Task execution method of agent with body and agent with body

Abstract

The invention is suitable for the technical field of artificial intelligence, and provides a task execution method of an agent and the agent, and the method comprises the steps: analyzing a user instruction after the user instruction is received, obtaining a target task needing to be executed and task elements of the target task, enabling the agent to accurately understand the intention of a user, and improving the user experience. And execution deviation caused by semantic ambiguity or information loss is avoided. Performing feasibility analysis on the target task according to the target task and the task elements to obtain a task analysis result; and when the task is determined to be feasible, executing the target task according to the target task and the current environment information. According to the method, whether the task can be completed or not can be dynamically evaluated before the task is executed, failure or system risks caused by forcibly executing the task when resources or capacities are insufficient are avoided, and meanwhile energy consumption resources are saved.

Description

Technical Field

[0001] This application belongs to the field of artificial intelligence technology, and in particular relates to a task execution method for an embodied intelligent agent and an embodied intelligent agent. Background Technology

[0002] In recent years, the rapid development of artificial intelligence technology has driven the application of intelligent agents in diverse environments, especially in the field of embodied AI, where agents are endowed with the ability to perceive their environment, understand tasks, and perform operations. These agents are typically deployed in real or simulated environments, and complete complex tasks such as home services, autonomous navigation, and industrial operations by perceiving, planning, and executing a series of actions.

[0003] Currently, embodied agents execute tasks according to user instructions after receiving them. However, due to insufficient understanding of the tasks by the embodied agents, the system may perform invalid operations or even cause the tasks to fail. Therefore, how to improve the task completion rate of embodied agents and reduce invalid operations has always been a concern in the industry. Summary of the Invention

[0004] This application provides a task execution method and an embodied intelligent agent, which can improve the task completion rate of the embodied intelligent agent and reduce invalid operations.

[0005] In a first aspect, embodiments of this application provide a task execution method for an embodied intelligent agent, including: In response to receiving a user instruction, the user instruction is parsed to determine the target task to be executed and the task elements of the target task, wherein the task elements include at least one of the target tool to be used to execute the task, the target task object, the geographical scope of the target task, and the target task type; Based on the capabilities of the embodied intelligent agent and the task elements, a feasibility analysis is performed on the target task to obtain the task analysis results; If the task analysis results indicate that the task is feasible, the target task is executed based on the target task and the collected environmental information of the embodied intelligent agent. If the task analysis results indicate that the task is not feasible, the execution of the target task shall be terminated.

[0006] In this application, upon receiving a user instruction, the instruction is parsed to obtain the target task and its elements, enabling the embodied agent to accurately understand the user's intent and avoid execution deviations caused by semantic ambiguity or missing information. Then, a feasibility analysis is performed on the target task based on its elements, yielding the analysis results. If feasibility is determined, the target task is executed based on the task and current environmental information. This application performs a feasibility analysis on the target task before execution to determine its feasibility. If feasible, the probability of completion is high, and the task can be executed; if infeasible, the probability is low, and the task can be skipped. This application dynamically assesses task completion before execution, preventing failures or system risks caused by the embodied agent forcibly executing tasks when resources or capabilities are insufficient, while also conserving energy resources.

[0007] In one possible implementation of the first aspect, the inherent capabilities of the embodied intelligent agent include invokeable tools, a range of activities, executable objects, and types of executable tasks; When the task elements include the target tools required to perform the task, the target task object, the geographical scope of the target task, and the target task type, the feasibility analysis of the target task based on the embodied intelligent agent's own capabilities and the task elements is performed to obtain the task analysis results, including: The target tool is searched from the available tools to obtain the tool search results; Determine whether the geographical range of the target task exists within the active range, and obtain the range determination result; Check if the target object exists in the executable object to obtain the object search result; Determine whether the executable task type contains the target task type to obtain the task type determination result; Based on the tool search results, the range judgment results, the object search results, and the task type judgment results, a feasibility analysis is performed on the target task to obtain the task analysis results.

[0008] In one possible implementation of the first aspect, executing the target task based on the target task and the collected current environmental information of the embodied intelligent agent includes: Based on the target task, the current environmental information of the embodied intelligent agent, and the sub-tasks already executed, plan the i-th step sub-task to be executed, where i is a positive integer greater than or equal to 1, the i-th step sub-task is the sub-task to be executed to complete the target task, and the executed sub-tasks are the sub-tasks executed before the i-th step. During the execution of the i-th step subtask, the contextual events of the environment in which the embodied intelligent agent is located are detected in real time, wherein the contextual events include user interaction events and physical environment events; If it is determined that the situational event is related to the target task, a new task target corresponding to the situational event is determined based on the event content of the situational event; The target task is updated based on the new task objective to obtain the updated target task; Based on the updated target task and executed subtasks, the subtasks that need to be executed at the current moment are re-determined, wherein the executed subtasks are the subtasks that were already executed when the target task was executed; Execute the redefined subtasks.

