Equipment action planning method, system and related device

Abstract

The invention discloses an equipment action planning method and system and a related device. The method comprises the following steps: receiving a current request input by a target user; obtaining target user portrait data formed based on at least one target memory tag corresponding to the target user, wherein the target memory tag represents a health state and / or equipment use preference corresponding to the target user; acquiring target comprehensive environment data based on the at least one target device; the current request, the target user portrait data and the target comprehensive environment data are input into an action planning model, target control parameters for at least one target device are output, and the action planning model is obtained through training based on sample training data of multiple known action planning results; the action planning is carried out in combination with multi-source information such as the current request, the target user portrait and the target comprehensive environment, so that the accuracy and the automation efficiency of action planning are improved, the automation regulation and control level of equipment is improved, and more personalized and intelligent service experience is provided for the user.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a method, system and related apparatus for planning device motion. Background Technology

[0002] As a core scenario for deep interaction between people and services, and between people and the environment, the home is becoming an important arena for the application of artificial intelligence. AI to Home (AI2H), as a new generation of smart home technology, is leading the global telecommunications industry into a new development cycle for home services. Health status, as one of the basic human states and needs, is closely related to individual human characteristics and personalized memories. How to better utilize human health status and memory information to drive the automatic adjustment and planning of smart devices is a crucial issue that urgently needs to be addressed. Summary of the Invention

[0003] This application provides a device motion planning method, system, and related apparatus. By utilizing a motion planning model combined with multi-source information such as "current request, target user profile, and target comprehensive environment" to generate target control parameters, it replaces the limitations of traditional rule-based control parameter matching or motion planning based on a single information source. This improves both the accuracy and automation efficiency of motion planning, as well as the level of automated device control, providing users with a more personalized and intelligent service experience. The above technical solution is as follows: In a first aspect, embodiments of this application provide a device motion planning method, the method comprising: Receive the current request input by the target user; Obtain target user profile data corresponding to the aforementioned target users; the aforementioned target user profile data is composed of at least one target memory tag corresponding to the aforementioned target users; the aforementioned target memory tag represents the relevant health status and / or device usage preferences of the aforementioned target users; Acquire comprehensive environmental data of the target based on at least one target device; The above-mentioned current request, the above-mentioned target user profile data, and the above-mentioned target comprehensive environment data are input into the action planning model, and the output is the target control parameters for at least one target device; the above-mentioned action planning model is trained based on sample training data of multiple known action planning results.

[0004] Secondly, embodiments of this application provide a device for planning equipment movements, the device comprising: The receiving module is used to receive the current request input by the target user; The first acquisition module is used to acquire target user profile data corresponding to the aforementioned target user; the aforementioned target user profile data is composed of at least one target memory tag corresponding to the aforementioned target user; the aforementioned target memory tag represents the relevant health status and device usage preferences of the aforementioned target user. The second acquisition module is used to acquire comprehensive environmental data of the target based on at least one target device; The action planning module is used to input the current request, the target user profile data, and the target comprehensive environment data into the action planning model, and output the target control parameters for at least one target device; the action planning model is trained based on sample training data of multiple known action planning results.

[0005] Thirdly, embodiments of this application provide a device motion planning system, the system comprising a server and at least one target device connected to the server; wherein: The aforementioned target device is used to report the current request it received from the target user to the aforementioned server; The aforementioned server is configured to execute the device action planning method steps provided in the first aspect of the present application based on the aforementioned current request, and to transmit corresponding target control parameters to each of the aforementioned target devices, so that each of the aforementioned target devices performs a corresponding target action based on the corresponding target control parameters.

[0006] Fourthly, embodiments of this application provide an electronic device, including: a processor and a memory; The processor is connected to the memory. The aforementioned memory is used to store executable program code; The processor reads the executable program code stored in the memory to run a program corresponding to the executable program code, so as to execute the method provided in the first aspect of the embodiments of this specification.

[0007] Fifthly, embodiments of this application provide a computer storage medium storing a plurality of instructions adapted for loading by a processor and executing the method steps provided in the first aspect of embodiments of this application.

[0008] Sixthly, embodiments of this specification provide a computer program product containing instructions that, when run on a computer or processor, cause the computer or processor to perform the method steps provided by the first aspect of the embodiments of this specification or any possible implementation thereof.

[0009] The beneficial effects of the technical solutions provided in some embodiments of this application include at least the following: In one or more embodiments of this application, firstly, by receiving the current request input by the target user, the user's immediate device control needs are directly captured, ensuring that the subsequent action planning process can start from the user's current real needs, avoiding misunderstandings of needs due to information transmission delays or deviations, and providing a clear and accurate initial direction for the entire intelligent action planning process. Secondly, target user profile data based on at least one target memory tag corresponding to the target user is obtained, and a personalized profile is constructed using the target memory tags representing the target user's health status and / or device usage preferences. This shifts the basis of the action planning model when planning device actions from generalized assumptions to the user's real personalized characteristics and needs, effectively avoiding the incompatibility of general solutions with different user health needs or usage habits, and ensuring that the subsequently output target control parameters can accurately match the target user's personalized needs. Simultaneously, by obtaining comprehensive target environment data based on at least one target device, the target user's action planning needs are linked with their real-time scene conditions, enabling the generation of device control parameters to adapt to the current environmental conditions, avoiding rigid planning detached from the actual scene, and significantly improving the scene rationality and practicality of action planning. Finally, the current request, target user profile data, and target comprehensive environment data are input into the trained action planning model, which outputs target control parameters for at least one target device for the target user. By combining multi-source information such as the current request, target user profile, and target comprehensive environment for action planning, the limitations of traditional rule-based control parameter matching or action planning based on a single information source are replaced. This improves both the accuracy and automation efficiency of action planning, as well as the level of automated control of the equipment, providing users with a more personalized and intelligent service experience.

[0010] The above description is merely an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and in order to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 A schematic diagram illustrating an application scenario of the device motion planning method provided in an exemplary embodiment of this application; Figure 2 A flowchart illustrating a device motion planning method provided for an exemplary embodiment of this application; Figure 3 A schematic diagram illustrating the process of constructing target user profile data, provided as an exemplary embodiment of this application; Figure 4 A schematic diagram illustrating another process for constructing target user profile data, provided as an exemplary embodiment of this application; Figure 5 A flowchart illustrating another device motion planning method provided as an exemplary embodiment of this application; Figure 6 A schematic diagram of a device memory data management process provided for an exemplary embodiment of this application; Figure 7 A flowchart illustrating another device motion planning method provided as an exemplary embodiment of this application; Figure 8 A schematic diagram of the structure of a device motion planning apparatus provided for an exemplary embodiment of this application; Figure 9 This is a schematic diagram of the structure of an electronic device provided as an exemplary embodiment of this application. Detailed Implementation

[0013] To make the features and advantages of this application more apparent and understandable, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0014] The terms "first," "second," "third," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or apparatuses.

[0015] Please refer to the following. Figure 1 , Figure 1 This is a schematic diagram illustrating an application scenario of the device motion planning method provided in an exemplary embodiment of this application. For example... Figure 1As shown, the application scenarios include, but are not limited to, smart home scenarios, smart office scenarios, smart campus scenarios, and smart healthcare scenarios. Each application scenario typically includes at least one artificial intelligence device 101 and multiple associated devices 102. The associated devices 102 are connected to the artificial intelligence device 101 via a network 103. The artificial intelligence device 101 can be a terminal device (such as a main control device in a home environment) or a server device, including but not limited to hardware servers, virtual servers, and cloud servers. The associated devices 102 may be partially the same or completely different in different application scenarios, and their configuration can be tailored to the functional positioning of the scenario and user needs. The network 103 can be a medium providing communication links or the Internet, which includes network devices and transmission media, and is not limited to these. The transmission media can be wired links, such as, but not limited to, coaxial cables, fiber optic cables, and digital subscriber lines (DSL), or wireless links, such as, but not limited to, wireless internet access (Wi-Fi), Bluetooth, and mobile device networks.

[0016] Taking a smart home scenario as an example, each target device in the smart home scenario can collect the current requests (such as the current interactive voice request) of family members (target users) in real time and send them to the artificial intelligence device. Upon receiving the current request, the artificial intelligence device can obtain target user profile data corresponding to the target user. This target user profile data is based on at least one target memory tag corresponding to the target user, which represents the target user's relevant health status and / or device usage preferences. It can also obtain comprehensive target environmental data based on at least one target device. Finally, the current request, the target user profile data, and the comprehensive target environmental data are input into an action planning model, which outputs target control parameters for at least one target device. The action planning model is trained based on sample training data of multiple known action planning results. For example, if each target device in the smart home scenario receives the voice message "I'm a little hot, please turn on the air conditioner" from family member A (target user), the artificial intelligence device can combine the target user profile data with the voice and comprehensive target environmental information uploaded by each target device to determine the air conditioner temperature. Alternatively, it can further control other devices (such as a humidifier) ​​to maintain the indoor environment within a comfortable range to meet the user's actual health needs.

