A personalized life skill intention representation and structured modeling method based on human behavior input

Abstract

The application discloses a kind of individualized life skill intention representation and structured modeling method based on human behavior input, it is related to artificial intelligence, human-computer interaction, service robot and embodied intelligent technical field, specifically a kind of individualized life skill intention representation and structured modeling method based on human behavior input, the behavior input of user in real scene is collected and structured analysis, constructs intention model generation data, after user confirmation, solidification storage, and the like executor robot generates instruction accordingly.The data is independent of the action trajectory, only based on the behavior input, cannot be executed without confirmation, the structure is not changed in execution, deviates from the constraint and can be aborted or rolled back, suitable for family service, old-age care and the like scene, support multi-skill conflict verification.The robot system for executing the method is also provided.

Description

Technical Field

[0001] This invention relates to the fields of artificial intelligence, human-computer interaction, service robots and embodied intelligence, specifically a method for personalized life skills intent representation and structured modeling based on human behavioral input. Background Technology

[0002] In today's era of rapid technological advancement, service robots are gradually becoming integrated into our daily lives. As these robots begin to take on various tasks in the home, the technological challenges they face are becoming increasingly prominent.

[0003] Traditional service robot technology relies on several methods to achieve its functions. One common approach is rule-based execution using preset action templates. This method involves pre-setting a series of action templates, which the robot follows when performing tasks. However, this approach is limited by its lack of flexibility, making it difficult to adapt to the personalized needs of different users in different scenarios. Each user has unique lifestyle habits and operational preferences, and preset action templates often fail to accurately meet these personalized requirements.

[0004] Behavior inference based on large models or reinforcement learning is also a commonly used technique. This method attempts to infer the robot's behavior using large amounts of data and algorithms. However, this inference process is often complex, and the accuracy of the results is difficult to guarantee. Due to a lack of in-depth understanding of the user's true intentions, the robot may perform behaviors that do not meet the user's expectations.

[0005] Imitation learning based on direct mapping of demonstrated actions also has some problems. This method involves having the robot observe human demonstrations and then directly mapping these actions into its own execution process. However, the drawback of this method is that it relies too heavily on specific action details. Once the external environment or task requirements change, the robot may be unable to perform the task accurately.

[0006] Furthermore, existing technologies have other shortcomings. For example, they struggle to accurately reflect the operational sequences, implicit preferences, and prohibitions of different users in real life. Each user has their own operational habits and preferences in daily life, some of which may be implicit and not easily noticed. Moreover, in certain situations, there may be prohibitions, such as certain items not being able to be moved at will or certain operations not being able to be performed at specific times. Existing technologies find it difficult to comprehensively consider these factors.

[0007] Life skills are highly dependent on specific actions or model parameters, and the lack of transferability is a serious problem. This means that when a robot needs to perform different tasks or work in different environments, it needs to relearn and adapt to new actions and parameters, which not only increases the robot's learning cost but also reduces its work efficiency.

[0008] The untraceable and unverifiable origins of robot behavior pose safety risks. Because the source and basis of robot behavior are unclear, it is difficult to trace the root cause and effectively resolve any errors or anomalies. This is particularly dangerous in home settings, where the safety and privacy of family members are at stake.

[0009] In summary, existing technologies have certain limitations in achieving personalization, flexibility, portability, and security of service robots, thus urgently requiring a new technological solution to address these issues. Summary of the Invention

[0010] The purpose of this invention is to provide a personalized life skills intent representation and structured modeling method based on human behavioral input. By collecting and structurally analyzing users' real behavioral input, an intent model confirmed by the user is constructed, enabling robots and other actuators to accurately understand and perform tasks according to stable intents, thereby improving the adaptability and safety of home services and special care scenarios.