[0009] In this application, during the execution of subtasks, environmental information is detected in real time, and the tasks to be executed are re-determined based on the environmental information. This allows the embodied intelligent agent to make corresponding task modifications according to changes in the environment, thereby improving the task completion rate.

[0010] In one possible implementation of the first aspect, the step of redetermining the subtasks to be executed at the current moment based on the updated target task and the executed subtasks includes: Determine the priority of the new task objective and the i-th step sub-task; If the priority of the new task objective is greater than the priority of the i-th step sub-task, then stop executing the i-th step sub-task; Based on the updated target task and the executed subtasks, determine the (i+1)th subtask that needs to be executed at the current moment.

[0011] In one possible implementation of the first aspect, after determining the priority of the new task objective and the i-th step sub-task, the method further includes: If the priority of the new task objective is lower than the priority of the i-th step sub-task, then continue to execute the i-th step sub-task; After the i-th step subtask is completed, the (i+1)-th step subtask to be executed is determined based on the updated target task and the executed subtasks.

[0012] In one possible implementation of the first aspect, terminating the execution of the target task includes: If the task analysis results indicate that the task is not feasible, a prompt message is output, which is used to indicate the reason why the user command cannot be executed and the user command cannot be completed. Terminate the execution of the target task.

[0013] In one possible implementation of the first aspect, after real-time detection of situational events in the environment of the embodied agent during the execution of the i-th step subtask, the method further includes: If it is determined that the situational event is not related to the target task, the situational event is ignored, and the i-th step subtask is executed.

[0014] In one possible implementation of the first aspect, the method further includes: After the i-th subtask is completed, the execution result of the i-th subtask is evaluated based on the usage status of the target tool used to execute the i-th subtask, and the task execution evaluation result is obtained. If the task execution evaluation result indicates that the execution result of the i-th step sub-task does not meet the preset requirements, the i-th step sub-task is re-planned according to the task execution evaluation result to obtain the task planning result of the i-th step sub-task, wherein the re-planning includes execution path planning and / or navigation target planning; Execute the i-th subtask based on the task planning results.

[0015] In one possible implementation of the first aspect, after the i-th step subtask is completed, the execution result of the i-th step subtask is evaluated based on the usage status of the target tool used to execute the i-th step subtask, and after obtaining the task execution evaluation result, the method further includes: If the task execution evaluation result indicates that the execution result of the i-th step sub-task meets the preset requirements, then based on the target task and the current environment information of the embodied intelligent agent, the i+1-th step sub-task to be executed is planned.

[0016] Secondly, embodiments of this application provide an embodied intelligent agent, including: The task parsing module is used to respond to a received user instruction, parse the user instruction, and determine the target task to be executed and the task elements of the target task. The task elements include at least one of the target tool to be used to execute the task, the target task object, the geographical scope of the target task, and the target task type. The feasibility analysis module is used to perform a feasibility analysis on the target task based on the embodied intelligent agent's own capabilities and the task elements, and obtain the task analysis results; The task execution module is used to execute the target task based on the target task and the collected current environment information of the embodied intelligent agent if the task analysis result indicates that the task is feasible. The task termination module is used to terminate the execution of the user instruction if the task analysis results indicate that the task is not feasible.

[0017] Thirdly, embodiments of this application provide an embodied intelligent agent, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the task execution method of the embodied intelligent agent as described in any of the first aspects above.

[0018] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the task execution method of the embodied intelligent agent described in any one of the first aspects.

[0019] Fifthly, embodiments of this application provide a computer program product that, when run on an embodied intelligent agent, causes the embodied intelligent agent to execute the task execution method of the embodied intelligent agent described in any of the first aspects above. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art 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.