[0017] Understandably, Figure 1The number and type of AI devices 101 and associated devices 102 in the application scenario shown are only examples. In specific implementations, any number of AI devices 101 and associated devices 102 can be included, and any type of AI devices 101 and associated devices 102 can be used. This specification does not specifically limit this.

[0018] Next, combine Figure 1 This document describes the device motion planning method provided in the embodiments of this application. Please refer to the following for details. Figure 2 This is a flowchart illustrating a device motion planning method provided in an exemplary embodiment of this application. The executing entity of this device motion planning method is an electronic device providing artificial intelligence services (AI device). This electronic device can be a server-side device or a terminal device; this embodiment of the application does not limit this. Figure 2 As shown, the device motion planning method may include, but is not limited to, the following steps: S201, Receive the current request input by the target user.

[0019] Specifically, the target user can input the current request through various methods such as voice, text, or touch operation. When at least one target device associated with the target user (such as, but not limited to, associated devices in the same environment as the target user or associated devices that the target user can control through commands or operations) receives the current request input by the target user, each target device will send the current request to the artificial intelligence device. The artificial intelligence device can then plan corresponding actions based on the current request received by each target device from the target user.

[0020] Understandably, the aforementioned "same environment" refers to the same physical space, including but not limited to home environments, office environments, campus environments, and medical environments. Devices in different environments may be partially the same or completely different, and their configuration needs to be tailored to the functional positioning of the scenario and user needs. For example, a home environment may include, but is not limited to, wearable health monitoring devices (smart bracelets, sleep monitoring mattresses, heart rate monitors, etc.), smart interactive terminals (smart speakers, central control touchscreens, voice remote controls, etc.), environmental control devices (air conditioners, humidifiers, air purifiers, fresh air systems, smart curtains, etc.), and entertainment devices (smart TVs, projectors, home theater systems, etc.). An office environment may include, but is not limited to, environmental control devices (air conditioners, fresh air systems, air purifiers, etc.), meeting collaboration devices (projectors, smart meeting tablets, microphones, cameras, etc.), smart interactive terminals (smart speakers, voice assistants, etc.), and office automation devices (printers, scanners, shredders, etc.).

[0021] Understandably, the aforementioned artificial intelligence device can also be an intelligent target device capable of providing artificial intelligence services among at least one target device related to the target user. That is, if the target device has sufficient computing power, the pre-trained action planning model in this application embodiment can be directly deployed on the target device to realize action planning on the target user's local device. This eliminates the need for the target user to add additional artificial intelligence devices for action planning on each target device, reducing the hardware cost of action planning control for one or more devices, and also protects the user's device privacy information to a certain extent, improving the security of user data during the action planning process.

[0022] S202, Obtain target user profile data corresponding to the target user; the target user profile data is composed of at least one target memory tag corresponding to the target user; the target memory tag represents the target user's relevant health status and / or device usage preferences.

[0023] Specifically, target user profile data is used to characterize the personalized features of target users, such as, but not limited to, personalized behavioral features and / or personalized health needs in specific scenarios, and is dynamically constructed from multi-dimensional data accumulated over a long period. This target user profile data is structured data with content arranged according to certain format rules. Target memory tags may, but are not limited to, data structures constructed from personalized feature information related to target users (such as, but not limited to, health status features, device usage preference features, etc.) according to a preset structure. This data structure may, but is not limited to, consist of features extracted and structured from basic attribute information, device usage preference information, and / or health status information. Basic attribute information includes static information such as tag identity, tag creation time, and tag validity period; device usage preference information records the user's device operation habits, frequently used settings, etc., which helps to understand the user's personalized needs for device operation. For example, a user may prefer to set the air conditioner temperature to a specific temperature, or prefer to use a printer with a specific printing speed and quality. Through structured device usage preference information, the action planning model can more accurately and efficiently plan device actions for the user, providing a personalized intelligent control experience. Health status information records the user's health status, such as but not limited to physical indicators such as medical history, allergy history, sleep quality, and fatigue level. Through structured health status information, it helps the action planning model analyze and understand the user's special needs for environmental equipment, such as whether an air purifier is needed to improve air quality to suit allergies, or whether the air conditioner temperature needs to be adjusted to suit a specific physical condition.

[0024] Optionally, the aforementioned target user profile data may be based on, but is not limited to, at least one of the following target memory tags corresponding to the target user: target individual memory tags and target group memory tags. The aforementioned target individual memory tags represent the target user's individual health status and device usage preferences. The aforementioned target group memory tags represent the device usage preferences of the target group to which the target user belongs.

[0025] In this embodiment, by introducing target individual memory tags and / or target group memory tags to construct target user profile data, it is possible to more comprehensively and accurately depict the personalized characteristics of users. Target individual memory tags focus on individual users, deeply recording their unique health status and device usage preferences, enabling the motion planning model to tailor device actions for users based on this detailed information, meeting their personalized device control needs. Target group memory tags, on the other hand, focus on the target group to which the user belongs, extracting the shared device usage preferences of that target group. This helps to quickly plan device actions that meet the general needs of the corresponding target group when facing target users belonging to that target group, thereby also meeting the personalized device control needs of target users belonging to that target group to a certain extent, improving motion planning efficiency and personalization. At the same time, the target user's personal memory tags and target group memory tags can be used to build a more comprehensive target user profile. This allows the motion planning model to integrate the relevant personalized features of the target user's personal and group dimensions to achieve more accurate and personalized motion planning, significantly improving the accuracy and adaptability of device motion planning and bringing users a better and more considerate intelligent control experience.

[0026] Optionally, the aforementioned target user profile data may be, but is not limited to, based on target group memory tags and the target relationship group member memory tags corresponding to the target user's target relationship group. The aforementioned target relationship group member memory tags consist of the target user's corresponding target personal memory tag and the personal memory tags corresponding to other members of the target relationship group besides the target user.

[0027] In this embodiment, by integrating the personal memory tags of each member in the target user's target relationship group, the system can capture more detailed information such as the health status and device usage preferences of all members in the target relationship group. This allows the action planning model to not only consider the target user's personalized needs but also the collaborative device usage preferences between the target user and other members of the target relationship group, such as family members or company members, and the health status of each member in the same environment as the target user. For example, when adjusting the air conditioner temperature, the action planning model can automatically balance the device usage preferences and health status of each family member to the most suitable indoor environmental parameters. Furthermore, through multi-level memory tags, such as target group memory tags, the system not only enriches the dimensions of the target user profile data but also ensures that the action planning model achieves a dynamic balance between individual needs and group collaboration, enhancing its adaptability to complex social scenarios.

[0028] Furthermore, the aforementioned target group memory tags may include, but are not limited to, at least one of the following: target relationship group memory tags and target association group memory tags. The aforementioned target relationship group memory tags represent the target user's target relationship group's device usage preferences for each target device and the shared preferences of the target relationship group. The aforementioned target association group memory tags represent the target association group's device usage preferences for similar devices corresponding to the target device and the corresponding health status of the target association group. The aforementioned target relationship group is a group composed of members in the same environment as the target user, such as a family in a smart home environment or a company in a smart office environment; that is, for example, but not limited to, it may include the target user's family relationship group, organizational relationship group (such as people in the same organization as the target user), etc. The aforementioned target association group may include, but is not limited to, people belonging to the same category as the target user (e.g., the same age group, the same region, etc.).

[0029] Understandably, the aforementioned group memory tags and personal memory tags can be structured data (i.e., data content arranged according to certain format rules). Of course, in other embodiments, group memory tags and personal memory tags can also be unstructured data (i.e., data content arranged without rules), and this is not limited here. Device usage preferences include a user's (group or member's) usage preferences for a single device, shared usage preferences for multiple devices, and / or linked usage preferences for multiple devices (for example, a user often turns on a humidifier at the same time as turning on the air conditioner in winter). Typically, different types of devices correspond to different types of usage preferences; for example, the usage preferences for air conditioners include temperature preferences, and the usage preferences for smart speakers include volume preferences.

[0030] In some possible embodiments, such as Figure 3As shown, the implementation process of obtaining the target user profile data corresponding to the target user in S202 above may include, but is not limited to: S301, construct target personal memory tags based on target user memory data on various target devices.