[0011] To achieve the above objectives, the present invention provides the following technical solution: a method for personalized life skills intent representation and structured modeling based on human behavioral input, comprising: Collect human behavioral inputs from users in real-life scenarios, formed through body movements, operation sequences, ways of using objects, and interactions with the environment; The collected human behavioral inputs are subjected to behavioral element extraction, sequential relationship classification, and contextual association analysis to form identifiable behavioral patterns. Based on the results of behavioral pattern analysis, user goals and operational priorities are identified, a life skills intent model reflecting the user's personalized needs is constructed, and life skills intent data with a structured expression is generated based on the model. The life skills intention data is presented in a visual interface or fed back to the user in a perceptible way for the user to check and confirm. After the user completes the confirmation operation, the life skills intent data of the current version is marked and stored in a designated storage medium to ensure its stability and reusability; Upon receiving a task instruction, the robot or actuator reads the stored life skill intent data, analyzes its structural elements and sequential logic, and generates corresponding execution instructions to achieve the corresponding life skill task.

[0012] Furthermore, the life skill intention data is independent of specific action trajectories or execution parameters. Its representation content includes task objectives, behavioral stages, object associations, and sequential logic, excluding the specific movement path, speed setting, angle transformation, or force control parameters of the execution device, so as to ensure that the intention has universality and transferability under different physical conditions.

[0013] Furthermore, the generation process of the life skills intention data does not introduce preset knowledge templates or external reasoning results. All of its constituent information comes directly from the product of structured analysis of the collected human behavioral input, so as to ensure that the intention content is consistent with the individual's actual behavior and without external intervention.

[0014] Furthermore, life skills intent data that has not been confirmed by the user item by item or as a whole, regardless of whether its generation process is complete, shall not be used as the basis for execution or invoked by the robot or the execution entity for any operation, in order to prevent the accidental execution of unauthorized or inaccurate task content.

[0015] Furthermore, during the execution of life skills tasks, modifications to the structure, element relationships, and sequential logic of the confirmed and invoked life skills intent data are prohibited to ensure the stability of the execution process and the continuous consistency of the task objectives.

[0016] Furthermore, during execution, if the result or path of the current execution behavior deviates from the behavioral constraints specified in the life skills intention data, the termination command or rollback mechanism is immediately triggered through sensor information or status feedback, causing the executor to stop the current operation and return to the previous valid state or terminate the task.

[0017] Furthermore, the method is deployed and implemented in a home environment, and is applicable to typical life skill tasks performed by home service robots, such as organizing items, cleaning, meal preparation, and clothing handling, so as to support them in achieving operational responses that conform to the habits of family members without direct human control.

[0018] Furthermore, the method is also applicable to elderly care or disability assistance scenarios, supporting the collection and modeling of user behavior patterns in activities such as moving around, daily eating, personal cleaning, and medication access, so as to generate assistive operation intentions that fit the individual abilities and habits of users, thereby improving the feasibility and adaptability of self-care support.

[0019] Furthermore, when multiple life skills intents need to be invoked in parallel within the same time period, the system should pre-verify the potential resource competition, behavioral conflicts, or order contradictions between the intents, and prompt the user to adjust or reconfirm the intent content when inconsistencies or uncoordinated situations are found, in order to avoid task execution disorder or failure.

[0020] Furthermore, the system includes a behavior acquisition module, a behavior analysis module, an intent modeling module, a user confirmation module, a data storage module, an instruction generation module, and a task execution module. Each module works in concert and is configured to sequentially execute the method described in any one of claims 1 to 9 to achieve full-process control of the identification, modeling, confirmation, invocation, and execution of personalized life skills based on human behavioral input.

[0021] This invention provides a personalized method for representing and structurally modeling life skills intentions based on human behavioral input, which has the following beneficial effects: 1. Enable highly personalized and explainable life skills learning, and improve the naturalness of human-computer collaboration. This method uses natural human behavioral input from users in real-world scenarios as its source, and after structured analysis, forms a personalized intent model, enabling the robot to accurately capture individual habits and preferences. Because intent data is independent of specific action trajectories or execution parameters, it describes "why" rather than "how," thus preserving the user's personalized logic while possessing clear interpretability. Users can intuitively review and correct the model through a confirmation process, preventing the machine from executing rigid procedures. This allows for truly personalized human-machine collaboration in everyday tasks such as making tea and tidying up, significantly enhancing user trust and providing a natural interactive experience.