[0021] Figure 1 This is a schematic diagram of the structure of an existing embodied intelligent agent provided in an embodiment of this application; Figure 2 This is a schematic diagram of the structure of the embodied intelligent agent provided in one embodiment of this application; Figure 3 This is a flowchart illustrating a task execution method for an embodied intelligent agent provided in an embodiment of this application; Figure 4 This is a flowchart illustrating a task replanning method provided in an embodiment of this application; Figure 5 This is a flowchart illustrating a method for task reflection provided in an embodiment of this application; Figure 6 This is a flowchart illustrating a task execution method for an embodied intelligent agent provided in another embodiment of this application; Figure 7 This is a flowchart illustrating a task execution method for an embodied intelligent agent provided in another embodiment of this application; Figure 8 This is a schematic diagram of the structure of an embodied intelligent agent provided in an embodiment of this application; Figure 9 This is a schematic diagram of the structure of an embodied intelligent agent provided in an embodiment of this application. Detailed Implementation

[0022] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0023] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0024] References to "one embodiment" or "some embodiments" in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized.

[0025] An embodied intelligent agent (also known as an embodied robot) refers to an artificial intelligence system that has a physical body and can interact with the environment in real time and dynamically through this body.

[0026] Specifically, such as Figure 1 As shown, the existing embodied agent includes a task planning module, a task execution module, a task observation module, and a task response module. Upon receiving a user instruction, the embodied agent analyzes the instruction using a large language model in the task planning module and plans the tasks to be executed, resulting in sub-tasks to be performed at each step. The task execution module executes the sub-tasks planned by the task planning module, obtaining the sub-task execution results. The task observation module feeds back the sub-task execution results to the task planning module, allowing it to re-plan the task based on these results. When all sub-tasks have been executed or the user instruction cannot be completed, the task response module outputs the task execution result.

[0027] Current embodied intelligent agents generally adopt a serial model, lack real-time supervision of environmental changes, and are unable to cope with complex environmental changes. When the environment changes, they do not plan tasks according to the changes, resulting in task execution failure and low task execution efficiency.

[0028] In addition, existing embodied intelligent agents, upon receiving task instructions, directly execute user tasks according to the instructions. If a task cannot be executed during the execution process, the agent determines the reason for the failure and finally feeds back the task execution failure and the reason for the failure to the user.

[0029] For example, after a smart speaker receives a weather query command, it immediately begins to perform the weather query task. If it needs to access network information during the task execution but finds that the smart speaker cannot access network information, it cannot continue to execute the task, and therefore, the task is determined to have failed.

[0030] To address the aforementioned issues, this application proposes a task execution method for an embodied intelligent agent. Before task execution, the agent first conducts a feasibility analysis of the task based on its own capabilities to determine whether the task is feasible. If the task is determined to be infeasible, it can be skipped to reduce energy consumption; if the task is determined to be feasible, then the task can be executed.

[0031] During the execution of a task, the environment is monitored in real time. When changes are detected in the environment, the ongoing task is replanned based on these changes, new task objectives are determined, and the task is continued according to these new objectives. This improves task execution efficiency and reduces the chance of task failure.

[0032] Specifically, such as Figure 2 As shown, the embodied intelligent agent of this application includes a feasibility analysis module, a task planning module, a task execution module, an environmental perception module, and a task update judgment module.

[0033] The feasibility analysis module uses a large language model to perform feasibility analysis and determines if the task is feasible. The task planning module performs task planning for the target task, uses a large language model to perform single-step task planning, and executes the task according to the task plan.

[0034] During the execution of a single-step task, the environmental perception module collects environmental information and sends it to the task update judgment module.

[0035] The task update determination module calls the visual language big model to determine whether the task has been updated and outputs the determination result; the determination result is then sent to the task planning module.

[0036] If a task update is required, the task planning module will re-plan the task based on the environment information; if no task update is required, the task planning module will continue to execute the currently executing task.

[0037] For example, if the initial task is: go to the bedroom to call Xiaobao to eat; when the embodied intelligent agent detects that Xiaobao is playing in the living room while performing this task, if the embodied intelligent agent continues to go to the bedroom to call Xiaobao to eat according to the initial task, then the task will not be completed; therefore, in this application, according to the change of environment, the task objective of the embodied intelligent agent will be modified to: call Xiaobao to eat in the living room, and the embodied intelligent agent can complete the task by performing the new task objective.

[0038] The following combination Figure 3 The task execution method of the embodied intelligent agent in the embodiments of this application will be described in detail.

[0039] Figure 3 A schematic flowchart illustrating the task execution method of the embodied intelligent agent provided in this application is shown, with reference to... Figure 3 The method is described in detail below: S101, in response to receiving a user instruction, the user instruction is parsed to determine the target task to be executed and the task elements of the target task, wherein the task elements include at least one of the target tool to be used to execute the task, the target task object, the geographical scope of the target task, and the target task type.