[0031] Specifically, the aforementioned target user memory data may include, but is not limited to, target device usage information and target user personal information for each target device. The data in the target personal memory tag mainly includes the target user's health status and their usage preferences for each target device (device usage preferences), and may also include the tag creation time and validity period. The tag creation time is the creation time of the device usage preference and health status, and the validity period is the validity period of the corresponding device usage preference or health status. The target personal memory tag is associated with and bound to the target user identifier (target user ID). The health status can be determined based on the target user's personal information in each target user memory data, reflecting whether the target user is sick, if so, what illness they have, and further reflecting the severity of the illness. For example, if the target user memory data records the personal information of target user A, "Target user A didn't cover themselves properly last night and has a slight fever today," then the analysis can determine that target user A's health status is "mild cold and fever." The target user's usage preference for a single target device can be determined based on the target user's memory data regarding the target device's usage information. For example, if the target device is a bedroom air conditioner, and the target user's memory data records the target user A's specific adjustment information for the bedroom air conditioner multiple times within a month, then the target user A's device usage preference for the bedroom air conditioner (e.g., but not limited to temperature preference) can be summarized based on this adjustment information.

[0032] For example, taking a smart home environment as an example, the aforementioned target personal memory tag can be, but is not limited to, structured data in the following format: {Member: The ID of the specific family member (i.e., the target user).}

[0033] Name: The member's name.

[0034] Age: The age of the member.

[0035] Role: The role of a member (e.g., adult male, adult female, child, elderly).

[0036] Device usage preferences: Tag name: For example, a specific temperature preference tag.

[0037] Creation time: The time the tag was created.

[0038] Validity period: The validity period of the label, in days. If it is valid indefinitely, the value is "permanent".

[0039] Health status: Tag name: Specific health status tag.

[0040] Creation time: The time the tag was created.

[0041] Validity period: The validity period of the label, in days. If it is valid indefinitely, the value is "permanent".

[0042] } S302, construct target group memory tags based on the device memory data and / or group knowledge base corresponding to each target device.

[0043] Optionally, by mining and organizing the device's memory data, common characteristics and individual differences exhibited by different users when using the same target device can be identified. Common characteristics reflect general patterns in device usage, such as most users tending to use the device within specific time periods or having a high frequency of use for certain functions. Individual differences, on the other hand, reveal the unique preferences and needs of different users; for example, some users may have more precise requirements for the device's temperature settings.

[0044] Alternatively, a group knowledge base can be used as a supplementary information source. The aforementioned group knowledge base includes device usage preference data for different types of groups. Through the group knowledge base, target group memory tags that are both universal and personalized can be constructed.

[0045] Optionally, the aforementioned target group memory tags include at least one of the following: target relationship group memory tags and target association group memory tags. The implementation process of constructing target group memory tags based on device memory data and / or group knowledge base corresponding to each target device in S302 may include, but is not limited to: constructing target relationship group memory tags based on device memory data corresponding to each target device. The device memory data includes interaction data (i.e., device usage data) corresponding to each member in the target relationship group where the target user resides on each target device. The data in the target relationship group memory tags mainly includes device information of all devices in the entire environment where the target user is currently located, as well as the device usage preferences of the target relationship group. It may also include tag creation time and validity period. The tag creation time is the creation time of the device usage preference, and the validity period is the validity period of the corresponding device usage preference. The device information includes device identifiers (device IDs), device names (device types), device installation locations, etc., and the device usage preferences reflect the usage preferences of the entire group for a single device or multiple devices. For example, in a home environment, the family as a whole tends to set the living room air conditioner to a "moderate" temperature or the bedroom air conditioner to a "slightly cool" temperature.

[0046] For example, taking a smart home environment as an example, the aforementioned target relationship group memory tags (family memory tags) can be, but are not limited to, structured data in the following format: {Family: Specific family (i.e., target relationship group) ID} 1) Equipment: Device 1: The specific device ID.

[0047] Equipment Name: The name of the equipment.

[0048] Room: The room where the equipment is located.

[0049] Device usage preferences: Tag name: For example, a specific temperature preference tag.

[0050] Creation time: Tag creation time Validity period: The validity period of the label, in days. If it is valid indefinitely, the value is "permanent".

[0051] Device 2: The specific device ID.

[0052] Equipment Name: The name of the equipment.

[0053] Room: The room where the equipment is located.

[0054] Device usage preferences: Tag name: For example, a specific temperature preference tag.

[0055] Creation time: Tag creation time Validity period: The validity period of the label, in days. If it is valid indefinitely, the value is "permanent".

[0056] 2) Shared preferences: Device usage preferences: Tag name: For example, a specific temperature preference tag.

[0057] Creation time: The time the tag was created.

[0058] Validity period: The validity period of the label, in days. If it is valid indefinitely, the value is "permanent".

[0059] } and / or Optionally, the aforementioned target group memory tags include at least one of the following: target relationship group memory tags and target association group memory tags. The implementation process of constructing target group memory tags based on device memory data and / or group knowledge bases corresponding to each target device in S302 may include, but is not limited to: determining target association group knowledge of the target-related group corresponding to the target user from the group knowledge base based on the target basic information corresponding to the target user, and constructing target association group memory tags based on the target association group knowledge. The aforementioned group knowledge base includes device usage preference data and health status of different types of groups.

[0060] First, a detailed analysis of the target user's basic information is conducted. This information may include, but is not limited to, age, gender, occupation, and lifestyle habits. Based on this information, precise matching and filtering are performed in a group knowledge base to identify the target user's associated groups. Next, relevant knowledge such as device usage preferences and health status of this associated group is extracted from the group knowledge base. This knowledge forms the basis for constructing target-associated group memory tags. Finally, specific algorithms and models are used to integrate and analyze this extracted target-associated group knowledge, constructing representative and targeted target-associated group memory tags. This allows for more accurate device action planning services tailored to the target user's associated group characteristics.

[0061] For example, taking a smart home environment as an example, the aforementioned target-related group memory tags can be, but are not limited to, structured data in the following format: {Target group name (e.g., but not limited to children, elderly):} Device usage preferences: Label name: For example, specific temperature preference label (e.g., cool, moderate).

[0062] Creation time: The time the tag was created.

[0063] Validity period: The validity period of the label, in days. If it is valid indefinitely, the value is "permanent".

[0064] Health status: Tag name: Specific health status tag.

[0065] Creation time: The time the tag was created.

[0066] Validity period: The validity period of the label, in days. If it is valid indefinitely, the value is "permanent".

[0067] } In this embodiment, by integrating and analyzing the target user memory data on each target device and the device memory data corresponding to each target device, the tagging and contextual association of memory data can be achieved. This not only extracts the common device usage patterns of the group (such as the overall air conditioning temperature preference of the family), but also accurately binds the health status and personalized device usage needs of individual users (members) (such as the air purifier operation preference of allergic members). This transforms memory data from scattered information into effective basis that can be directly called by the action planning model, providing standardized core features for the subsequent construction of multi-level target user profile data.

[0068] S303, construct target user profile data corresponding to the target user based on the target individual memory tags and / or target group memory tags.

[0069] Specifically, when the target individual memory tags and target group memory tags are structured data, they can be directly identified as single-level target user profile data. Alternatively, the target individual memory tags and target group memory tags can be assembled and integrated to obtain multi-level target user profile data. When the target individual memory tags and target group memory tags are unstructured data, key features need to be extracted using natural language processing techniques and transformed into a structured representation before constructing single-level or multi-level target user profile data using the aforementioned methods.

[0070] Optionally, when the target user profile data has a multi-level structure, the target user profile data and the target group identifier (such as a family ID) can be associated and bound. By associating the structured user profile data with the target group identifier, it is convenient to dynamically update and maintain multi-level user profiles, enabling collaborative perception and intelligent decision-making across devices and scenarios. For example, in a home environment, based on children's preference for cooler temperatures and the allergies of specific members, the system can automatically adjust the temperature and humidity of the air conditioner and the operating mode of the air purifier, improving service accuracy and user experience.

[0071] In some possible embodiments, the aforementioned current request is a voice request. For example... Figure 4 As shown, the implementation process of obtaining the target user profile data corresponding to the target user in S202 above may also include, but is not limited to: S401, perform information identification on the current request to obtain the target basic information corresponding to the target user.

[0072] Specifically, the aforementioned information recognition may include, but is not limited to, recognition of basic user information such as identity information, gender, and age. For example, but not limited to, recognizing the target user's identity through voiceprint algorithms, and identifying whether the user is a child (male / female), an adult (male / female), or an elderly person (male / female) through voice gender and age recognition algorithms.

[0073] Optionally, in step S201 above, after receiving the current request input by the target user, the action planning method may further include, but is not limited to, inputting the current request into a disease state detection model and outputting a disease state detection result corresponding to the target user based on the voice state of the current request. For example, but not limited to, the disease state detection model may determine whether the target user has symptoms of a cold, cough, or respiratory illness based on the voice state in the current voice request, such as by analyzing features like the degree of voice tremor, speech rate changes, and pronunciation clarity. And based on the disease state detection result, the target user's corresponding target personal memory tag in the memory bank is updated, that is, the target user's health state tag in the target personal memory tag is updated. In other words, the health state tag in the target personal memory tag may be dynamically updated and determined based on, but is not limited to, the voice features of the target user's input voice request to identify possible illness states such as a cold or cough and / or health abnormality feedback information manually entered or voice input by the target user.

