the operation state optimization method provided in the embodiments of this application; Figure 3 This is a schematic diagram of the operating state optimization device provided in the embodiments of this application; Figure 4 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application.

[0023] Explanation of icon numbers: 101-Terminal device; 301-Acquisition module; 302-Type determination module; 303-Strategy determination module; 304-Optimization module; 401-Processor; 402-Memory; 403-Power supply; 404-Input unit. Detailed Implementation

[0024] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] In the description of the embodiments of this application, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships based on the orientation or positional relationships shown in the accompanying drawings, are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention. Furthermore, the terms "first," "second," "third," and "fourth" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, features defined with "first," "second," "third," and "fourth" may explicitly or implicitly include one or more features. In the description of the present invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0026] In this application, the term "exemplary" is used to mean "serving as an example, illustration, or description." Any embodiment described as "exemplary" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.

[0027] This application provides a method, apparatus, electronic device, and storage medium for optimizing operating status. Specifically, this application provides an apparatus for optimizing the operating status of electronic devices, including terminal devices such as mobile phones, tablets, laptops, televisions, air conditioners, or washing machines.

[0028] For example, please see Figure 1 , Figure 1This is a schematic diagram illustrating a scenario where a terminal device, according to an embodiment of this application, executes the operation state optimization method. Specifically, the execution process of the terminal device executing the operation state optimization method is as follows: Terminal device 101 acquires its own status information in at least one dimension, the status information in at least one dimension is used to determine the operating status of the terminal device, determines optimization scenario type information based on the status information in at least one dimension, determines optimization strategy information for the terminal device based on the optimization scenario type information, and optimizes the operating status of the terminal device through the optimization strategy information.

[0029] For example, taking a mobile phone as a terminal device, the mobile phone detects state information in multiple dimensions through sensors, determines the optimization scenario type information based on this state information, and then controls the generation of optimization strategy information that is specific to the terminal device and matches the optimization scenario type information according to the optimization scenario type information. The optimization strategy is then executed to optimize the operating state of the terminal device.

[0030] In summary, the embodiments of this application identify the optimization scenario type information corresponding to at least one dimension of the terminal device's state information, and control the determination of the terminal device's optimization strategy based on the optimization scenario type information. This enables the formulation of differentiated optimization strategies based on the optimization scenario type, which helps to improve the accuracy of device state optimization and enhance user experience.

[0031] The following sections provide detailed descriptions of each example. It should be noted that the order in which the embodiments are described is not intended to limit the priority of the embodiments.

[0032] Please see Figure 2 , Figure 2 This is a flowchart illustrating the operation state optimization method provided in this application embodiment. Although the flowchart shows a logical order, in some cases, the steps shown or described can be executed in a different order than that shown in the flowchart. Specifically, the specific flow of the operation state optimization method is as follows: S201. Obtain the status information of the terminal device in at least one dimension, wherein the status information in at least one dimension is used to determine the operating status of the terminal device.

[0033] Here, dimension refers to the source of state information. In the embodiments of this application, dimension includes hardware layer, system layer, application layer or environment layer, etc.

[0034] Among them, state information refers to the state values ​​corresponding to each dimension. The state information of each dimension can be obtained through sensors or state data stream monitoring. For example, state information for the hardware layer may include CPU power consumption, GPU power consumption, baseband power consumption, screen power consumption, Wi-Fi power consumption, or Bluetooth power consumption; state information for the system layer may include process CPU time slice utilization, memory usage, I / O operation frequency, interrupt frequency, scheduler status, etc.; state information for the application layer may include foreground application package name, background service activity status, network request frequency, rendering frame rate, etc.; state information for the environment layer may include screen brightness, ambient light intensity, network signal strength, battery temperature, charging status, or device posture, etc.

[0035] Among them, the operating status is an abstract or comprehensive representation of the terminal device's operating condition. It is used to represent the current operating status of the terminal device. For example, the operating status can be abstractly represented as excellent, good, medium or poor, or the operating status can also include comprehensive representations such as GPS being idle for a long time, the current ambient light being very bright but the brightness being manually set to 80%, and heavy load in the game scene.

[0036] Understandably, acquiring status information from multiple dimensions helps to comprehensively determine the current operating status of terminal devices, improves the accuracy of operating status analysis, and thus helps to improve the rationality and effectiveness of subsequent operating status optimization.

[0037] S202. Determine the optimized scenario type information based on the state information of at least one dimension.

[0038] Among them, the optimization scenario type information is category information determined based on the current state of the terminal device, used to characterize the decision complexity, execution risk, user impact, or collaborative requirements of the current optimization task. Accordingly, the specific optimization scenario type to which the current state of the terminal device belongs can be determined based on whether it involves decision complexity, execution risk, user impact, or collaborative requirements.

[0039] In this embodiment, the optimization scenario type information can be determined by analyzing the characteristics of each state. For example, in this embodiment, the optimization scenario type information includes direct execution scenarios, cloud collaboration scenarios, or user confirmation scenarios.

[0040] S203. Determine optimization strategy information for the terminal device based on the optimization scenario type information.

[0041] For example, different optimization scenario types correspond to different generation strategies or methods. In this embodiment, the corresponding strategy or method is selected based on the optimization scenario type information corresponding to the current operating state of the terminal device to control the generation of the corresponding optimization strategy information.

[0042] It is understandable that by optimizing scenario type information to determine the optimization strategy information for the terminal device, and realizing the formulation of differentiated strategies based on the optimization scenario type, it is helpful to improve the accuracy of strategy formulation and the adaptability of optimization strategies to the operating status of terminal devices.

[0043] S204. Optimize the operating status of the terminal device using the optimization strategy information.

[0044] For example, the operating status of the terminal device can be adjusted according to the specific adjustment parameters corresponding to the optimization strategy information. For instance, the terminal device may perform actions such as turning off the GPS module, reducing the CPU frequency to 2.0GHz, and reducing the GPU frequency to 500MHz.

[0045] In summary, the embodiments of this application identify the optimization scenario type information corresponding to at least one dimension of the terminal device's state information, and control the determination of the terminal device's optimization strategy based on the optimization scenario type information. This enables the formulation of differentiated optimization strategies based on the optimization scenario type, which helps to improve the accuracy of device state optimization and enhance user experience.

[0046] Specifically, in this application embodiment, feature extraction can be performed on state information from multiple dimensions, and the extracted features can be used to analyze the optimization scenario type information corresponding to the current state of the terminal device. That is, optionally, in some embodiments of this application, the step "determining optimization scenario type information based on the at least one dimension of state information" includes: Feature extraction is performed on the state information of at least one dimension to obtain state feature information, which includes at least one of the following: whether the source of power consumption anomaly is clear, whether it involves multi-module collaboration, whether it affects user experience, whether it involves cross-device linkage, and whether the risk of direct optimization is below a threshold. The optimized scenario type information is determined based on the state feature information.