[0040] In this embodiment, user commands can be voice information or text information, etc., and there are no restrictions.

[0041] Specifically, the user instructions are input into the text extractor for processing to determine the user's intent, and the target task and task elements to be executed are determined based on the user's intent.

[0042] In this embodiment, the target tool is the tool needed to complete the target task. For example, for the task of picking up an item, a robotic arm is needed, so the robotic arm is the target tool.

[0043] The target task object is the objective of the task, such as a person or an object. For example, in the task of calling Xiaobao to eat in the bedroom, the target task object is "Xiaobao".

[0044] The geographic scope of a target task indicates the locations where the target task needs to be performed. For example, if the target task is "go to the bedroom to call Xiaobao for dinner," then the target task scope is at home. If the target task is "help me pick up a package," then the target task scope is outside the home.

[0045] The target task type is the type of task that needs to be performed. For example, the target task type may include medical tasks, privacy tasks, media asset tasks, laundry tasks, cooking tasks, etc.

[0046] S102, based on the embodied intelligent agent's own capabilities and the task elements, perform a feasibility analysis on the target task to obtain the task analysis results.

[0047] In this embodiment, the capabilities of the embodied intelligent agent include callable tools, a range of movement, executable objects, and types of executable tasks. Callable tools are stored in a tool library, which records the tools and capabilities possessed by the embodied intelligent agent, such as navigation, room positioning, and visual understanding. The range of movement is pre-set within the embodied intelligent agent; for example, the embodied intelligent agent stores a preset map, outlining its range of movement. Executable objects can be objects pre-input by the user, such as family member profiles and relationships between family members. For example, if the target task is "go to the bedroom to call Xiaobao for dinner," and the embodied intelligent agent does not store Xiaobao's information, then Xiaobao is not the target of the embodied intelligent agent's execution, and the target task is not feasible. Executable task types are pre-set; for example, tasks such as cooking and conversation can be performed, but medical and privacy tasks cannot be performed.

[0048] In this embodiment, the task analysis result is finally generated through the thought chain reasoning method, and the task analysis result represents whether the task is feasible or not.

[0049] Specifically, the methods for determining the results of task analysis include: For the detection of the target tool, the target tool is searched from the callable tools to obtain the tool search results. Specifically, it is determined whether the target tool exists among the callable tools. If the target tool exists, the capability boundaries of the target tool in the tool library are determined. Based on the capability boundaries of the target tool, it is determined whether the target tool in the tool library has the ability to perform the target task. For tasks that exceed the capability range of the target tool, it is determined that the tool resources are insufficient, and the corresponding generated tool query results indicate insufficient tool resources. For tasks that do not exceed the capability range of the target tool, it is determined that the tool resources are sufficient, and the corresponding generated tool query results indicate sufficient tool resources.

[0050] For detecting the geographic scope of the target task, it is determined whether the geographic scope of the target task exists within the permitted activity area, thus obtaining a scope judgment result. If the geographic scope of the target task is within the permitted activity area, the generated scope judgment result indicates that it is within the scope; if the geographic scope of the target task is outside the permitted activity area, the generated scope judgment result indicates that it is not within the scope. For example, if the geographic scope of the target task is within a residential area, but the permitted activity area is the interior of a home, then it is determined that the geographic scope of the target task is not within the permitted activity area.

[0051] For the detection of the target object, the system checks whether the target object exists in the executable object to obtain the object search result. If the target object exists in the executable object, the generated object search result indicates that the object exists; if the target object does not exist in the executable object, the generated object search result indicates that the object does not exist.

[0052] For the detection of the target task type, it is determined whether the executable task type contains the target task type, thus obtaining a task type judgment result. If the executable task type does not include the target task type, the generated task type judgment result represents a non-existent type; if the executable task type includes the target task type, the generated task type judgment result represents a present type. For example, if the target task type is a task involving security or legal risks, such as medical diagnosis, it is determined that the target task type does not exist in the executable task types.

[0053] Based on the tool search results, the range judgment results, the object search results, and the task type judgment results, a feasibility analysis is performed on the target task to obtain the task analysis results.

[0054] Specifically, if at least one of the tool search results, the range judgment results, the object search results, and the task type judgment results fails to meet the requirements, the task analysis result is determined to be that the task is infeasible. If all three results meet the requirements, the task analysis result is determined to be that the task is feasible.

[0055] S103, if the task analysis result indicates that the task is feasible, the target task is executed according to the target task and the collected current environment information of the embodied intelligent agent.