[0022] 2. Strengthen safety controls and execution reliability to reduce the risk of misoperation. Data not confirmed by the user is strictly prohibited from being executed, and the intent structure must not be altered during execution, thus eliminating the possibility of the machine arbitrarily interpreting or tampering with user needs. Once an execution deviates from preset constraints, it is triggered to abort or rollback, effectively setting up a dynamic safety barrier for the user. In scenarios involving personal or property safety, such as home services and elderly care, this mechanism effectively prevents dangerous actions caused by environmental changes or perception errors, allowing the robot to remain controllable and trustworthy while being relevant to daily life, enhancing users' reliance on and sense of security with the intelligent system.

[0023] 3. Adapt to diverse life scenarios and the needs of special groups, expanding application boundaries. The method is explicitly applicable to home service robots and scenarios such as elderly care and disability assistance, meaning its modeling process is compatible with different spatial layouts, object placements, and individual ability differences. For example, for people with mobility impairments, an intention model that matches their physical condition can be extracted based on their gesturing movements; for the elderly, steps can be simplified and safety points highlighted. This generalization ability based on real behavioral input allows the system to be quickly deployed to diverse environments without reprogramming for each group, providing customized support for a wider range of social groups and promoting the universality and compassion of intelligent assistive technology.

[0024] 4. Ensure the stability of intent and the orderly collaboration of multiple skills, and optimize performance in complex tasks. When multiple life skills are invoked in parallel, the system pre-checks for skill conflicts to ensure that each intention is logically and resource-intensively compatible. Simultaneously, the permanently stored intention structure remains unchanged throughout execution, avoiding cascading errors caused by real-time adjustments. This feature enables robots to handle multi-objective, complex tasks such as "preparing meals while caring for children," with each sub-skill progressing according to a predetermined order and conditions, minimizing mutual interference. For daily processes requiring high reliability, this mechanism significantly improves completion quality and efficiency, making intelligent systems better suited for long-cycle, multi-stage continuous services.

[0025] 5. Promote active user participation and continuous iteration to form a virtuous cycle of skill evolution. By introducing a user confirmation process, model generation is transformed from a one-way derivation into a two-way communication. Users can not only verify accuracy but also inject detailed preferences, making system learning a collaborative human-machine process. The confirmed and stored data ensures the stability of current execution and provides a baseline for subsequent comparison and optimization. As users repeatedly input behaviors in different scenarios, the system can accumulate diverse samples, gradually refining intent representations and achieving incremental skill improvement. This sense of participation and growth feedback not only enhances satisfaction but also drives the robot to integrate more deeply into personal lives, forming a sustainably evolving service ecosystem. Attached Figure Description

[0026] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0027] Figure 1 This is the overall flowchart for personalized life skills intent modeling and execution in this invention; Figure 2 This is a flowchart illustrating the intended data characteristics of life skills in this invention; Figure 3This is a flowchart illustrating the user confirmation and data storage process of this invention. Figure 4 This is a flowchart illustrating the constraints and exception handling during the execution process of this invention. Figure 5 This is a flowchart illustrating the multi-skill parallelism and system applicability of the present invention. Detailed Implementation

[0028] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses consistent with some aspects of this disclosure as detailed in the appended claims.

[0029] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0030] How to use: Step 1: Collect human behavioral input. Users provide human behavioral input to the system in real-life scenarios through natural interactions (such as demonstrations, verbal descriptions, or operational demonstrations), which serves as the original basis for intent modeling.

[0031] Step 2: Structured analysis of behavioral input. The system parses the collected behavioral input, extracts key behavioral features and target orientations, sorts out the logical connections between behaviors, and forms structured analysis results that can represent the intentions of life skills.

[0032] Step 3: Construct and generate intent data. Based on the structured analysis results, construct a life skills intent model to generate life skills intent data that is independent of specific action trajectories or execution parameters; this data is generated solely based on the currently collected human behavioral input and does not involve other irrelevant information.

[0033] Step 4: User Confirmation of Intent Data. The generated intent data must be submitted for user confirmation to ensure it meets the user's actual needs; intent data that has not been confirmed by the user must not proceed to subsequent steps and is prohibited from use.

[0034] Step 5: Solidify and store confirmation data. After user confirmation, the intent data is solidified and stored to form a reusable personalized intent resource library.