[0047] For example, by analyzing the status information of terminal devices in multiple dimensions, it can be determined whether the source of power consumption anomalies is clear, whether it involves multi-module collaboration, whether it affects user experience, whether it involves cross-device linkage, and whether the risk of direct optimization is below the threshold, thereby determining the optimization scenario type information corresponding to the current operating status of the terminal device.

[0048] For example, if the status feature information includes multi-module collaboration or cross-device linkage, the current optimization scenario type is determined to be a cloud collaboration scenario; if the status feature information indicates that the source of power consumption anomaly is clear or the direct optimization risk is below the threshold, the current optimization scenario type is determined to be a direct execution scenario; if the status feature information includes impact on user experience, the current optimization scenario type is determined to be a user confirmation scenario.

[0049] In this embodiment, state feature information can be extracted through parameter normalization, threshold judgment, correlation analysis, etc. For example, the state feature information corresponding to the current running state can be extracted through mean, maximum value, state change correlation, user perception, whether other devices are connected, etc.

[0050] In this embodiment of the application, differentiated optimization strategy information can be generated based on the determined optimization scenario type. Specifically, in some embodiments of this application, the step "determining optimization strategy information for the terminal device based on the optimization scenario type information" includes: If the optimized scenario type information includes a direct execution scenario, then the optimized strategy information for the state information of the at least one dimension is generated by the local agent; If the optimized scenario type information includes a cloud collaboration scenario, then a strategy request message for the status information of the at least one dimension is sent to the cloud, and the optimized strategy information for the strategy request message returned by the cloud is received, wherein the optimized strategy information is generated by the cloud through a cloud intelligent agent; If the optimized scenario type information includes a user confirmation scenario, then the optimized strategy information for the state information of the at least one dimension is generated through a local intelligent agent or a cloud intelligent agent.

[0051] Specifically, for local execution scenarios or user confirmation scenarios, optimization strategy information can be generated directly through the local intelligent agent on the terminal device. For cloud collaboration scenarios or user confirmation scenarios, optimization strategy information can be generated through the cloud intelligent agent in the cloud.

[0052] Among them, direct execution scenarios refer to scenarios where the strategy can be executed directly, such as optimization scenarios that have been fully verified and have low execution risk, such as turning off idle GPS modules or reducing WiFi scanning frequency when the screen is off; user confirmation scenarios refer to scenarios that require user confirmation before execution, such as adjusting screen brightness or restricting background application activities that affect user experience; cloud collaboration scenarios refer to complex optimization scenarios, such as scenarios involving multiple components and cross-device linkage, such as heavy-load scenarios in games, which require cloud-based intelligent agents to provide optimization suggestions after in-depth analysis.

[0053] Accordingly, in the embodiments of this application, for scenarios requiring user confirmation, the optimization strategy information requires user confirmation before it can be executed. That is, optionally, in some embodiments of this application, the step "optimizing the operating state of the terminal device through the optimization strategy information" includes: If the optimization strategy information involves user experience, then a graphical interface is generated and displayed based on the optimization strategy information, and the user's selection operation on the graphical interface is received, and the strategy action corresponding to the selection operation is executed. If the optimization strategy information involves sufficient verification or the execution risk is below a threshold, then the strategy action corresponding to the optimization strategy information is executed directly.

[0054] Understandably, having users confirm whether to execute a set strategy through a graphical interface allows the strategy to be executed in a way that better meets user needs, avoiding direct execution that might negatively impact the user experience.

[0055] Among these, the directly executed policy actions, in addition to shutting down when the hardware is idle, can also include switching to the mobile network when the WiFi signal is <-75dBm and the mobile signal is >-85dBm, and freezing the application process after the application has been in the background for more than 30 minutes and is not a communication application.

[0056] In summary, the embodiments of this application implement a differentiated optimization mechanism for terminal devices based on different optimization scenario types. For example, for a long-term idle GPS operation state that can be directly executed locally, the optimization strategy information for shutting down the GPS module can be directly generated by the local intelligent agent, and the GPS module can be shut down by executing the shutdown of the GPS module. For an operation state where the screen brightness is too high and requires user confirmation, a pop-up window to reduce the screen brightness can be generated by the local intelligent agent or the cloud intelligent agent. The pop-up window asks the user whether the screen brightness can be reduced. If the user chooses to reduce it, the instruction or operation to reduce the screen brightness is executed. For a heavy-load operation state in a game scenario, the optimization strategy request information is sent to the cloud. The optimization strategy information is generated collaboratively by the cloud and then sent to the terminal from the cloud. The terminal can then execute the optimization strategy directly or confirm whether it can be executed through a pop-up window.

[0057] In this embodiment, a local policy library can be constructed to match optimization policy information suitable for the current operating state. This local policy library represents the unique optimization experience of the terminal device, recording optimization policy information applicable to multiple operating states. For example, the local policy library includes personalized policies, adaptive policies, and basic policies. Personalized policies include user preference records (e.g., refusing to reduce brightness by more than 20%), habitual optimization modes (e.g., automatically activating subway mode during commuting), and interaction history statistics (85% acceptance rate for Category B policies). Adaptive policies include hardware feature adaptation (e.g., voltage-frequency curves for specific SoCs), system version optimization (tuning parameters for different Android versions), and long-term effect records (policy execution history and effect tracking). Basic policies include scene recognition models (scene features such as games / videos / reading), quick response policies (e.g., a set of directly executable policies), and security boundary parameters (safe operating range of each hardware component). It is understood that configuring multi-layered policies enhances the comprehensiveness of the local policy library.

[0058] In this application embodiment, the local policy library can be optimized based on the reward policy to improve its accuracy. Specifically, in some embodiments of this application, updating the local policy library includes: Determine at least one candidate strategy information corresponding to the running status; The reward estimate for each candidate strategy is designed based on a reward function, which is based on at least one of power saving, performance loss, or negative user feedback. The optimization strategy information corresponding to the running state is determined based on the exploration probability, utilization probability, and reward estimation corresponding to each candidate strategy information. The local policy library is updated based on the operating status and the optimization strategy information.

[0059] It is understandable that designing reward estimates for each optimization strategy by constructing a reward function based on multiple rewards such as power saving, performance loss, and negative user feedback can help improve the performance of the final selected strategy across multiple reward dimensions. In other words, it makes the selected strategy more conducive to reducing power consumption, performance loss, and negative user feedback.

[0060] In this embodiment, the strategy selection mechanism can employ a contextual gambling machine learning model. For example, for each running state s, there is a corresponding set of candidate strategy information A(s) = {a1, a2, ..., an}, and then the expected reward estimate for each candidate strategy information is configured as: Q(s,a) = (1-η)×Q(s,a) +η×R(s,a). A high Q(s,a) value indicates that performing action a in state s yields a good reward; a low Q(s,a) value indicates that historical experience shows the action is ineffective.