[0074] Please continue to refer to the following. Figure 4 ,like Figure 4 As shown, in step S401 above, after identifying the information of the current request and obtaining the basic target information corresponding to the target user, the action planning method may also include, but is not limited to, the following: S402, based on the target's basic information, query the memory bank for at least one target memory tag corresponding to the target user; the memory bank includes individual memory tags corresponding to multiple users and / or group memory tags corresponding to multiple groups.

[0075] Specifically, after obtaining the target user's target disease information, the AI ​​device first performs a precise search in the memory bank based on basic target information, such as the target user's identity identifier. The memory bank, as a vast data storage center, can contain, but is not limited to, unique personal memory tags for each user, and can also, but is not limited to, group memory tags corresponding to various groups. Through the above search mechanism, at least one target memory tag matching the target user can be quickly located, providing strong data support for subsequent action planning.

[0076] Optionally, but not limited to, the above-mentioned step of querying at least one target memory tag corresponding to the target user in the memory bank based on the target basic information may be performed only when the current request is identified as a device control request. This avoids the problem of invalid querying memory tag data when the current request does not require a memory tag at all, such as when the current request is a chat request from the target user asking for the current time.

[0077] Optionally, the AI ​​device can also, but is not limited to, update the relevant group memory tags in the memory bank at target time intervals based on device memory data from each device connected to it within the target time interval (e.g., but not limited to a month, a year, a week, or a day). This ensures that the group memory tags remain synchronized with the actual usage of the devices and reflect the latest group user behavior. Through regular updates, the AI ​​device can more accurately understand the preferences and needs of the user group, thereby providing users with more personalized services. At the same time, this updating method also improves the flexibility and adaptability of the memory bank, enabling it to respond to the ever-changing needs of the group users.

[0078] S403, construct target user profile data corresponding to the target user based on at least one target memory tag.

[0079] Specifically, the process of constructing the target user profile data in S403 is similar to the process of constructing the target user profile data in S303, and will not be repeated here.

[0080] Please continue to refer to the following. Figure 2 ,like Figure 2 As shown, in step S201 above, after receiving the current request input by the target user, the action planning method may also include, but is not limited to, the following: S203, acquire comprehensive environmental data of the target based on at least one target device.

[0081] Specifically, the comprehensive environmental data for the target may include, but is not limited to, the current device status and current environmental perception data corresponding to each target device. The current environmental perception data may be obtained in real time by sensors deployed in the target devices, collecting data on the current environment, and may include, but is not limited to, multi-dimensional information such as light intensity, temperature and humidity, sound field distribution, and personnel activity density. The current device status may reflect, but is not limited to, the current operating mode, current power level, and current network connection status of each target device.

[0082] Optionally, in order to facilitate a more accurate and efficient understanding by the subsequent action planning model, the aforementioned target integrated environment data can be data that is structured from the current device status and current environmental perception data corresponding to each target device. This structured format is consistent with the structured format of the aforementioned memory labels, which facilitates unified processing and feature fusion by the action planning model and improves inference efficiency.

[0083] S204, input the current request, target user profile data and target comprehensive environment data into the action planning model, and output the target control parameters for at least one target device; the action planning model is trained based on sample training data of multiple known action planning results.

[0084] Specifically, the action planning results include device control parameters corresponding to the sample training data. The action planning model can collaboratively infer from the current request, target user profile data, and comprehensive target environment data to output target control parameters adapted for at least one target device. This action planning model supports dynamic decision optimization across devices and scenarios, ensuring that control parameters meet user preferences and real-time environmental requirements. For example, when a user requests "creating a suitable reading environment," the action planning model can comprehensively assess factors such as insufficient light intensity, user nearsightedness, high ambient noise, and sufficient battery power and network connectivity of the desk lamp on the target device. It can then automatically adjust the desk lamp to eye-protection brightness mode, activate white noise compensation in conjunction with the noise reduction speaker, and optimize the e-book device's screen refresh rate to a paper-like display.

[0085] Optionally, the number of target devices in the at least one target device for action planning in the above action planning model can be the same as or different from the number of target devices in the at least one target device for uploading the current request to the artificial intelligence device. When the numbers are different, the action planning model can dynamically add or remove target devices participating in control according to the actual scenario requirements to ensure that the decision covers the most suitable combination of devices.

[0086] Optionally, the aforementioned sample training data is generated by the target large model based on sample seed data carrying generation state annotations and data rationality annotations. The generation state annotations represent the types of control parameters generated based on the sample seed data for action planning; the data rationality annotations represent the rationality of the planned control parameters in the sample seed data; the sample seed data includes sample interaction requests corresponding to sample users with known action planning results, sample user profile data, and sample comprehensive environment data. When generating sample training data, the target large model will accurately determine the types of control parameters to be generated based on the generation state annotations in the sample seed data, such as brightness adjustment parameters, volume adjustment parameters, or screen refresh rate adjustment parameters. At the same time, based on the data rationality annotations, the generated control parameters are evaluated and screened for rationality to ensure that the control parameters in the generated sample training data not only meet the needs of the actual scenario but also effectively improve the performance of the action planning model. This carefully annotated and screened sample training data will be used to train the action planning model, enabling it to more accurately understand user intentions and plan device actions more intelligently, thereby providing users with a more personalized and comfortable device control experience.

[0087] and / or The aforementioned action planning model is obtained by preference alignment optimization based on preference optimization training data. This preference optimization training data includes sample requests with multiple known preference planning parameter pairs. Each preference planning parameter pair includes positive and negative preference parameters planned for the sample requests. Positive preference parameters are those that align with the user's device usage preferences and health status, i.e., reasonable parameters; negative preference parameters are those that do not align with the user's device usage preferences or health status, i.e., unreasonable parameters. In this embodiment, using preference optimization training data to perform preference alignment optimization on the action planning model allows the model to better understand the user's preference tendencies in different scenarios. This, in turn, enables the model to better meet the user's personalized needs when planning device actions, thereby improving user satisfaction with device control.

[0088] Optionally, in step S204 above, the process of inputting the current request, target user profile data, and target comprehensive environment data into the action planning model and outputting the implementation process of target control parameters for at least one target device may include, but is not limited to: inputting the current request, target user profile data, and target comprehensive environment data into the action planning model and outputting a time series of target control parameters for at least one target device. The aforementioned time series of target control parameters may include, but is not limited to, target control parameters corresponding to multiple time points arranged in chronological order. These multiple time points include the current time and at least one future time. That is, the action planning model can also be a multi-step action planning model, capable of ensuring that the target device performs actions at multiple subsequent time points in a manner most consistent with user preferences, health status, and environmental status.

[0089] In this embodiment, firstly, by receiving the current request input by the target user, the system directly captures the user's immediate device control needs, ensuring that subsequent action planning processes start from the user's current real needs. This avoids misunderstandings of needs due to information transmission delays or deviations, providing a clear and accurate initial direction for the entire intelligent action planning process. Secondly, the system acquires target user profile data based on at least one target memory tag corresponding to the target user. Using the target memory tags representing the target user's health status and / or device usage preferences, a personalized profile is constructed. This shifts the basis of the action planning model when planning device actions from generalized assumptions to the user's real personalized characteristics and needs, effectively avoiding the incompatibility of general solutions with different user health needs or usage habits. This ensures that the subsequently output target control parameters accurately match the target user's personalized needs. Simultaneously, by acquiring comprehensive target environment data based on at least one target device, the system links the target user's action planning needs with their real-time scene conditions. This allows the generation of device control parameters to adapt to the current environmental conditions, avoiding rigid planning detached from the actual scenario and significantly improving the scenario rationality and practicality of action planning. Finally, the current request, target user profile data, and comprehensive target environment data are input into the trained action planning model, which outputs target control parameters for at least one target device for the target user. By combining action planning models with multi-source information such as "current request, target user profile, and target comprehensive environment" for action planning, the limitations of traditional rule-based control parameter matching or action planning based on a single information source are replaced. This not only improves the accuracy and automation efficiency of action planning, but also enhances the level of automated control of equipment, providing users with a more personalized and intelligent service experience.

[0090] Please refer to the following. Figure 5 This is a flowchart illustrating another device motion planning method provided in an exemplary embodiment of this application. Figure 5 As shown, the device motion planning method may include, but is not limited to, the following steps: S501, Receive the current request input by the target user.

[0091] Specifically, S501 is the same as S201, and will not be repeated here.