[0035] Step 6: Generate and execute instructions. The robot or actuator calls the stored intent data to generate corresponding instructions and implement the action; the execution strictly follows the intent structure, and changes to its inherent framework are prohibited. If the execution deviates from the intent constraints (such as target deviation or logical conflict), the system immediately triggers an abort or rollback mechanism to ensure the accuracy of the intent.

[0036] Special scenario adaptation: If applied to home service robots, elderly care or disability assistance scenarios, potential conflicts between skills need to be verified when multiple skills are called in parallel to avoid intention interference; all links must ensure that the intention data comes only from the user's current behavior input and does not depend on external redundant data.

[0037] Example 1: A Personalized Life Skill Intent Representation and Structured Modeling Method for Making Fruit Salad in a Home Service Robot Scenario This method is applied to home service robots, targeting users' life skill needs for "making a fruit salad." First, it collects human behavioral input from real-life scenarios: the user stands in front of the kitchen counter, holding fruits such as apples, bananas, and strawberries, and verbally says "make a refreshing fruit salad," while picking up the apple and cutting it into chunks, peeling and slicing the banana, and removing the stems from the strawberries and cutting them in half. Finally, the user places these prepared fruits into a glass bowl—the entire set of behaviors is a natural demonstration of a life scenario, constituting the original input.

[0038] Next, the human behavioral input is subjected to structured analysis: the system analyzes the user's presentation and language, extracts key features—the core goal of "making a fruit salad" is "to obtain a bowl of mixed and pre-cut fruit"; the behavioral logic is "selecting fruit → processing fruit (cutting into chunks / slices / removing stems) → mixing and placing in a bowl"; the target is clearly "a refreshing combination of mixed fruits", forming a structured analysis result.

[0039] Then, based on the analysis results, a life skills intent model was constructed to generate life skills intent data: the model focuses on the essential intent of "making a fruit salad," generating intent data independent of specific action trajectories (such as whether to cut apples into 3 cm or 5 cm pieces) and execution parameters (such as whether to use a ceramic bowl or a glass bowl). The content is "The user needs to mix various fruits after cutting and preparing them into a refreshing fruit salad." This data is generated solely based on this user demonstration and language input, without introducing any external preset rules.

[0040] The system then confirms with the user: the system presents the intent data to the user, and after the user checks it, they say, "This is what I want to do, no need to change it"—data that has not been confirmed cannot proceed to the next step.

[0041] Confirm and store: Store the intent data in the robot's "personalized skill library" and mark it as "user-exclusive fruit salad making intent".

[0042] Finally, the execution instructions are generated: the robot calls the data and generates the execution instructions "select common fruits → process in the usual way (no need for precise size) → mix and put in a bowl"; during execution, it strictly follows the intent structure (without adjusting the logic of "cutting and mixing"). If the robot mistakenly adds vegetables (deviating from the intent constraint of "fruit"), the system immediately triggers a rollback and re-executes the steps of cutting and mixing the fruit.

[0043] Example 2: A Personalized Method for Representing and Structured Modeling Intent to Assist with Dressing in Elderly Care Scenarios This method is applied to assistive devices in elderly care scenarios, specifically addressing the skill support needs of seniors for "self-adjusting their coats." First, human behavioral input is collected: a caregiver sits beside the elderly person, picks up a cardigan, and while saying, "Let's put this cardigan on slowly," demonstrates "first, put your right hand into the right sleeve → then your left hand into the left sleeve → adjust the collar → button the middle button," with slow movements throughout, conforming to the elderly person's physical habits—this is behavioral input in a real care scenario.

[0044] Next, a structured analysis was performed: the system analyzed the demonstration and language, and extracted key features—the goal of “assisting in putting on a coat” is “to enable the elderly to complete the cardigan dressing independently”; the behavioral logic is “extend the right hand into the right sleeve → extend the left hand into the left sleeve → adjust the collar → button the middle button”; the target is “an independent dressing process that conforms to the elderly’s physical abilities”, forming a structured result.

[0045] Then, a model is built to generate intent data: the model ignores specific action details (such as the angle of the right hand extending into the sleeve, the force with which the button is fastened), generating intent data independent of the execution parameters: "The user needs to follow the process of 'extending the right hand into the right sleeve → extending the left hand into the left sleeve → adjusting the collar → fastening the middle button' to independently put on the cardigan." This data is derived solely from the caregiver's demonstration and verbal input in this instance.