[0061] Where η refers to the learning rate, which decays over time: η_t = η0 / (1 + decay×t), decay refers to the decay factor / decay coefficient, used to reduce the weight of historical data or historical feedback on the current decision; R(s,a): refers to the immediate reward function, the design of which is as follows: R(s,a)=λ1×ΔPower_saving-λ2×ΔPerformance_loss-λ3×User_penalty; Wherein, ΔPower_saving refers to power saving, ΔPower_saving = (P_before - P_after) / P_before; ΔPerformance_loss refers to the performance loss, ΔPerformance_loss = max(0, Perf_before - Perf_after) / Perf_before; User_penalty refers to negative user feedback; it is 1 when the user explicitly refuses, and 0 otherwise.

[0062] Among them, λ1, λ2, and λ3 can be pre-configured based on experience.

[0063] Furthermore, based on the action selection strategy ε-greedy, max_a Q(s,a) is selected according to the utilization probability 1-ε, and new actions are randomly explored with the detection probability ε to filter the optimization strategy information corresponding to the motion state.

[0064] For example, if the current motion state s is a high-load game or high-temperature sub-scene, its corresponding candidate strategy information A(s) includes three options: a1 = CPU frequency reduction by 10%, a2 = CPU frequency reduction by 15%, and a3 = no adjustment. The estimated rewards are Q(s,a1) = 0.7, Q(s,a2) = 0.3, and Q(s,a3) = 0.2, respectively. The system selects a1 (CPU frequency reduction by 10%) with a high probability according to the ε-greedy strategy and outputs the corresponding hardware control instruction for execution.

[0065] After execution, power consumption decreased by 15% (ΔPower_saving=0.15), performance decreased by 2% (ΔPerformance_loss=0.02), and the user did not object (User_penalty=0). The reward value R(s,a1)=λ1×0.15 -λ2×0.02 -λ3×0 is a relatively high positive value. The system uses this R value to update Q(s,a1), further improving the Q value of this action from 0.7. This increases the system's confidence in choosing to reduce the CPU frequency by 10% when encountering similar scenarios in the future.

[0066] In this embodiment of the application, the local policy library can also be optimized and updated in conjunction with the group policy distributed from the cloud. That is, optionally, in some embodiments of this application, the step "generating the optimized policy information for the state information of the at least one dimension through the local intelligent agent" includes: The optimized strategy information for the state information of the at least one dimension is matched from the local policy library by the local agent or the optimized strategy information for the state information of the at least one dimension is generated directly by the local agent. The local policy library is optimized based on the group policy issued by the cloud. The group policy is generated by the cloud based on the knowledge graph construction and analysis of the knowledge summary reported by the cloud agent on at least one terminal device or by federated learning. The knowledge summary includes at least one of the following: policy hash for optimization policy information, score for optimization effect, or status identification information for running status.

[0067] For example, the cloud receives knowledge summaries reported by multiple terminal devices and generates a group policy for these devices through knowledge graph construction and analysis, or federated learning. This group policy is then distributed to the current terminal devices, which update their local policy libraries based on the group policy, resulting in an optimized local policy library. It is understandable that this cloud-based group policy-updated local policy library can take into account both local and cloud / group overall needs and trends.

[0068] Understandably, this local policy library can not only meet local personalized needs, but also serve as a basis for decision-making when the terminal device is offline or disconnected from the network, thus avoiding the inability to provide optimization strategies due to offline or disconnected network conditions.

[0069] In this embodiment, when generating optimization strategy information through a local intelligent agent, optimization strategies can be directly matched in parallel through a local strategy library, and the instructions corresponding to the optimization strategies can be preloaded into the execution queue in advance to improve the device optimization efficiency.

[0070] The knowledge summary encapsulates raw power consumption data, policy data, and performance data to prevent the leakage of users' private data, such as raw power consumption data. For example, the knowledge summary includes policy hashes for optimization policy information, scores for optimization effects, or status identifiers for running status. In other words, what is uploaded is the encapsulated identifier data or transformed data from the raw data, not the user's raw data directly. This protects user privacy. Furthermore, abstracting / encapsulating raw data into identifier data also helps reduce the amount of data that needs to be uploaded.

[0071] In this embodiment, the local policy library and the cloud-based group policy can be fused based on fusion weights to obtain the final optimized policy information. That is, optionally, in some embodiments of this application, the specific steps for optimizing the local policy library based on the cloud-deployed group policy include: Receive the group policy sent from the cloud; The first fusion weight corresponding to the group strategy, the second fusion weight corresponding to the local strategy library, and the third fusion weight corresponding to user personalization are determined based on the strategy influencing factors. The strategy influencing factors include at least one of confidence level, historical effect, data coverage, scene familiarity, user status, device health status, and network status. Based on the first fusion weight, the second fusion weight, and the third fusion weight, the group policy and the local policy in the local policy library are fused to obtain the fused policy. The local policy library is updated based on the merged policy.

[0072] It is understandable that weighted fusion improves the rationality and effectiveness of the fused strategy. A fusion weight determination scheme based on the factors influencing the strategy further enhances the rationality and effectiveness of the fusion weights. For example, when the confidence level of a cloud-based strategy is high, the cloud-based group strategy is assigned a higher fusion weight; when the historical performance of a local strategy is good, the local strategy library is assigned a higher fusion weight; and since user personalization is less affected by the confidence level and historical performance of cloud-based strategies, the fusion weight corresponding to user personalization is lower.

[0073] Understandably, besides the confidence level of cloud-based strategies and the historical performance of local strategies, other policy-influencing factors include data coverage, scenario familiarity, user status, device health, and network status. These factors are used to boost or reduce the final fusion weight. For example, if a strategy is proven effective on a large number of devices (broad coverage), the fusion weight corresponding to the group strategy will be increased (i.e., the weight value will be biased towards the upper limit of the range), because broad coverage implies high reliability of the conclusion. On the other hand, the coverage of local strategies is naturally limited to the current device (narrow coverage), so the fusion weight corresponding to the local strategy library will be relatively conservative, with no increase or even a slight discount. User preferences only represent personal will and there is no comparison of coverage, so this factor has virtually no impact on the personalized fusion weight for users, and a fixed median value is taken.

[0074] For general scenarios (such as video playback), the cloud has massive amounts of data, making it more reliable, so the fusion weight corresponding to the group strategy is increased; for scenarios specific to the current device (such as specific user operating habits), local strategies have more say, so the fusion weight corresponding to the local strategy library is increased; the user's historical preferences directly reflect the acceptance in that scenario, serving as the basis for the user's personalized fusion weight.

[0075] The device health status includes the device's current real-time status, such as the device's temperature, remaining battery power, and charging status. Since the cloud cannot perceive the device's current real-time status, the fusion weight corresponding to the local policy library can be appropriately increased, while the fusion weight corresponding to the cloud-based group policy can be appropriately decreased.