[0092] S502, obtain target user profile data corresponding to the target user; the target user profile data is composed of at least one target memory tag corresponding to the target user; the target memory tag represents the target user's relevant health status and / or device usage preferences.

[0093] Specifically, S502 is the same as S202, and will not be repeated here.

[0094] S503, acquire comprehensive environmental data of the target based on at least one target device.

[0095] Specifically, S503 is the same as S203, and will not be repeated here.

[0096] S504, obtain the historical interaction data corresponding to the target user.

[0097] Specifically, historical interaction data includes records of historical interaction requests and dialogues between the target user and AI devices or various target devices within a historical time period (e.g., but not limited to the past week, the past month, etc.). This includes, but is not limited to, the dialogue content and timing of historical interaction requests stored according to a preset data structure. Analyzing recent historical interaction data of target users can identify dynamic changes in their behavioral patterns and potential needs, overcoming the timeliness limitations of static profile data. By combining the implicit intentions in current requests and historical interactions, and capturing the evolution of user preferences through time-series modeling, target control parameters can not only be more adapted to the immediate scenario but also predict future user action needs, achieving a leap from passive response to proactive prediction.

[0098] S505 inputs the current request, target user profile data, target comprehensive environment data, and historical interaction data into the action planning model, and outputs target control parameters for at least one target device.

[0099] Specifically, the aforementioned action planning model is trained based on current sample requests, sample user profile data, sample comprehensive environment data, and sample historical interaction data with known device control parameters in various scenarios. After inputting the current request, target user profile data, target comprehensive environment data, and historical interaction data, the action planning model comprehensively considers the specific content of the current request, the user characteristics reflected in the target user profile data, the scenario conditions presented by the target comprehensive environment data, and the user behavior patterns, preferences, or historical interaction intentions reflected in the historical interaction data. This allows it to accurately generate target control parameters for at least one target device. These target control parameters ensure that the target device operates in the most appropriate way in the current scenario, meeting the user's current needs while also considering their long-term usage habits and potential intentions.

[0100] In this embodiment, by jointly inputting historical interaction data, current requests, target user profiles, and comprehensive target environmental data into the action planning model, the model's understanding of user intent and device usage needs is enhanced. This improves the coherence and accuracy of control parameter generation, making device responses more aligned with actual user expectations and further optimizing the naturalness, intelligence, and smoothness of human-machine collaboration. Simultaneously, by inputting historical interaction data and current requests, the action planning model, based on multi-dimensional data fusion, strengthens its ability to capture the temporal characteristics of user interaction behavior. This enables control decisions to not only respond to current commands but also anticipate future interaction trends, continuously improving the subtlety and seamlessness of the user experience.

[0101] In some possible embodiments, an exemplary embodiment of this application also provides a method for managing device memory data related to device motion planning. For example... Figure 6 As shown, the process for managing the memory data of this device may include, but is not limited to, the following steps: S601 receives device memory data uploaded by each target device.

[0102] Specifically, the aforementioned device memory data is data obtained by processing the original interaction data uploaded by the target devices, and can be considered as the data stored in the memory of each target device itself. Taking a smart home environment as an example, each home device can collect the original interaction data between each family member and the device in real time, and synchronize it to the artificial intelligence device through IoT communication links (such as Bluetooth, WiFi, ZigBee). This original interaction data can be text data entered by the user through the control panel, voice data collected through devices such as microphones, or other forms, without limitation here. Then, the artificial intelligence device extracts relevant information about the device and the user from the original interaction data and adds it to the device memory data associated with that device.

[0103] The device memory data of the aforementioned target device may include all user information regarding the device's usage and / or at least one user's personal information. Device usage information relates to device operating parameters, such as, but not limited to, air conditioner temperature, smart speaker volume, and humidifier humidity. Personal information includes personal physiological information and / or basic personal information. Personal physiological information primarily refers to health-related bodily information, including but not limited to heart rate, blood pressure, sleep quality, and medical history. Basic personal information includes, but is not limited to, gender, age, name, and user role (e.g., parent).

[0104] It's important to note that each device's memory data is typically stored in conjunction with its device identifier to ensure traceability and ensure ownership, providing a reliable data source for subsequent data analysis and retrieval. The storage area can be located on server-side devices or on terminal devices with sufficient storage space. The device memory data is updated in real-time. For example, if a user adjusts a device's settings (e.g., turns on the air conditioner and sets the temperature to 24°C), the device usage information in the corresponding memory data needs to be updated accordingly. The storage area can store only the device memory data for a fixed period (e.g., one month) for each device, or it can allocate a fixed-capacity storage area for each device, automatically deleting the oldest data stored in the area once the capacity is exceeded.

[0105] S602 distributes the device memory data of different target devices in different device structures.

[0106] Specifically, after receiving device memory data uploaded by various target devices, the AI ​​device will distribute and store the device memory data corresponding to each target device in different device structures. For example, for air conditioning devices, lighting devices, and security devices in a smart home system, corresponding air conditioning device structures, lighting device structures, and security device structures are created respectively. Each structure records detailed device memory data such as device identifier, device usage information, device control records, and device location information. The aforementioned device structures are a data structure organization form corresponding to device memory data, ensuring the classified storage and efficient retrieval of device memory data.

[0107] Through this structured storage method, AI devices can quickly locate and retrieve historical memory data for specific devices, supporting cross-device contextual linkage and collaborative control. For example, when a user enters the bedroom at night, the AI ​​device can automatically adjust the light brightness based on usage habits recorded in the lighting device's structure, simultaneously activate the cooling mode based on temperature preference data from the air conditioning device's structure, and retrieve information from the security device's structure to confirm that the door is locked, achieving multi-device collaborative response. This memory data management mechanism based on device structures not only improves data access efficiency but also supports dynamic updates and access control, ensuring the independence of data from each device and the overall stability of the system, providing personalized intelligent services while protecting user privacy.

[0108] Meanwhile, this application embodiment also ensures the independence and integrity of the original memory data of each device through distributed storage. By distributing the device memory data of different target devices with different device structures, the confusion of data from multiple devices is avoided, which facilitates the subsequent tracing, maintenance or analysis of the historical memory of a single device and ensures the accuracy of device-level data.

[0109] S603 integrates the user memory data for the same user from the device memory data corresponding to each target device and stores it in the user structure corresponding to the user.

[0110] Specifically, after receiving device memory data uploaded by various target devices, the AI ​​device will further filter and aggregate user memory data for the same user across all target devices based on the user's unique identifier. This user memory data covers multi-dimensional information such as the user's operating habits, preference settings, and usage frequency on different target devices. For example, it includes the user's preferred air conditioning temperature settings, lighting brightness adjustment records, and operating modes of security devices. By integrating and storing the user memory data scattered across various devices into a single user structure corresponding to that user, a complete user-centric data set is formed. This integration method not only facilitates the AI ​​device's comprehensive understanding of user behavior patterns but also provides a precise and complete data foundation from a global perspective for subsequent personalized action planning services, meeting the demand for comprehensive data in multi-device collaborative scenarios.

[0111] In this embodiment, distributed storage disperses the data management load, ensuring the independence of data from each device and the overall stability of the system. At the same time, by centrally integrating and storing the user memory data of the same user in each target device, it breaks down the silos of memory data on a single device, realizes the effective integration of multiple memory data, avoids the data management chaos caused by memory data redundancy and inconsistent formats, simplifies the query and retrieval process of cross-device data, improves data management efficiency, and adapts to the application scenario requirements of "single device or multi-device collaboration" in the subsequent action planning process.

[0112] In some possible embodiments, the above-described process of obtaining target user profile data corresponding to the target user may also include, but is not limited to: determining target personal memory tags based on the target user structure corresponding to the target user, and determining target group memory tags based on the device structure corresponding to each target device; finally, constructing target user profile data corresponding to the target user based on the target personal memory tags and / or target group memory tags.

[0113] Upon receiving a current request from a target user uploaded by various target devices, the AI ​​device can, but is not limited to, query and retrieve the corresponding target user structure from the memory database based on the target user's unique identifier (e.g., identity identifier), and query and retrieve the corresponding device structure from the memory database based on the device identifier of each target device. Subsequently, key user memory features are extracted from the target user structure and analyzed and mapped according to a preset structural form. For example, device usage preference features (e.g., air conditioning temperature preference) and health status features (e.g., allergy history) are extracted from the target user structure, structured, and directly mapped as components of the target individual's memory tag. Simultaneously, statistical analysis is performed on the device structures of the target devices to identify the target group's usage preferences and common patterns for each target device, and directly mapped as corresponding target group memory tags. Finally, by combining the target individual memory tags and / or target group memory tags, target user profile data encompassing individual characteristics and / or group commonalities is generated.