[0056] In this embodiment, if the target task is feasible, the process proceeds to single-step task planning, where the subtasks to be executed in the current step are planned. For example, if the target task is "go to the bedroom to call Xiaobao for dinner," the first subtask is: navigate to the bedroom door and plan the navigation path; after completing the current subtask, the next subtask is planned.

[0057] In this embodiment, the task analysis results indicate that the task is feasible and can be displayed to the user so that the user can understand the task progress in a timely manner.

[0058] S104, if the task analysis result indicates that the task is not feasible, terminate the execution of the target task.

[0059] Specifically, if the task analysis results indicate that the task is infeasible, a prompt message is output. This prompt message indicates why the user command cannot be executed and why it cannot be completed; the execution of the target task is then terminated. The prompt message can be output as a voice message or displayed to the user as text.

[0060] For example, the embodied intelligent agent's voice module can play messages such as "The embodied intelligent agent is not equipped with a robotic arm and cannot perform a picking operation."

[0061] In one possible implementation, environmental information is detected in real time during the execution of a subtask, and it is determined whether the currently executing subtask needs to be updated based on the environmental information, so that the embodied agent can adjust the task strategy in a timely manner according to the external environment.

[0062] Specifically, such as Figure 4 As shown, the implementation process of step S103 above may include: S1031, based on the target task, the collected current environment information of the embodied intelligent agent, and the executed sub-tasks, plan the i-th step sub-task to be executed, where i is a positive integer greater than or equal to 1, the i-th step sub-task is the sub-task to be executed to complete the target task, and the executed sub-tasks are the sub-tasks executed before the i-th step.

[0063] In this embodiment, the embodied intelligent agent is equipped with multiple intelligent agents with different functions, and environmental information is collected through different intelligent agents to obtain the current environmental information.

[0064] When executing the target task, this application adopts a single-step execution strategy, planning the next subtask only after completing one subtask. Specifically, it employs an exploratory planning strategy, eliminating the need to plan a complete task chain at the initial stage of the task execution; instead, through a step-by-step planning strategy, the next subtask is determined only after the previous subtask has been completed.

[0065] In this embodiment, the planned i-th step subtask may include an interaction plan and an execution plan to facilitate more precise task completion. The interaction plan is used to plan tasks that interact with the user, such as inquiring about the details of the target task or providing feedback on the task execution results. The execution plan may include the name of the tool to be invoked, the interface parameters of the invoked tool, and the task objective.

[0066] Specifically, task planning is performed based on current environmental information, the long-term memory of the embodied intelligent agent, the short-term memory of the embodied intelligent agent, the target task, and the executed sub-tasks to obtain the i-th step sub-task.

[0067] Long-term memory can include the user's personal information and the environment in which the embodied intelligent agent resides. Personal information can include name, age, nickname, and lifestyle habits. There can be multiple users, such as all family members. The environment information can include a list of items.

[0068] Short-term memory can include information from question-and-answer exchanges between the user and the embodied intelligent agent, as well as the execution status of sub-tasks.

[0069] To ensure task planning aligns with expectations, constraints must be met during the planning process. Therefore, constraint planning can include: 1. Prerequisite tasks between subtasks, such as considering the robot's head pitch before visual understanding; 2. Guidance on how to implement a given task, such as in a person-finding task, first identifying the area where the target person is most likely to be active, navigating to that area, and then finding the target person; 3. Query-type tasks require a response, while instruction-type tasks may not require one.

[0070] In this embodiment, when planning tasks, the sub-goals and the corresponding tools and parameters to be called are first determined to obtain the sub-tasks.

[0071] For example, if the target task is "help me find Li Ming", then the first sub-task might be "navigate to the room where Li Ming is most likely to be", which requires calling the workstation navigation tool.

[0072] S1032, during the execution of the i-th step sub-task, the contextual events of the environment in which the embodied intelligent agent is located are detected in real time, wherein the contextual events include user interaction events and physical environment events.

[0073] In this embodiment, real-time detection of current environmental data, including user voice and visual information, enables environmental monitoring. Contextual events are determined based on user voice and visual information. Visual information can be collected using devices such as cameras. Analysis of the visual information reveals the environmental state, which may include changing objects and existing objects. User voice can reflect new user commands.

[0074] It should be noted that the task planning thread (main thread) and the context event detection thread (secondary thread) exist in parallel. The execution of the i-th subtask and the detection of the context event are parallel. The main thread and the secondary thread communicate with the event queue through shared memory, achieving decoupled collaboration. For shared memory, a centralized memory system is used to support efficient and timely information sharing within the embodied intelligent agent, minimizing memory sharing latency and ensuring that the embodied intelligent agent always uses the latest information for better coordination and efficiency.