[0076] For example, taking a specific parameter p to be optimized as an example, its final value V_fused after fusion is calculated as follows: V_fused(p) = w_cloud(p) × V_cloud(p) + w_local(p) × V_local(p) + w_user(p) × V_user(p); in: w_cloud(p)=α×Confidence_cloud(p)×Coverage_cloud(p)×(1- Network_latency); w_local(p) = β × Effectiveness_local(p) × Scene_familiarity ×Device_health; w_user(p) = γ×Consistency_user(p) × Recent_feedback(p) ×Engagement_level; Constraints: w_cloud + w_local + w_user = 1.

[0077] Wherein, V_cloud(p) represents the suggested value given by the cloud-based agent based on the knowledge graph for the optimization parameter p, such as the cloud suggesting a screen brightness of 60%; V_local(p) represents the suggested value given by the local agent based on the private policy library for the optimization parameter p, such as the local agent suggesting a screen brightness of 65% based on historical learning; V_user(p) represents the suggested value given by the user based on historical preferences for the optimization parameter p, such as the user's historical behavior showing a preferred brightness of 70%.

[0078] Among them, Confidence_cloud (cloud-based policy confidence): the value ranges from 0 to 1, and is determined by the sample size and performance consistency of the policy in the knowledge graph. The larger the sample size and the smaller the performance variance, the higher the value, indicating that the cloud is more confident in the effectiveness of the policy.

[0079] Coverage_cloud (Policy Coverage Percentage): A value ranging from 0 to 1 indicates the percentage of terminal devices that have verified the policy within the knowledge graph. A higher coverage percentage indicates broader applicability and higher credibility of the policy.

[0080] Effectiveness_local (Local Strategy Historical Performance Rating): Values ​​range from 0 to 1, calculated from the Q-values ​​recorded in the local strategy library. The better the strategy's historical performance in this scenario, the higher this value.

[0081] Scene_familiarity: The value ranges from 0 to 1, and is obtained by normalizing the frequency of the scene's occurrence on the current device. The more times it occurs, the higher the value, indicating that the local agent has richer decision-making experience with that scene.

[0082] Consistency_user (User Preference Consistency): Values ​​range from 0 to 1, measuring the stability of a user's historical feedback to similar strategies. A higher value indicates consistent feedback when a user repeatedly executes the same type of strategy; a lower value indicates fluctuating feedback, such as sometimes accepting and sometimes rejecting.

[0083] Recent_feedback: Values ​​range from 0 to 1, recording the percentage of users who accepted similar strategies within a recent period (e.g., the last 7 days or the last 10 interactions). The higher the recent acceptance rate, the higher this value.

[0084] Engagement_level (User Engagement): Values ​​range from 0 to 1, measuring the level of user activity in interacting with Type B strategies. The higher the frequency with which users actively adjust optimization preferences and view optimization reports, the higher this value, indicating a stronger willingness for users to participate in power consumption optimization.

[0085] Device_health (Device Health Status): The value ranges from 0 to 1 and is calculated based on real-time physical indicators such as the current device temperature, remaining battery power, and charging status. This value is higher when the device temperature is normal and the battery power is sufficient; and lower when the temperature is too high or the battery power is extremely low, so as to remind the algorithm to prioritize a conservative strategy when the device is in poor condition.

[0086] Network_latency: The value ranges from 0 to 1 and is obtained by normalizing the real-time network communication latency between the cloud and the terminal. The higher the latency, the higher this value, and the more obvious the discount effect on w_cloud—when the network is poor, the weight of the cloud policy will be actively reduced to avoid untimely policy delivery due to excessive latency.

[0087] Where w_cloud(p) represents the first fusion weight corresponding to the parameter p to be optimized. This weight is jointly determined by α, Confidence_cloud (cloud policy confidence), Coverage_cloud (policy coverage terminal ratio), and Network_latency (network latency).

[0088] Where w_local(p) represents the second fusion weight corresponding to the parameter p to be optimized. This weight is jointly determined by β, Effectiveness_local (historical effect score of local strategy), Scene_familiarity (scene familiarity), and Device_health (device health status).

[0089] Where w_user(p) represents the third fusion weight corresponding to the parameter p to be optimized. This weight is jointly determined by γ, Consistency_user (consistency of user preferences), Recent_feedback (recent user feedback), and Engagement_level (user engagement).

[0090] Wherein, α, β, and γ are adjustment coefficients, with default values ​​of α=0.4, β=0.35, and γ=0.25, which can be dynamically adjusted based on empirical values.

[0091] For example, suppose the optimization parameter p that needs to be decided is "screen brightness percentage". The cloud-based group strategy suggests V_cloud = 60% (based on the best statistical value under massive similar scenarios), the local private strategy library suggests V_local = 65% based on historical learning (because the device has repeatedly achieved good results with 65% brightness in this scenario), and user preference records show user habits V_user = 70% (historical manual adjustment records show that the user prefers higher brightness).

[0092] Assuming the three calculated weights are w_cloud = 0.3, w_local = 0.4, and w_user = 0.3 (the sum of which is 1), the final actual screen brightness is: V_fused = 0.3×60% + 0.4×65% + 0.3×70% = 18% + 26% + 21% = 65%. The system ultimately sets the screen brightness to 65%, incorporating big data experience from the cloud, respecting the historical learning results of local strategies, and also partially considering user preferences.

[0093] If the decision is made on a different parameter, such as "maximum CPU frequency," the suggested values ​​for that parameter will differ among the cloud, local, and user parties, and the corresponding weights will be recalculated (because factors such as policy influencing factors may have different values ​​for different parameters). Finally, these weighted values ​​are independently merged to arrive at the execution value for that parameter. Each optimization parameter undergoes the above weighted fusion calculation independently.

[0094] In this embodiment of the application, the group strategy in the cloud can be implemented by constructing and analyzing a knowledge graph. For example, the knowledge summary uploaded to the cloud can be organized into a knowledge network of state, strategy, device, and effect. This knowledge network is the knowledge graph. The analysis method is to find the best strategy (optimization strategy information) suitable for the state from the knowledge network and answer questions such as 'which strategy is most effective in which state'.

[0095] For example, in this embodiment of the application, the knowledge graph construction and analysis includes constructing a knowledge graph and generating cloud strategy information, and several cloud strategy information constitutes the group strategy; The knowledge graph includes scene nodes, strategy nodes, device nodes, and effect nodes. Each node includes its own attributes and relationship descriptions, and edges between nodes are established through the relationship descriptions. The generated cloud-based strategy information includes: Based on the strategy effect, sample size, and risk screening strategy nodes indicated by the strategy nodes, cloud-based strategy information is obtained.