[0046] User confirmation: The intent data is read to the elderly person, and the elderly person nods and indicates "Yes, that's right, I usually do it this way" - the data is not effective before confirmation.

[0047] Fixed storage: The data is stored in the "elderly skills library" of the care device and labeled with "elderly exclusive cardigan wearing intention".

[0048] Generate execution instructions: The device calls up data and generates a voice prompt to "prompt the elderly to follow the procedure"; if the device incorrectly guides the elderly to "extend the left hand first" (deviating from the intention constraint of "extend the right hand first") during execution, the system immediately stops the prompt and replays the correct procedure.

[0049] Example 3: A Personalized Method for Representing and Structured Modeling Intents of Autonomously Pouring Warm Water in Assisted Living Skills in a Disabled Scenarios This method is applied to the executioner in assistive scenarios for people with disabilities, targeting the user's skill training needs for "self-pouring warm water". First, human behavioral input is collected: The therapist stands next to the water dispenser, holding the user's special water cup, and says, "Let's learn to pour warm water, don't burn yourself", while demonstrating "walk to the water dispenser → press the 'warm water' button → wait for the water flow to stabilize and then take the water cup to fill it → fill it to eight-tenths and put it down" - the actions are adapted to the user's mobility (such as walking slowly and holding the cup lightly).

[0050] Next, structured analysis: The system analyzes the demonstration and language, extracting key features—the goal of "pouring warm water" is "to safely obtain warm water at a suitable temperature"; the behavioral logic is "go to the water dispenser → press the warm water button → fill the dispenser with a steady flow of water → fill it to eight-tenths full"; the target is "a safe water-filling process to avoid scalding", forming a structured result.

[0051] Then, a model is built to generate intent data: the model does not focus on specific movement trajectories (such as walking speed or hand posture when filling the water dispenser), generating intent data independent of execution parameters: "The user needs to press the 'warm water' button at the water dispenser, fill it to eight-tenths full after the water flow stabilizes, and complete the self-pouring of warm water." This data is based solely on the input from the therapist's demonstration in this instance.

[0052] User confirmation: Ask the user to repeat the process. If the user says, "Yes, it's pressing the warm water button and filling it to eight-tenths full," the data is valid after confirmation.

[0053] Fixed storage: The data is stored in the assistive device's "disability skills library" and marked as "user exclusive - pouring warm water intention".

[0054] Generate execution instructions: The device calls up data and generates instructions that provide step-by-step voice prompts for user operation; if the device incorrectly prompts "Press the hot water button" (which deviates from the intended constraint of "warm water") during execution, the system immediately stops and corrects it to "Please press the warm water button".

[0055] Example 4: A Personalized Life Skill Intent Representation and Structured Modeling Method for Parallel Invocation of Multi-Skill Coffee and Hot Bread This method is applied to home service robots to handle users' multi-skill needs of "making coffee and heating bread simultaneously". First, human behavioral input is collected: the user stands in the kitchen and says, "I want to drink hot coffee and eat hot bread in the morning", while demonstrating "making coffee: put in coffee powder → add warm water → stir; heating bread: put in bread slices → press the 'heat' button → wait for prompt sound" - two independent sets of behaviors constitute the input.

[0056] Next, structured analysis: The system analyzes the two sets of behaviors separately, extracting the key features of "making coffee" (adding coffee powder → adding warm water → stirring, with the goal of "getting hot coffee") and the key features of "heating bread" (adding bread slices → pressing the heating button → waiting for the prompt sound, with the goal of "getting hot bread"), forming two independent structured analysis results.

[0057] Then, the model was built to generate intent data: independent intent data was generated for the two skills: "Make coffee: Follow the process of 'put in coffee powder → add warm water → stir' to make hot coffee" and "Heat bread: Follow the process of 'put in bread slices → press the heating button → wait for the prompt sound' to heat the bread", both based solely on the user demonstration input.

[0058] User confirmation: After the user checks, they indicate "Yes, I want both" - data is unavailable before confirmation.

[0059] Fixed storage: Store the two intent data into the robot's "multi-skill library" and label them as "parallel coffee + bread intent".