[0075] S1033, if it is determined that the situational event is related to the target task, a new task target corresponding to the situational event is determined based on the event content of the situational event.

[0076] In this embodiment, if the contextual event is related to the target task, it means that the contextual event is an event that affects the execution of the target task. It is determined that the current task needs to be updated. Therefore, the content of the contextual event needs to be considered to redetermine the sub-tasks that should be executed.

[0077] For example, if the current target task is "go to the bedroom to call Xiaobao for dinner", and the received user voice message is "don't go", then the contextual event clearly conflicts with the target task. Therefore, it is determined that the contextual event is related to the target task, and the current task needs to be updated.

[0078] In this embodiment, the new task objective is the task that needs to be performed in the event content. For example, if the currently executing objective task is "go to the bedroom to call Xiaobao for dinner", and the received user voice message is "don't go", then the new task objective is "don't go to call Xiaobao for dinner".

[0079] S1034, Update the target task based on the new task objective to obtain the updated target task.

[0080] For example, if the objective task is "go to the bedroom to call Xiaobao to eat", and the new objective task is "Xiaobao enters the living room from the bedroom", the updated objective task is "go to the living room to call Xiaobao to eat".

[0081] If the objective task is "go to the kitchen to check if the gas is off", the new objective task is "go to the kitchen to check if the window is closed", and the updated objective task is "go to the kitchen to check if the gas is off, and go to the kitchen to check if the window is closed".

[0082] S1035, based on the updated target task and executed subtasks, redetermine the subtasks that need to be executed at the current moment, wherein the executed subtasks are the subtasks that were already executed when the target task was executed.

[0083] For example, if the updated target task is "go to the kitchen to see if the gas is turned off and go to the kitchen to see if the window is closed", the executed sub-tasks include the agent navigating to the kitchen door. Then, based on the updated target task and the executed sub-tasks, the sub-task to be executed at the current moment is determined to be: "navigate to the gas stove".

[0084] When determining that the current task needs updating, it's also necessary to consider whether the currently executing subtask i needs to be stopped immediately. If it's determined that subtask i needs to be stopped immediately, the method to stop all tools is called to stop subtask i. Then, the subtasks to be executed are re-determined based on the event content of the context event. If it's determined that subtask i does not need to be stopped immediately, it can continue executing. After subtask i finishes execution, the subtasks to be executed are re-determined based on the event content of the context event.

[0085] Specifically, the implementation process of this step may include: S11, determine the priority of the new task objective and the i-th step sub-task.

[0086] In this embodiment, the urgency of the new task objective and the i-th step sub-task is determined. If the new task objective is more urgent, the new task objective has a higher priority; if the i-th step sub-task is more urgent, the i-th step sub-task has a higher priority.

[0087] For example, the current target task is "go to the bedroom to call Xiaobao for dinner", and the i-th step sub-task is "navigate to the bedroom"; if the received user voice is "don't go", and the contextual event is "no need to call Xiaobao for dinner", then the contextual event clearly has a higher priority than the i-th step sub-task.

[0088] S12, if the priority of the new task target is greater than the priority of the i-th step sub-task, then stop executing the i-th step sub-task.

[0089] In this embodiment, tasks with higher priority are executed first; if a new task objective has a higher priority, it means that the new task objective is more urgent, and therefore, the currently executing sub-task i needs to be stopped immediately.

[0090] S13, based on the updated target task and the executed sub-tasks, determine the (i+1)th step sub-task that needs to be executed at the current moment.

[0091] S14, if the priority of the new task target is lower than the priority of the i-th step sub-task, then continue to execute the i-th step sub-task.

[0092] For example, if the target task is "go to the kitchen to see if the gas stove is turned off", the i-th step subtask being executed is "navigate to the kitchen"; if the situational event is "check if the kitchen window is closed", then the priority of the new task target is lower than the priority of the i-th step subtask, and the i-th step subtask can continue to be executed.

[0093] S15, after the i-th step subtask is completed, determine the (i+1)-th step subtask to be executed based on the updated target task and the executed subtask.

[0094] In another implementation, if it is determined that the contextual event is not related to the target task, it means that the current task does not need to be updated, the contextual event is ignored, and the i-th step subtask is continued.

[0095] For example, if the target task is "go to the balcony to see if the laundry is finished" and the contextual event is "go to the balcony to see if the laundry is finished", then the contextual event is unrelated to the target task. Since the contextual event and the target task are determined to be unrelated, the current subtask can continue to be executed.