[0114] Optionally, the aforementioned target personal memory tags can be adjusted in real time as the target user structure data changes. For example, when the target user's device usage preferences or health status in the target user structure data are updated, the target personal memory tags can also be automatically updated synchronously, providing a dynamic basis for subsequent user profile data iterations, maintaining the timeliness and accuracy of the data, and ensuring that the user profile data always reflects the latest behavioral characteristics and status needs. At the same time, the target group memory tags can also be periodically optimized based on the continuous aggregation and analysis of multi-device data, strengthening the ability to identify group patterns and improving the accuracy and personalization of service recommendations in cross-device collaborative scenarios.

[0115] Optionally, the aforementioned target individual memory tags and target group memory tags can be extracted using, but are not limited to, a pre-trained feature tag extraction model. The aforementioned feature tag extraction model can be trained, but is not limited to, on a training dataset of known memory tags. The aforementioned training dataset of these structures is generated by a large-parameter model based on seed structure data (seed user structure data, seed device structure data) of known memory tags (individual memory tags, group memory tags).

[0116] In this embodiment, by using structured data query and dynamic tag feature extraction mapping, the accuracy of memory tags is determined, and the transformation from fragmented data to structured profiles is realized. This enables efficient generation and dynamic updating of user profile data, providing reliable and personalized data support that the action planning model can understand for AI devices to perform action planning.

[0117] Please refer to the following. Figure 7 This is a flowchart illustrating another device motion planning method provided in an exemplary embodiment of this application. Figure 7 As shown, the device motion planning method may include, but is not limited to, the following steps: S701 receives the current request input by the target user.

[0118] Specifically, S701 is the same as S201, and will not be repeated here.

[0119] S702, Obtain target user profile data corresponding to the target user; the target user profile data is composed of at least one target memory tag corresponding to the target user; the target memory tag represents the target user's relevant health status and / or device usage preferences.

[0120] Specifically, S702 is the same as S202, and will not be repeated here.

[0121] S703, acquire comprehensive environmental data of the target based on at least one target device.

[0122] Specifically, S703 is the same as S203, and will not be repeated here.

[0123] S704 inputs the current request, target user profile data, and target comprehensive environment data into the action planning model, and outputs the target control parameters for at least one target device and the target action thought chain corresponding to the target control parameters.

[0124] Specifically, the target action thought chain includes a complete reasoning path from parsing the current request to generating target control parameters, enhancing the interpretability and decision-making transparency of the action planning model. The action planning model is capable of in-depth analysis and processing based on the input current request, target user profile data, and target comprehensive environmental data. By comprehensively considering this data, the action planning model can accurately output target control parameters for at least one target device. These target control parameters directly guide the target device to execute the corresponding target action. Furthermore, to make the action planning process more transparent and interpretable, the action planning model also outputs the target action thought chain corresponding to the target control parameters. This target action thought chain records in detail the entire reasoning process of the action planning model from input data to output control parameters, demonstrating how the action planning model comprehensively analyzes the correlation between the current request, target user profile data, and target comprehensive environmental data, greatly improving the user's understanding and trust in the action planning model's decision-making process.

[0125] Optionally, when the target control parameters and the target action thought chain are information in the same output of the action planning model, a special symbol or phrase can be selected as a special token, either by customizing it or by selecting it from the vocabulary of the trained base model. When the action planning model outputs, the part before the special token is output as the target action thought chain in a streaming manner, printed one by one on the target interactive display page corresponding to the current request; while the part after the special token is stored in the cache and output as the action planning parameters (target control parameters) in a non-streaming manner, and passed to the control module of the downstream hardware device (e.g., the target device). This embodiment of the application can achieve the separation of non-streaming and streaming output through the special token method, ensuring that the target control parameters and the target action thought chain are processed and displayed in different ways.

[0126] Optionally, the target action thought chain may consist of, but is not limited to, multiple sub-chains, such as, but not limited to, a request intent recognition chain, a user preference mapping chain, an environment adaptation chain, and a parameter optimization chain. Each sub-chain corresponds to a logical module within the model, ensuring that the output control parameters not only meet the user's immediate needs but also adapt to long-term health status, device usage habits, and overall environmental conditions.

[0127] Optionally, when the action planning model outputs the time series of target control parameters for each target device, it can also simultaneously generate the corresponding target action thought chain time series. The aforementioned action thought chain time series includes target action thought chains corresponding to multiple time nodes arranged in chronological order. These multiple time nodes include the current time and at least one future time. The target action thought chain corresponding to the future time node can be, but is not limited to, the predictive inference path for the target control parameters of each target device corresponding to the future time node, generated based on predictive information such as user request intent, comprehensive environmental changes, user health status, and usage preference shifts at the future time node.

[0128] In some possible embodiments, such as Figure 7 As shown, in step S704 above, after inputting the current request, target user profile data, and target comprehensive environment data into the action planning model, and outputting the target control parameters for at least one target device and the target action thought chain corresponding to the target control parameters, the action planning method may also include, but is not limited to: S705 transmits each target control parameter to the corresponding target device so that the target device can perform the corresponding target action based on the target control parameters.

[0129] Specifically, after the motion planning model outputs the target control parameters for each target device, these parameters can be directly transmitted to the corresponding target device. Upon receiving the target control parameters from the AI ​​device, the target device will convert them into specific execution instructions based on its built-in control logic and algorithms, thereby driving the device to perform the corresponding target action. These target actions may include, but are not limited to, starting, stopping, adjusting operating parameters, and switching operating modes, aiming to meet the user's current and future device usage needs while optimizing device operating efficiency and energy consumption.

[0130] Optionally, the target device can provide real-time feedback on its operating status or the final execution status of the target action (e.g., whether the user agrees to the target device operating with the planned target control parameters) to the AI ​​device during execution, forming a closed-loop control. Based on the feedback data, the AI ​​device dynamically adjusts subsequent target control parameters and action thought chains to ensure continuous optimization of the control strategy. Especially in multi-device collaborative scenarios, devices can share action thought chain information to achieve coordinated decision-making and synchronized actions, further enhancing the overall system's intelligence level and user experience. This enables precise control of individual devices and builds a deep collaborative relationship between people, equipment, and the environment, allowing technology to truly serve the natural behavioral logic of humans.

[0131] Optionally, the aforementioned target control parameters may also carry a corresponding target control time, which is used to instruct the target device to perform the corresponding target action at a specified time node. The target device will accurately and synchronously perform the target action according to the target control time, ensuring that the device completes the corresponding operation at the time predetermined by the user.

[0132] S706, the target action thought chain is transmitted to the target interactive display device corresponding to the target user, so that the target interactive display device displays the target action thought chain on the target interactive display page corresponding to the current request.

[0133] Specifically, the aforementioned target interactive display device may be, but is not limited to, a central control display screen, a user's mobile phone, or a designated target device capable of interactive display among at least one target device requiring action planning. The specific design can be tailored to the actual application scenario and user needs, and this application embodiment does not impose such limitations. After receiving the target action thought chain transmitted by the artificial intelligence device, the target interactive display device will analyze the key node information in the target action thought chain and present it in the form of a visual flowchart or natural language description on the target interactive display page corresponding to the current request, enabling the user to clearly understand the system's decision-making logic and execution plan.

[0134] Optionally, users can confirm, adjust, or reject the target action plan corresponding to each target device through the target interaction display page. Their feedback instructions will be synchronized to the artificial intelligence device in real time, triggering the dynamic reconstruction of the action thought chain and the replanning of control parameters. This strengthens the trust foundation of human-machine collaboration while ensuring the transparency of action planning operations. It is especially suitable for scenarios with high requirements for safety and operational certainty, such as home and medical settings, ensuring that users always maintain control and achieving full traceability and explainability of the target action execution process.

[0135] Please refer to the following. Figure 8 This is a schematic diagram of the structure of a device motion planning apparatus provided in an exemplary embodiment of this application. Figure 8 As shown, the device motion planning device 800 includes: The first receiving module 810 is used to receive the current request input by the target user; The first acquisition module 820 is used to acquire target user profile data corresponding to the target user; the target user profile data is based on at least one target memory tag corresponding to the target user; the target memory tag represents the relevant health status and device usage preferences of the target user. The second acquisition module 830 is used to acquire comprehensive environmental data of a target based on at least one target device; The action planning module 840 is used to input the current request, the target user profile data, and the target comprehensive environment data into the action planning model and output the target control parameters for at least one target device; the action planning model is trained based on sample training data of multiple known action planning results.

[0136] In one possible implementation, the aforementioned target user profile data is based on at least one of the following target memory tags corresponding to the aforementioned target user: target personal memory tag and target group memory tag; the aforementioned target personal memory tag represents the health status and device usage preferences of the aforementioned target user; the aforementioned target group memory tag represents the device usage preferences of the target group corresponding to the aforementioned target user.