[0060] Generate execution instructions and conflict verification: When the robot calls data, it first verifies that there is no conflict between the two skills (the left side of the countertop is used for making coffee, and the right side is used for heating bread, so they do not interfere with each other); if the "heating button" for "heating bread" is mistakenly pressed as the "defrosting button" (deviating from the intention constraint of "heating") during execution, the system triggers a rollback and re-executes the heating step.

[0061] Example 5: A Personalized Method for Representing and Structuredly Modeling Life Skills Intents to Remind Patients to Take Medication on Time in Elderly Care Scenarios This method is applied to reminder devices in elderly care scenarios, specifically addressing users' need to "take antihypertensive medication on time." First, human behavioral input is collected: a family member holds a box of antihypertensive medication, saying, "Take this antihypertensive medication every morning at 8 am, just one pill," while demonstrating "opening the box → taking one pill → swallowing with warm water"—behavioral input from a real-life scenario.

[0062] Next, structured analysis was performed: the system parsed the demonstration and language, and extracted key features—the goal of “reminding to take medication” is “to take antihypertensive drugs on time and in the correct dosage”; the behavioral logic is “8 am → open the medicine box → take 1 pill → take with warm water”; the target is “timely reminders to ensure accurate medication”, forming a structured result.

[0063] Then, a model is built to generate intent data: the model ignores specific action details (such as how the pillbox is opened, the amount of water taken), and generates intent data independent of the execution parameters: "The user needs to take the antihypertensive medication every morning at 8:00 AM by following the process of 'opening the pillbox → taking 1 pill → taking it with warm water'." This data is solely based on the input from the family member in this instance.

[0064] User confirmation: The elderly person repeats, "Yes, one pill every day at 8 o'clock" - data is not valid before confirmation.

[0065] Fixed storage: Store the data in the "Medication Skills Library" of the reminder device and label it with "Senior-specific - Antihypertensive Drug Reminder Intent".

[0066] Generate execution instructions: The device triggers a voice reminder at 8:00 AM to guide the user to operate according to the procedure; if the device incorrectly reminds the user of "8:00 PM" during execution (deviating from the intention constraint of "8:00 AM"), the system immediately stops the incorrect reminder and re-broadcasts "It's time to take your medication at 8:00 AM".

[0067] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for personalized life skills intent representation and structured modeling based on human behavioral input, characterized in that, include: Collect user input based on human behavior in real-life scenarios; Perform structured analysis on the aforementioned human behavioral input; Based on the analysis results, a life skills intention model is constructed, and life skills intention data is generated. User confirmation is required for the aforementioned life skills intent data; After confirmation, the life skills intention data is solidified and stored. Execution instructions are generated on the robot or actuator based on the life skills intent data.

2. The method for personalized life skills intent representation and structured modeling based on human behavioral input according to claim 1, characterized in that: The life skills intent data is independent of specific action trajectories or execution parameters.

3. The method for personalized life skills intent representation and structured modeling based on human behavioral input according to claim 1, characterized in that: The life skills intent data is generated solely based on human behavioral input.

4. The method for personalized life skills intent representation and structured modeling based on human behavioral input according to claim 1, characterized in that: Life skills intent data that has not been confirmed by the user must not be used for execution.

5. The method for personalized life skills intent representation and structured modeling based on human behavioral input according to claim 1, characterized in that: The structure of the life skills intent must not be altered during implementation.

6. The method for personalized life skills intent representation and structured modeling based on human behavioral input according to claim 1, characterized in that: When the execution behavior deviates from the intended constraints, an abort or rollback mechanism is triggered.

7. The personalized life skills intent representation and structured modeling method based on human behavioral input according to claim 1, characterized in that: The method is applicable to home service robots.

8. The method for personalized life skills intent representation and structured modeling based on human behavioral input according to claim 1, characterized in that: The method is applicable to elderly care or disability assistance scenarios.

9. The method for personalized life skills intent representation and structured modeling based on human behavioral input according to claim 1, characterized in that: Validate skill conflicts when multiple skills are invoked in parallel.

10. A robot system, characterized in that, The system is configured to perform the method according to any one of claims 1 to 9.

Images