[0096] S1036, execute the redefined subtask.

[0097] In one possible implementation, after each subtask is completed, the execution process of the subtask can be adjusted based on the tool's state during the subtask execution, or the next subtask can be continued. This allows for self-examination and reflection on one's own behavior, learning from mistakes and drawing lessons to provide improvement suggestions for subsequent actions, thereby effectively improving the quality of the final result.

[0098] Specifically, such as Figure 5 As shown, the method of this application may also include: S201, after the i-th step sub-task is completed, the execution result of the i-th step sub-task is evaluated based on the usage status of the target tool used to execute the i-th step sub-task, and a task execution evaluation result is obtained.

[0099] In this embodiment, a tool from the tool library is invoked to execute the i-th subtask. During the execution of the i-th subtask, the usage information (or usage status) of the invoked target tool is stored, such as whether the navigation tool has successfully reached the target location, or whether the visual understanding tool has identified the target object.

[0100] Based on the usage status of the target tool, a comprehensive evaluation is conducted on the execution process of the i-th step sub-task to determine whether the execution result of the i-th step sub-task meets the expectations (preset requirements), thus obtaining the task execution evaluation result.

[0101] S202, if the task execution evaluation result indicates that the execution result of the i-th step sub-task does not meet the preset requirements, the i-th step sub-task is re-planned according to the task execution evaluation result to obtain the task planning result of the i-th step sub-task, wherein the re-planning includes execution path planning and / or navigation target planning.

[0102] For example, if the embodied agent fails to reach the target location, the execution path can be replanned.

[0103] S203, Execute the i-th step subtask according to the task planning result.

[0104] In this embodiment, the i-th step subtask is re-executed so that the execution process of the i-th step subtask is closer to the expectation, thereby making the execution of the target task more accurate.

[0105] S204, if the task execution evaluation result indicates that the execution result of the i-th step sub-task meets the preset requirements, then plan the i+1-th step sub-task to be executed based on the target task and the current environment information of the embodied intelligent agent.

[0106] In this embodiment, the (i+1)th step subtask is executed, and the execution process of the above subtasks is repeated. The embodied intelligent agent evaluates whether the target task has been completed. If it determines that the target task has been completed, it outputs the task completion result.

[0107] The following two specific examples further illustrate the method of this application.

[0108] Example 1: The user command is "Clean the living room." The task planning module in the embodied agent begins task planning, determining the sub-tasks to be executed. During the execution of the sub-tasks, the multimodal perception module collects current environmental information and sends it to the environmental reflection module. The environmental reflection module obtains the current environmental information and determines the contextual event based on it. For example, if the contextual event is that a pet cat runs out of the bedroom, leaving fur and paw prints in the living room, the new task objective is determined to be "focus on cleaning the pet cat's fur and paw prints." This contextual event is related to the currently executed target task; however, it is not urgent. Therefore, its priority is lower than the priority of the currently executed sub-task. Thus, the current sub-task can continue, and the task objective of focusing on cleaning the pet cat's fur and paw prints can be added during subsequent cleaning processes.

[0109] like Figure 6 As shown, the process of an embodied intelligent agent processing user commands includes: S21, after receiving the user's instruction, the task planning module determines the target task to be executed as "cleaning the living room" based on the task instruction.

[0110] S22, the task planning module plans the current sub-task based on the target task, and starts the cleaning mode to begin cleaning the living room according to the current sub-task.

[0111] S23, the multimodal perception module collects current environmental information.

[0112] S24, the multimodal perception module sends the current environmental information to the environmental reflection module.

[0113] S25, the environmental reflection module obtains the current environmental information and determines the situational event as "focusing on cleaning the pet cat's fur and paw prints" based on the current environmental information.

[0114] S26, the environmental reflection module determines that the situational event is related to the target task being executed, and updates the target task to: "Clean the living room, focusing on cleaning up the pet cat's fur and paw prints", but the priority of the situational event is lower than the priority of the current subtask being executed.

[0115] S27, the environmental reflection module sends the first task adjustment instruction to the task planning module, which includes the situational event.

[0116] S28. After receiving the first task adjustment instruction, the task planning module continues to execute the current subtask. After the current subtask ends, it plans the next subtask based on the updated target task and the executed subtasks.