[0096] Scene nodes, also known as status nodes, indicate the operating status of a device. The attributes of a scene node include scene labels, feature vectors, and frequency of occurrence. The relationships between scene nodes include hardware configuration, optimization strategies, and performance metrics. The feature vector is a quantitative indicator of the scene node; it can be a set of data, such as "the average CPU load of this scene is X%, the average GPU load is Y%, and the average screen-on power consumption is Z milliwatts." These data collectively constitute the mathematical description of the scene. Frequency of occurrence refers to the cumulative number of times the scene / operating status is identified across all terminals, used to measure the prevalence of the scene; it is also a statistical attribute of the node itself. Relationships describe the connections between the scene node and other nodes. Hardware configuration, optimization strategies, and performance metrics correspond to three types of connections between the scene node and device nodes, strategy nodes, and performance nodes, respectively. These connections constitute the edges between nodes.

[0097] Similarly, policy nodes are used to indicate policies. The attributes of a policy node include a policy hash, a parameter set, or a confidence level. The policy hash is the hash representation of the policy, and the parameter set is a collection of parameters designed for the policy, such as screen brightness and CPU load. The relationships within a policy node include applicable scenarios, effect records, and terminal types, used to identify connections with status nodes, effect nodes, and device nodes, respectively.

[0098] Device nodes are used to indicate devices / device types, such as which / which type of device a policy applies to in a particular operational state. The attributes of a device node include device fingerprint, hardware configuration, or user profile. The relationships between device nodes include usage scenarios / operational states, private policies, or effect feedback.

[0099] Effect nodes indicate the effect of a strategy applied to a given state. Attributes of these effect nodes include power saving rate, performance retention, and user acceptance. Relationships within these effect nodes include strategy-scenario (state) pairs, execution time, and terminal type.

[0100] For example, the construction of a knowledge graph includes the following steps: After receiving an experience summary from a terminal, the cloud first parses out the "scene tag," "policy hash," and "effect data." For the "scene tag," the cloud first checks the existing graph to see if a scene node with the same name or feature vector already exists. If it already exists, the node is reused, and its "occurrence frequency" attribute is updated (incremented by 1). If it does not exist, a new scene node is created, using the scene tag as the node ID, storing the scene feature vector from the experience summary in the node attribute, and initializing the occurrence frequency to 1. For the "policy hash," the cloud similarly checks the existing graph to see if a policy node with the same hash value already exists. If it already exists, it is reused; if it does not exist, a new policy node is created, using the hash value as the node ID, storing the policy parameter set in the node attribute, and initializing the confidence score to a low base value (because the sample size is insufficient). The effect data is processed in the same way. If the "scene-policy-effect" combination is appearing for the first time, a new effect node is created; otherwise, new statistical data (such as updating the average power saving rate, cumulative sample count, etc.) is aggregated on the existing effect node. Understandably, the entire process is incremental. Each time a new summary is received, the map performs "query-reuse or create-update attributes" operations on the existing data to continuously add new points and new annotations.

[0101] After establishing the knowledge graph, superior cloud-based strategy information can be selected by clustering similar scenarios / states, identifying anomalous nodes, and extracting high-frequency strategies. For example, clustering similar scenarios identifies scenarios / states with similar power consumption characteristics, identifying anomalous nodes filters to anomalous nodes, and ranking strategy effectiveness through strategy effectiveness scoring. Based on this, strategy nodes are selected as cloud-based strategy information based on strategy effectiveness, sample size, and execution risk. Specifically: for a target scenario (or a class of similar scenarios / states), all "strategy-effect" records related to that scenario are retrieved from the knowledge graph. These records are sorted according to their effectiveness scores, and strategies that meet the following conditions are selected as cloud-based strategy information: sufficient sample size (to avoid small sample bias), average power saving rate higher than a threshold, average performance loss rate lower than a threshold, and user acceptance rate higher than a threshold.

[0102] In addition, to integrate and align local and cloud policies, or to select matching policies based on status and device, this embodiment of the application also performs joint encapsulation based on scene / device status, cloud policy information, and statistical results. For example, the encapsulated structured rule entry is: When the terminal is in scene X (or scene group G), it is recommended to execute policy P. Based on the statistical data of N devices, the expected power saving rate is A%, the expected performance loss rate is B%, and the user acceptance rate is C%.

[0103] In this application embodiment, a cloud-based group policy can also be constructed using federated learning. Specifically, in some embodiments of this application, the group policy generated based on federated learning includes: A knowledge graph is constructed based on the knowledge summary. The knowledge graph includes scene nodes, strategy nodes, device nodes, and effect nodes. Each node includes its own attributes and relationship descriptions. Edges between nodes are established through the relationship descriptions. Based on the strategy node, at least one associated effect node is determined; Calculate at least one statistical indicator based on the effect node, the statistical indicator including at least one of average power saving rate, average performance loss rate, or sample device coverage; The strategy fitness is obtained by fusing the statistical indicators of the at least one dimension, and the strategy fitness is used to determine the degree of acceptance of the strategy corresponding to the strategy node in the knowledge graph. The target strategy is identified according to the fitness of the strategy, and the target strategy is used as the parent to undergo crossover mutation to obtain the offspring strategy. The offspring strategy was verified; The group strategy is constructed based on the target strategy and the validated offspring strategy.

[0104] For example, for a given strategy node, find all effect nodes connected to that strategy node in the knowledge graph (i.e., all effect records of the strategy historically). Extract the following three statistical metrics from these effect nodes: Average power saving rate: The average power saving rate across all effect records, representing the power saving effect of this strategy.

[0105] Average performance loss rate: The average performance loss rate across all performance records, representing the degree of negative impact of the strategy on user experience.

[0106] Sample device coverage: This indicates the number of different devices that have executed the strategy and reported the results, representing the adequacy of the strategy's validation. A larger sample size generally indicates more reliable statistical results.

[0107] The above three indicators are weighted and combined into a single comprehensive fitness score (strategy fitness) according to the following logic: Strategy fitness = w1 × average power saving rate - w2 × average performance loss rate + w3 × normalized value of sample coverage. Where w1, w2, and w3 are preset weighting coefficients that control the influence of power saving effect, performance protection, and reliability (determined based on sample device coverage) in the final score, respectively. Generally, w1 and w3 are positive, while w2 is negative to penalize strategies that save power but severely sacrifice performance.

[0108] Among them, the identification of target strategies is based on strategy fitness. For example, strategy nodes are sorted according to their strategy fitness, and the strategies ranked at the top (e.g., the top 10%) are identified as target strategies.

[0109] In this embodiment of the application, strategies with large sample sizes and stable effects can also be selected as target strategies based on the ranking of strategy fitness. These strategies may not be ranked high, but they have stable effects, good consistency, and certain selection value.

[0110] In this application embodiment, abnormal policies can also be filtered by identifying potential risks. For example, if a policy performs well on most devices, but its power consumption increases instead of decreasing or its performance is abnormally low on a specific model device, the cloud will mark the "scenario-policy-device model" combination as high-risk, and will automatically filter out the policy when it is subsequently sent to that model device. Alternatively, there is a potential risk if two different policies work simultaneously and conflict, such as "CPU downclocking by 10%" and "CPU upclocking to performance mode," which target the same hardware objective but have opposite directions. Policies with user acceptance rates below a threshold also pose a potential risk.