[0137] In one possible implementation, the target user profile data is based on the target group memory tags and the target relationship group member memory tags corresponding to the target relationship group to which the target user belongs; the target relationship group member memory tags are composed of the target personal memory tags corresponding to the target user and the personal memory tags corresponding to other members in the target relationship group besides the target user.

[0138] In one possible implementation, the target group memory tag includes at least one of the following: target relationship group memory tag and target association group memory tag; the target relationship group memory tag represents the device usage preferences of the target user's target relationship group for each of the target devices and the shared preferences of the target relationship group; the target association group memory tag represents the device usage preferences of the target association group for the target device and the health status of the target association group.

[0139] In one possible implementation, the first acquisition module 820 includes: The first construction unit is used to construct the target personal memory tag based on the target user's memory data on each of the target devices. The second construction unit is used to construct the target group memory tags based on the device memory data and / or group knowledge base corresponding to each of the above target devices; the group knowledge base includes device usage preference data of different types of groups. The third construction unit is used to construct target user profile data corresponding to the target user based on the aforementioned target individual memory tags and / or the aforementioned target group memory tags.

[0140] In one possible implementation, the aforementioned target group memory tags include at least one of the following: target relationship group memory tags, target association group memory tags; The third building block mentioned above is specifically used for: The target relationship group memory tags are constructed based on the device memory data corresponding to each of the aforementioned target devices; the device memory data includes the interaction data of each member in the target relationship group to which the target user belongs on each of the aforementioned target devices; and / or, based on the target basic information corresponding to the aforementioned target user, the target association group knowledge of the target association group corresponding to the aforementioned target user is determined from the group knowledge base, and the target association group memory tags are constructed based on the target association group knowledge; the group knowledge base includes device usage preference data and health status of different types of groups.

[0141] In one possible implementation, the motion planning device 800 further includes: The third acquisition module is used to acquire the historical interaction data corresponding to the aforementioned target users; The aforementioned action planning module 840 is specifically used to: input the aforementioned current request, the aforementioned target user profile data, the aforementioned target comprehensive environment data, and the aforementioned historical interaction data into the action planning model, and output target control parameters for at least one of the aforementioned target devices.

[0142] In one possible implementation, the aforementioned sample training data is generated by the target large model based on sample seed data carrying generation state labels and data rationality labels; the aforementioned generation state labels represent the control parameter types generated based on the aforementioned sample seed data for action planning; the aforementioned data rationality labels represent the rationality of the planned control parameters in the aforementioned sample seed data; the aforementioned sample seed data includes sample interaction requests corresponding to sample users with known action planning results, sample user profile data, and sample comprehensive environment data; and / or, the aforementioned action planning model is obtained by preference alignment optimization based on preference optimization training data; the aforementioned preference optimization training data includes sample requests with multiple known preference planning parameter pairs; the aforementioned preference planning parameter pairs include positive preference parameters and negative preference parameters planned for the aforementioned sample requests.

[0143] In one possible implementation, the aforementioned motion planning module 840 is specifically used for: Input the aforementioned current request, the aforementioned target user profile data, and the aforementioned target comprehensive environment data into the action planning model, and output the target control parameters for at least one target device and the target action thought chain corresponding to the aforementioned target control parameters.

[0144] In one possible implementation, the motion planning device 800 further includes: The first transmitting module is used to transmit each of the above-mentioned target control parameters to the corresponding target device, so that the target device performs the corresponding target action based on the above-mentioned target control parameters; The second sending module is used to transmit the aforementioned target action thought chain to the target interactive display device corresponding to the target user, so that the target interactive display device displays the aforementioned target action thought chain on the target interactive display page corresponding to the current request.

[0145] In one possible implementation, the aforementioned motion planning module 840 is specifically used for: The above-mentioned current request, the above-mentioned target user profile data, and the above-mentioned target comprehensive environment data are input into the action planning model, and the output is a time series of target control parameters for at least one target device; the above-mentioned target control parameter time series includes the target control parameters corresponding to multiple time nodes arranged in chronological order; the above-mentioned multiple time nodes include the current time and at least one future time.

[0146] In one possible implementation, the aforementioned current request is a voice request; The first acquisition module 820 mentioned above includes: The information identification unit is used to identify the information of the current request and obtain the target basic information corresponding to the target user. The query unit is used to query at least one target memory tag corresponding to the target user in the memory bank based on the aforementioned target basic information; the aforementioned memory bank includes individual memory tags corresponding to multiple users and / or group memory tags corresponding to multiple groups. The fourth construction unit is used to construct target user profile data corresponding to the target user based on at least one of the above-mentioned target memory tags.

[0147] In one possible implementation, the motion planning device 800 further includes: The status detection module is used to input the current request into the disease status detection model and output the disease status detection result corresponding to the target user based on the voice status of the current request. The first update module is used to update the target personal memory tags corresponding to the target users in the memory bank based on the above disease state detection results.

[0148] In one possible implementation, the above method is applied to an artificial intelligence device, and the above-mentioned motion planning device 800 further includes: The second update module is used to update the relevant group memory tags in the memory bank according to the target time interval, based on the device memory data of each device connected to the aforementioned artificial intelligence device within the target time interval.

[0149] In one possible implementation, the above method is applied to an artificial intelligence device, which is connected to the target device; the motion planning device 800 further includes: The second receiving module is used to receive device memory data uploaded by each of the aforementioned target devices; The distributed storage module is used to distribute and store the device memory data of different target devices in different device structures. An integrated storage module is used to integrate user memory data for the same user from the device memory data corresponding to each of the above target devices and store it in the user structure corresponding to the above user.

[0150] In one possible implementation, the first acquisition module 820 includes: The first determining unit is used to determine the target personal memory tag based on the target user structure corresponding to the target user. The second determining unit is used to determine the memory tags of the target group based on the device structure corresponding to each of the above-mentioned target devices. The third construction unit is used to construct target user profile data corresponding to the target user based on the aforementioned target individual memory tags and / or the aforementioned target group memory tags.

[0151] The division of modules in the above-described equipment motion planning device is for illustrative purposes only. In other embodiments, the equipment motion planning device can be divided into different modules as needed to complete all or part of the functions of the above-described equipment motion planning device. The implementation of each module in the equipment motion planning device provided in the embodiments of this specification can be in the form of a computer program. This computer program can run on a terminal or server. The program modules constituted by this computer program can be stored in the memory of the terminal or server. When the computer program is executed by a processor, it implements all or part of the steps of the equipment motion planning method described in the embodiments of this specification.

[0152] Please see below. Figure 9 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 9 As shown, the electronic device 900 may include: at least one processor 910, at least one network interface 920, user interface 930, memory 940, and at least one communication bus 950.

[0153] The communication bus 950 is used to enable communication between these components.

[0154] The network interface 920 may include, but is not limited to, a low-power Bluetooth module, a near field communication (NFC) module, a wireless Fidelity (Wi-Fi) module, etc.

[0155] The user interface 930 may include a display screen and a camera. Optionally, the user interface 930 may also include a standard wired interface and a wireless interface.

[0156] The processor 910 may include one or more processing cores. The processor 910 connects to various parts within the electronic device 900 using various interfaces and lines, and performs various functions and processes data by running or executing instructions, programs, code sets, or instruction sets stored in the memory 940, and by calling data stored in the memory 940. Optionally, the processor 910 may be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 910 may integrate one or a combination of several of the following: Central Processing Unit (CPU), Graphics Processing Unit (GPU), and modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the content required for display; and the modem handles wireless communication. It is understood that the modem may also be implemented as a separate chip without being integrated into the processor 910.

[0157] The memory 940 may include random access memory (RAM) or read-only memory (ROM). Optionally, the memory 940 may include a non-transitory computer-readable storage medium. The memory 940 can be used to store instructions, programs, code, code sets, or instruction sets. The memory 940 may include a program storage area and a data storage area. The program storage area may store instructions for implementing an operating system, instructions for at least one function (such as memory management functions, user profile building functions, device action planning functions, etc.), and instructions for implementing the various method embodiments described above. The data storage area may store data involved in the various method embodiments described above. Optionally, the memory 940 may also be at least one storage device located remotely from the aforementioned processor 910. Figure 9 As shown, the memory 940, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and program instructions.

[0158] exist Figure 9In the illustrated electronic device 900, the user interface 930 is mainly used to provide an input interface for the user and to acquire user input data; while the processor 910 can be used to call program instructions stored in the memory 940. The aforementioned electronic device 900 can, but is not limited to, [functions such as...]. Figure 8 The device motion planning device 800 shown herein specifically executes the actions described in this application. Figure 2-6 The illustrated embodiment provides any step of the motion planning method.