[0111] It is understandable that the target strategy is a set of devices that have undergone multiple screening processes, achieving optimization results while ensuring controllable performance loss, and covering a wide range of sample devices, thus possessing superior guidance for device optimization. In the embodiments of this application, more strategies can be developed through crossover mutation, and then the optimal strategy can be selected through verification.

[0112] The cross-validation process involves taking two effective target policies as "parent policies," breaking each down into fragments with multiple parameter dimensions, and then recombining these fragments to generate a "child policy." For example, suppose that after policy fitness evaluation, the knowledge graph identifies the following two target policies: Parent Policy A (highly effective in game scenarios): includes three parameters—CPU frequency reduced to 90% of nominal frequency, GPU frequency reduced to 85% of nominal frequency, and screen brightness remaining unchanged. Parent Policy B (highly effective in video playback scenarios): includes three parameters—CPU frequency reduced to 85% of nominal frequency, GPU frequency maintained at 100% (no frequency reduction), and screen brightness reduced by 20%.

[0113] A new strategy that performs better or is more balanced in a game scenario is generated through cross-operations. Specifically, the parameter fragment "CPU frequency reduced by 90%" is taken from parent A, and the parameter fragments "GPU frequency kept at 100%" and "screen brightness reduced by 20%" are taken from parent B. These three are combined into a new child strategy C. Child strategy C consists of: CPU frequency reduced to 90% of nominal frequency, GPU frequency kept at 100% (no frequency reduction), and screen brightness reduced by 20%. Understandably, this child strategy C may never have been tried before in the existing knowledge graph; for example, no terminal has previously executed this specific parameter combination simultaneously in a game scenario. The cloud uses it as a candidate child strategy, preparing to push it to some terminals for distributed verification.

[0114] Mutation involves randomly fine-tuning a parameter of an existing policy to generate a new policy that is similar to but slightly different from the original. Referring to the example above, a mutation operation is performed on child policy C: the parameter "reduce screen brightness by 20%" is randomly adjusted to "reduce screen brightness by 15%" (with a change of ±5%), thus generating a new child policy D. Child policy D contains the following: CPU frequency reduced to 90% of nominal frequency, GPU frequency maintained at 100% (no downclocking), and screen brightness reduced by 15%.

[0115] In this embodiment, the offspring strategies generated through crossover and mutation (such as strategy C and strategy D) are not directly distributed to all devices. Instead, they are first pushed to a small number of eligible terminal devices (e.g., devices with active game scenarios and user consent) through a canary distribution method. After these terminals execute the offspring strategies, they upload an experience summary of "scene tag + strategy hash + effect score" back to the cloud. The knowledge graph records the effect data of these new strategies. If the verification effect is good, the fitness score of these offspring strategies will gradually increase, and they will eventually be included in the group strategy library and deployed on a large scale; if the effect is poor, they will be naturally eliminated. Thus, expanding strategies through crossover and mutation and verifying the effect helps to select the optimal cloud strategy.

[0116] In summary, the embodiments of this application identify the optimization scenario type information corresponding to at least one dimension of the terminal device's state information, and control the determination of the terminal device's optimization strategy based on the optimization scenario type information. This enables the formulation of differentiated optimization strategies based on the optimization scenario type, which helps to improve the accuracy of device state optimization and enhance user experience.

[0117] Furthermore, terminal devices optimize their local policy libraries through local learning, while the cloud optimizes group policies through knowledge graphs and federated learning. Policy alignment and fusion achieve mutual promotion and address both general and personalized needs, which also helps improve user acceptance. Knowledge graphs can also improve the efficiency of generating optimization strategies in cloud-based collaborative scenarios.

[0118] In this embodiment, the terminal device and the cloud can adjust the frequency of strategy optimization to achieve real-time response or short-term periodic response updates.

[0119] Furthermore, the embodiments of this application screen optimization strategies from multiple aspects such as power consumption and performance to improve the energy efficiency balance during device optimization. Energy efficiency ratio = power saving rate / (1 + performance loss rate).

[0120] In this application embodiment, the operation state optimization method is also applicable to the scenario of device state prediction. For example, predicting the next state of the terminal device based on user habits, such as predicting the next usage scenario of the terminal device and thus predicting the operation state of the device in advance, and then matching optimization strategy information in advance through the predicted operation state, and then performing device preheating or cooling treatment in advance, so as to provide the operating environment or state for the next processing of the device in advance.

[0121] To facilitate better implementation of the operational state optimization method of this application, this application also provides an operational state optimization apparatus based on the above-described operational state optimization method. The meanings of the terms used are the same as in the above-described operational state optimization method, and specific implementation details can be found in the descriptions of the method embodiments.

[0122] Please see Figure 3 , Figure 3 This is a schematic diagram of the operating state optimization device provided in an embodiment of this application. The operating state optimization device is applied to a terminal device, and the operating state optimization device can be specifically as follows: The acquisition module 301 is used to acquire the status information of the terminal device in at least one dimension, and the status information in at least one dimension is used to determine the operating status of the terminal device. The type determination module 302 is used to determine the optimized scenario type information based on the state information of at least one dimension. The strategy determination module 303 is used to determine optimization strategy information for the terminal device based on the optimization scenario type information; The optimization module 304 is used to optimize the operating status of the terminal device through the optimization strategy information.

[0123] Optionally, in some embodiments of this application, determining the optimized scene type information based on the state information of the at least one dimension includes: Feature extraction is performed on the state information of at least one dimension to obtain state feature information, which includes at least one of the following: whether the source of power consumption anomaly is clear, whether it involves multi-module collaboration, whether it affects user experience, whether it involves cross-device linkage, and whether the risk of direct optimization is below a threshold. The optimized scenario type information is determined based on the state feature information.

[0124] Optionally, in some embodiments of this application, determining the optimization strategy information for the terminal device based on the optimization scenario type information includes: If the optimized scenario type information includes a direct execution scenario, then the optimized strategy information for the state information of the at least one dimension is generated by the local agent; If the optimized scenario type information includes a cloud collaboration scenario, then a strategy request message for the status information of the at least one dimension is sent to the cloud, and the optimized strategy information for the strategy request message returned by the cloud is received, wherein the optimized strategy information is generated by the cloud through a cloud intelligent agent; If the optimized scenario type information includes a user confirmation scenario, then the optimized strategy information for the state information of the at least one dimension is generated through a local intelligent agent or a cloud intelligent agent.

[0125] Optionally, in some embodiments of this application, the step of generating the optimization strategy information for the state information of the at least one dimension through a local intelligent agent includes: The optimized strategy information for the state information of the at least one dimension is matched from the local policy library by the local agent or the optimized strategy information for the state information of the at least one dimension is generated directly by the local agent. The local policy library is optimized based on the group policy issued by the cloud. The group policy is generated by the cloud based on the knowledge graph construction and analysis of the knowledge summary reported by the cloud agent on at least one terminal device or by federated learning. The knowledge summary includes at least one of the following: policy hash for optimization policy information, score for optimization effect, or status identification information for running status.

[0126] Optionally, in some embodiments of this application, the specific steps of optimizing the local policy library based on the group policy distributed from the cloud include: Receive the group policy sent from the cloud; The first fusion weight corresponding to the group strategy, the second fusion weight corresponding to the local strategy library, and the third fusion weight corresponding to user personalization are determined based on the strategy influencing factors. The strategy influencing factors include at least one of confidence level, historical effect, data coverage, scene familiarity, user status, device health status, and network status. Based on the first fusion weight, the second fusion weight, and the third fusion weight, the group policy and the local policy in the local policy library are fused to obtain the fused policy. The local policy library is updated based on the merged policy.

[0127] Optionally, in some embodiments of this application, the updating of the local policy library further includes: Determine at least one candidate strategy information corresponding to the running status; The reward estimate for each candidate strategy is designed based on a reward function, which is based on at least one of power saving, performance loss, or negative user feedback. The optimization strategy information corresponding to the running state is determined based on the exploration probability, utilization probability, and reward estimation corresponding to each candidate strategy information. The local policy library is updated based on the operating status and the optimization strategy information.

[0128] Optionally, in some embodiments of this application, the knowledge graph construction and analysis includes constructing a knowledge graph and generating cloud-based policy information, wherein several cloud-based policy information constitute the group policy; The knowledge graph includes scene nodes, strategy nodes, device nodes, and effect nodes. Each node includes its own attributes and relationship descriptions, and edges between nodes are established through the relationship descriptions. The generated cloud-based strategy information includes: Based on the strategy effect, sample size, and risk screening strategy nodes indicated by the strategy nodes, cloud-based strategy information is obtained.

[0129] Optionally, in some embodiments of this application, the group strategy generated based on federated learning specifically includes: A knowledge graph is constructed based on the knowledge summary. The knowledge graph includes scene nodes, strategy nodes, device nodes, and effect nodes. Each node includes its own attributes and relationship descriptions. Edges between nodes are established through the relationship descriptions. Based on the strategy node, at least one associated effect node is determined; Calculate at least one statistical indicator based on the effect node, the statistical indicator including at least one of average power saving rate, average performance loss rate, or sample device coverage; The strategy fitness is obtained by fusing the statistical indicators of the at least one dimension, and the strategy fitness is used to determine the degree of acceptance of the strategy corresponding to the strategy node in the knowledge graph. The target strategy is identified according to the fitness of the strategy, and the target strategy is used as the parent to undergo crossover mutation to obtain the offspring strategy. The offspring strategy was verified; The group strategy is constructed based on the target strategy and the validated offspring strategy.

[0130] Optionally, in some embodiments of this application, optimizing the operating state of the terminal device through the optimization strategy information includes: If the optimization strategy information involves user experience, then a graphical interface is generated and displayed based on the optimization strategy information, and the user's selection operation on the graphical interface is received, and the strategy action corresponding to the selection operation is executed. If the optimization strategy information involves sufficient verification or the execution risk is below a threshold, then the strategy action corresponding to the optimization strategy information is executed directly.

[0131] In this embodiment, the acquisition module 301 acquires the status information of the terminal device in at least one dimension, which is used to determine the operating status of the terminal device. The type determination module 302 determines the optimization scenario type information based on the status information in at least one dimension. The strategy determination module 303 determines the optimization strategy information for the terminal device based on the optimization scenario type information. The optimization module 304 optimizes the operating status of the terminal device through the optimization strategy information.

[0132] In summary, the embodiments of this application identify the optimization scenario type information corresponding to at least one dimension of the terminal device's state information, and control the determination of the terminal device's optimization strategy based on the optimization scenario type information. This enables the formulation of differentiated optimization strategies based on the optimization scenario type, which helps to improve the accuracy of device state optimization and enhance user experience.

[0133] In addition, this application also provides an electronic device, such as Figure 4 As shown, it illustrates a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Specifically: The electronic device may include components such as a processor 401 with one or more processing cores, a memory 402 with one or more computer-readable storage media, a power supply 403, and an input unit 404. Those skilled in the art will understand that... Figure 4 The electronic device structure shown does not constitute a limitation on the electronic device and may include more or fewer components than shown, or combine certain components, or have different component arrangements. Wherein: The processor 401 is the control center of the electronic device. It connects various parts of the electronic device via various interfaces and lines. By running or executing software programs and / or modules stored in the memory 402, and by calling data stored in the memory 402, it performs various functions and processes data, thereby providing overall monitoring of the electronic device. Optionally, the processor 401 may include one or more processing cores; preferably, the processor 401 may integrate an application processor and a modem processor, wherein the application processor mainly handles the operating system, user interface, and applications, and the modem processor mainly handles wireless communication. It is understood that the modem processor may not be integrated into the processor 401.

[0134] The memory 402 can be used to store software programs and modules. The processor 401 executes various functional applications and data processing by running the software programs and modules stored in the memory 402. The memory 402 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, applications required for at least one function, etc.; the data storage area may store data created according to the use of the electronic device, etc. In addition, the memory 402 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device. Accordingly, the memory 402 may also include a memory controller to provide the processor 401 with access to the memory 402.

[0135] The electronic device also includes a power supply 403 that supplies power to the various components. Preferably, the power supply 403 can be logically connected to the processor 401 through a power management system, thereby enabling functions such as charging, discharging, and power consumption management through the power management system. The power supply 403 may also include one or more DC or AC power supplies, recharging systems, power equipment debugging circuits, power converters or inverters, power status indicators, and other arbitrary components.

[0136] The electronic device may also include an input unit 404, which can be used to receive input digital or character information, and generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function control.

[0137] Although not shown, the electronic device may also include a display unit, etc., which will not be described in detail here. Specifically, in this embodiment, the processor 401 in the electronic device loads the executable files corresponding to the processes of one or more applications into the memory 402 according to the following instructions, and the processor 401 runs the applications stored in the memory 402, thereby implementing the steps in any of the running state optimization methods provided in the embodiments of this application.

[0138] The terminal device in this application embodiment acquires state information in at least one dimension, the state information in at least one dimension is used to determine the operating state of the terminal device, optimizes scenario type information is determined based on the state information in at least one dimension, optimizes strategy information for the terminal device is determined based on the optimized scenario type information, and optimizes the operating state of the terminal device through the optimization strategy information.

[0139] In this embodiment, by identifying the optimization scenario type information corresponding to at least one dimension of the terminal device's state information, and controlling the determination of the terminal device's optimization strategy based on the optimization scenario type information, the formulation of differentiated optimization strategies based on the optimization scenario type is realized, which helps to improve the accuracy of device state optimization and enhance user experience.

[0140] For details on the implementation of each of the above operations, please refer to the previous examples, which will not be repeated here.