[0159] This application also provides a computer storage medium storing instructions that, when run on a computer or processor, cause the computer or processor to execute one or more steps of any of the above methods. If the constituent modules of the above-described device action planning apparatus are implemented as software functional units and sold or used as independent products, they can be stored in the storage medium.

[0160] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted through the computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., digital versatile discs (DVDs)), or semiconductor media (e.g., solid-state disks (SSDs)).

[0161] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. The aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks. Unless otherwise specified, the technical features of this embodiment and its implementation can be combined arbitrarily.

[0162] The embodiments described above are merely preferred embodiments of this application and are not intended to limit the scope of this application. Any modifications and improvements made by those skilled in the art to the technical solutions of this application without departing from the spirit of this application should fall within the protection scope defined by the claims of this application.

Claims

1. A device motion planning method, characterized by, The method comprises: receiving a current request input by a target user; obtaining target user portrait data corresponding to the target user; the target user portrait data is composed based on at least one target memory label corresponding to the target user; the target memory label represents a related health status and / or device use preference of the target user; obtaining target comprehensive environment data based on at least one target device; inputting the current request, the target user portrait data and the target comprehensive environment data into an action planning model to output target control parameters for at least one target device; the action planning model is trained based on a plurality of known action planning result sample training data.

2. The method of claim 1, wherein, The target user portrait data is composed based on at least one of the following target memory labels corresponding to the target user: target personal memory label, target group memory label; the target personal memory label represents the health status and device use preference of the target user; the target group memory label represents the device use preference of the target group corresponding to the target user.

3. The method of claim 2, wherein, The target user portrait data is composed based on the target group memory label and the target relationship group member memory label corresponding to the target relationship group in which the target user is located; the target relationship group member memory label is composed of the target personal memory label corresponding to the target user and the personal memory label corresponding to other members in the target relationship group except the target user.

4. The method of claim 2 or 3, wherein, The target group memory label includes at least one of the following: target relationship group memory label, target associated group memory label; the target relationship group memory label represents the device use preference of the target relationship group for each target device and the sharing preference of the target relationship group; the target associated group memory label represents the device use preference of the target associated group for the same type of device corresponding to the target device and the health status corresponding to the target associated group.

5. The method of claim 2 or 3, wherein, The target user portrait data corresponding to the target user is obtained, comprising: constructing the target personal memory label based on the target user memory data of the target user on each target device; constructing the target group memory label based on the device memory data corresponding to each target device and / or a group knowledge base; the group knowledge base includes device use preference data of different types of groups; constructing the target user portrait data corresponding to the target user based on the target personal memory label and / or the target group memory label.

6. The method of claim 2 or 3, wherein, The target group memory label includes at least one of the following: target relationship group memory label, target associated group memory label; The target group memory label is constructed based on the device memory data corresponding to each target device and / or a group knowledge base, comprising: constructing the target relationship group memory label based on the device memory data corresponding to each target device; The device memory data includes interaction data corresponding to each member in the target relationship group in which the target user is located on each target device; and / or determine target associated group knowledge of a target associated group corresponding to the target user from a group knowledge base based on target basic information corresponding to the target user, and construct a target associated group memory tag based on the target associated group knowledge; the group knowledge base includes device use preference data and health status of different types of groups.

7. The method of any one of claims 1-3, wherein, Before the inputting the current request, the target user portrait data, and the target comprehensive environment data into the action planning model and outputting the target control parameters for at least one target device, the method further comprises: obtaining historical interaction data corresponding to the target user; the inputting the current request, the target user portrait data, and the target comprehensive environment data into the action planning model and outputting the target control parameters for at least one target device comprises: inputting the current request, the target user portrait data, the target comprehensive environment data, and the historical interaction data into the action planning model and outputting the target control parameters for the at least one target device.

8. The method of claim 1, wherein, The sample training data is generated by a target large model based on sample seed data carrying generation state annotations and data rationality annotations; the generation state annotations represent the type of control parameters generated based on the sample seed data for action planning; the data rationality annotations represent the rationality of the control parameters planned in the sample seed data; the sample seed data includes sample interaction requests, sample user portrait data, and sample comprehensive environment data of a sample user corresponding to a known action planning result; and / or the action planning model is obtained by preference alignment optimization based on preference optimization training data; the preference optimization training data includes sample requests of multiple known preference planning parameter pairs; the preference planning parameter pair includes a preference positive parameter and a preference negative parameter planned for the sample request.

9. The method of claim 1, wherein, the inputting the current request, the target user portrait data, and the target comprehensive environment data into the action planning model and outputting the target control parameters for at least one target device comprises: inputting the current request, the target user portrait data, and the target comprehensive environment data into the action planning model and outputting the target control parameters for at least one target device and the target action thought chain corresponding to the target control parameters.

10. The method of claim 9, wherein, After the inputting the current request, the target user portrait data, and the target comprehensive environment data into the action planning model and outputting the target control parameters for at least one target device and the target action thought chain corresponding to the target control parameters, the method further comprises: transmitting each target control parameter to the corresponding target device to enable the target device to perform a corresponding target action based on the target control parameter; transmitting the target action thought chain to a target interaction display device corresponding to the target user to enable the target interaction display device to display the target action thought chain on a target interaction display page corresponding to the current request.

11. The method of claim 1, wherein, The step of inputting the current request, the target user profile data, and the target comprehensive environment data into the action planning model, and outputting target control parameters for at least one target device, includes: The current request, the target user profile data, and the target comprehensive environment data are input into the action planning model, and the target control parameter time series for at least one target device is output. The target control parameter time series includes target control parameters corresponding to multiple time nodes arranged in chronological order. The multiple time nodes include the current time and at least one future time.

12. The method of claim 1, wherein, The current request is a voice request; The step of obtaining the target user profile data corresponding to the target user includes: Information identification is performed on the current request to obtain the target basic information corresponding to the target user; Based on the target's basic information, at least one target memory tag corresponding to the target user is queried from the memory bank; the memory bank includes multiple personal memory tags corresponding to each user and / or multiple group memory tags corresponding to each group. Based on the at least one target memory tag, construct the target user profile data corresponding to the target user.

13. The method of claim 12, wherein, After receiving the current request input by the target user, the method further includes: The current request is input into the disease state detection model, and the disease state detection result corresponding to the target user is output based on the voice state of the current request. The target personal memory tag corresponding to the target user in the memory bank is updated based on the disease state detection results.

14. The method of claim 12, wherein, The method is applied to an artificial intelligence device, and the method further includes: According to the target time interval, the relevant group memory tags in the memory bank are updated based on the device memory data of each device connected to the artificial intelligence device within the target time interval.

15. The method of claim 2 or 3, wherein, The method is applied to an artificial intelligence device, which is connected to the target device; the method further includes: Receive device memory data uploaded by each of the target devices; The device memory data of different target devices are distributed and stored using different device structures. The user memory data for the same user in the device memory data corresponding to each target device is integrated and stored in the user structure corresponding to the user.

16. The method of claim 15, wherein, The step of obtaining the target user profile data corresponding to the target user includes: The target personal memory tag is determined based on the target user structure corresponding to the target user; The target group memory tag is determined based on the device structure corresponding to each target device; Based on the target individual memory tags and / or the target group memory tags, construct the target user profile data corresponding to the target user.

17. A device motion planning system, comprising: The system includes an artificial intelligence device and at least one target device connected to the server; wherein: The target device is used to report to the artificial intelligence device the current request it has received from the target user; The artificial intelligence device is configured to execute the device action planning method steps as described in any one of claims 1-16 based on the current request, and to transmit corresponding target control parameters to each of the target devices, so that each of the target devices performs a corresponding target action based on the corresponding target control parameters.

18. An apparatus motion planning device, characterized by, The device includes: The first receiving module is used to receive the current request input by the target user; The first acquisition module is used to acquire target user profile data corresponding to the target user; the target user profile data is based on at least one target memory tag corresponding to the target user; the target memory tag represents the target user's relevant health status and device usage preferences. The second acquisition module is used to acquire comprehensive environmental data of the target based on at least one target device; The action planning module is used to input the current request, the target user profile data, and the target comprehensive environment data into the action planning model, and output target control parameters for at least one target device; the action planning model is trained based on sample training data of multiple known action planning results.

19. An electronic device, comprising: include: Processor and memory; The processor is connected to the memory; The memory is used to store executable program code; The processor runs a program corresponding to the executable program code stored in the memory to perform the steps of the method as described in any one of claims 1-16.

20. A computer storage medium, comprising, The computer storage medium stores a plurality of instructions adapted for loading by a processor and executing the steps of the method as described in any one of claims 1-16.

21. A computer program product comprising instructions that, when run on a computer or processor, cause the computer or processor to perform the steps of the method as claimed in any one of claims 1-16.

Images