[0141] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be performed by instructions, or by instructions controlling related hardware. These instructions can be stored in a computer-readable storage medium and loaded and executed by a processor.

[0142] Therefore, this application provides a storage medium, including a computer-readable storage medium on which a computer program is stored, the computer program being loadable by a processor to execute the steps in any of the runtime optimization methods provided in this application.

[0143] For details on the implementation of each of the above operations, please refer to the previous examples, which will not be repeated here.

[0144] The computer-readable storage medium may include: read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0145] Since the instructions stored in the computer-readable storage medium can execute the steps in any of the operation state optimization methods provided in this application, the beneficial effects that any of the operation state optimization methods provided in this application can achieve can be realized, as detailed in the preceding embodiments, and will not be repeated here.

[0146] The above provides a detailed description of the operation state optimization method, apparatus, electronic device, and computer-readable storage medium provided in this application. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, those skilled in the art will recognize that there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

[0147] It is understood that in the specific implementation of this application, data related to the status information and optimization effects of the terminal device in multiple dimensions are involved. When the embodiments in this application are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

Claims

1. A method for optimizing operational states, characterized in that, Applied to a terminal device, the method includes: The status information of the terminal device in at least one dimension is obtained, and the status information in at least one dimension is used to determine the operating status of the terminal device. Determine the optimization scenario type information based on the state information of at least one dimension; Based on the optimization scenario type information, optimization strategy information for the terminal device is determined; The operating status of the terminal device is optimized using the optimization strategy information.

2. The operation state optimization method according to claim 1, characterized in that, Determining the optimized scenario type information based on the state information of at least one dimension includes: Feature extraction is performed on the state information of at least one dimension to obtain state feature information, which includes at least one of the following: whether the source of power consumption anomaly is clear, whether it involves multi-module collaboration, whether it affects user experience, whether it involves cross-device linkage, and whether the risk of direct optimization is below a threshold. The optimized scenario type information is determined based on the state feature information.

3. The operation state optimization method according to claim 1, characterized in that, The step of determining the optimization strategy information for the terminal device based on the optimization scenario type information includes: If the optimized scenario type information includes a direct execution scenario, then the optimized strategy information for the state information of the at least one dimension is generated by the local agent; If the optimized scenario type information includes a cloud collaboration scenario, then a strategy request message for the status information of the at least one dimension is sent to the cloud, and the optimized strategy information for the strategy request message returned by the cloud is received, wherein the optimized strategy information is generated by the cloud through a cloud intelligent agent; If the optimized scenario type information includes a user confirmation scenario, then the optimized strategy information for the state information of the at least one dimension is generated through a local intelligent agent or a cloud intelligent agent.

4. The operation state optimization method according to claim 3, characterized in that, The optimization strategy information generated by the local agent for the state information of the at least one dimension includes: The optimized policy information for the state information of the at least one dimension is matched from the local policy library by the local agent or the optimized policy information for the state information of the at least one dimension is generated directly by the local agent. The local policy library is optimized based on the group policy issued by the cloud. The group policy is generated by the cloud based on the knowledge graph construction and analysis of the knowledge summary reported by the cloud agent on at least one terminal device or by federated learning. The knowledge summary includes at least one of the following: policy hash for optimization policy information, score for optimization effect, or status identification information for running status.

5. The operation state optimization method according to claim 4, characterized in that, The specific steps for optimizing the local policy library based on the group policies distributed from the cloud include: Receive the group policy sent from the cloud; The first fusion weight corresponding to the group strategy, the second fusion weight corresponding to the local strategy library, and the third fusion weight corresponding to user personalization are determined based on the strategy influencing factors. The strategy influencing factors include at least one of confidence level, historical effect, data coverage, scene familiarity, user status, device health status, and network status. Based on the first fusion weight, the second fusion weight, and the third fusion weight, the group policy and the local policy in the local policy library are fused to obtain the fused policy. The local policy library is updated based on the merged policy.

6. The operation state optimization method according to claim 4, characterized in that, The update to the local policy library also includes: Determine at least one candidate strategy information corresponding to the running status; The reward estimate for each candidate strategy is designed based on a reward function, which is based on at least one of power saving, performance loss, or negative user feedback. The optimization strategy information corresponding to the running state is determined based on the exploration probability, utilization probability, and reward estimation corresponding to each candidate strategy information. The local policy library is updated based on the operating status and the optimization strategy information.

7. The operation state optimization method according to claim 4, characterized in that, The knowledge graph construction and analysis includes constructing a knowledge graph and generating cloud-based strategy information, wherein several cloud-based strategy information constitute the group strategy. The knowledge graph includes scene nodes, strategy nodes, device nodes, and effect nodes. Each node includes its own attributes and relationship descriptions, and edges between nodes are established through the relationship descriptions. The generated cloud-based strategy information includes: Based on the strategy effect, sample size, and risk screening strategy nodes indicated by the strategy nodes, cloud-based strategy information is obtained.

8. The operation state optimization method according to claim 4, characterized in that, The group strategy, generated based on federated learning, specifically includes: A knowledge graph is constructed based on the knowledge summary. The knowledge graph includes scene nodes, strategy nodes, device nodes, and effect nodes. Each node includes its own attributes and relationship descriptions. Edges between nodes are established through the relationship descriptions. Based on the strategy node, at least one associated effect node is determined; Calculate at least one statistical indicator based on the effect node, the statistical indicator including at least one of average power saving rate, average performance loss rate, or sample device coverage; The strategy fitness is obtained by fusing the statistical indicators of the at least one dimension, and the strategy fitness is used to determine the degree of recognition of the strategy corresponding to the strategy node in the knowledge graph. The target strategy is identified according to the fitness of the strategy, and the target strategy is used as the parent to undergo crossover mutation to obtain the offspring strategy. The offspring strategy was verified; The group strategy is constructed based on the target strategy and the validated offspring strategy.

9. The operation state optimization method according to claim 1, characterized in that, The optimization of the terminal device's operating state using the optimization strategy information includes: If the optimization strategy information involves user experience, then a graphical interface is generated and displayed based on the optimization strategy information, and the user's selection operation on the graphical interface is received, and the strategy action corresponding to the selection operation is executed. If the optimization strategy information involves sufficient verification or the execution risk is below a threshold, then the strategy action corresponding to the optimization strategy information is executed directly.

10. An operating state optimization device, characterized in that, Applied to a terminal device, the device includes: The acquisition module is used to acquire the status information of the terminal device in at least one dimension, and the status information in at least one dimension is used to determine the operating status of the terminal device; A type determination module is used to determine the optimized scenario type information based on the state information of at least one dimension. The strategy determination module is used to determine optimization strategy information for the terminal device based on the optimization scenario type information; An optimization module is used to optimize the operating status of the terminal device using the optimization strategy information.

11. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the runtime optimization method as described in any one of claims 1-9.

12. A storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the steps of the operation state optimization method as described in any one of claims 1-9.

Images