[0117] Example 2: The user command is "Go to the bedroom and call Xiaobao for dinner." The task planning module in the embodied agent begins task planning, determines the sub-tasks to be executed, and begins going to the bedroom to call Xiaobao for dinner. The task planning module sends information about the sub-tasks being executed to the shared memory module. The environment reflection module obtains the sub-task information from the shared memory module.

[0118] During the execution of subtasks, the multimodal perception module collects current environmental information and sends it to the environmental reflection module. The environmental reflection module obtains the current environmental information and determines the contextual event based on it. For example, if the contextual event is detecting Xiaobao entering the corridor, the new task objective is determined to be "Xiaobao is in the corridor." This contextual event is related to the currently executing target task and is relatively urgent. Therefore, the priority of this contextual event is higher than the priority of the currently executing subtask. Thus, the currently executing subtask needs to be stopped, and a new subtask is determined based on the target task and the new task objective, such as "Xiaobao has been found; call Xiaobao to the living room for dinner."

[0119] If the current environmental information includes a kitten rolling on the ground, then the situational event is "the kitten is rolling on the ground". However, if this situational event is irrelevant to the target task, then no action will be taken on this situational event.

[0120] Specifically, such as Figure 7 As shown, the process by which an embodied intelligent agent processes user commands includes the following steps.

[0121] S31, after receiving the user's instruction, the task planning thread determines the target task to be executed as "go to the bedroom to call Xiaobao for dinner".

[0122] S32, the task planning module plans the current sub-task based on the target task, and executes the current sub-task, going to the bedroom to call Xiaobao for dinner.

[0123] S33, the task planning module sends the current subtask and task execution information to the shared memory module. The task execution information includes path planning, the state of the embodied intelligent agent, etc., so that the shared memory module can store the current subtask and task execution information.

[0124] S34, the multimodal perception module collects current environmental information.

[0125] S35, the multimodal perception module sends the current environmental information to the environmental reflection module.

[0126] S36, the environmental reflection module obtains the current environmental information and determines the situational event as "Xiao Bao is in the corridor" based on the current environmental information.

[0127] S37, the environment reflection module determines that the contextual event is related to the target task being executed, and determines that the priority of the contextual event is higher than the priority of the currently executing subtask, which is obtained from shared memory.

[0128] S38, the environment reflection module sends a second task adjustment instruction to the shared memory module. The second task adjustment instruction includes a stop instruction, a task replanning instruction, and a situational event.

[0129] S39, the task planning module obtains the second task adjustment instruction from the shared memory module, stops executing the current subtask, and redetermines a new subtask based on the context event and the target task, and executes the new subtask. The new subtask is "Xiao Bao has been found, call Xiao Bao to the living room for dinner".

[0130] S40, the task planning module sends a new subtask to the shared memory so that the shared memory module can store the subtask.

[0131] S41, the multimodal perception module collects current environmental information.

[0132] S42, the multimodal perception module sends the current environmental information to the environmental reflection module.

[0133] S43, the environmental reflection module obtains the current environmental information and determines the situational event as "the kitten is rolling on the ground" based on the current environmental information.

[0134] S37, the environmental reflection module determines that the situational event is irrelevant to the target task being performed and ignores the situational event.

[0135] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0136] Corresponding to the task execution method of the embodied intelligent agent described in the above embodiments, Figure 8 A structural block diagram of the embodied intelligent agent provided in the embodiments of this application is shown. For ease of explanation, only the parts related to the embodiments of this application are shown.

[0137] Reference Figure 8 The embodied intelligent agent 300 may include: a task parsing module 310, a feasibility analysis module 320, a task execution module 330, and a task termination module 340.

[0138] The task parsing module 310 is used to parse the user instruction in response to receiving the user instruction, and determine the target task to be executed and the task elements of the target task. The task elements include at least one of the target tool to be used to execute the task, the target task object, the geographical scope of the target task, and the target task type. The feasibility analysis module 320 is used to perform a feasibility analysis on the target task based on the embodied intelligent agent's own capabilities and the task elements, and obtain the task analysis results. The task execution module 330 is used to execute the target task based on the target task and the collected current environment information of the embodied intelligent agent if the task analysis result indicates that the task is feasible. The task termination module 340 is used to terminate the execution of the user instruction if the task analysis result indicates that the task is not feasible.

[0139] In one possible implementation, the embodied intelligent agent's own capabilities include callable tools, a range of activities, executable objects, and types of executable tasks; When the task elements include the target tools required to perform the task, the target task object, the geographical scope of the target task, and the type of the target task, the feasibility analysis module 320 can specifically be used for: The target tool is searched from the available tools to obtain the tool search results; Determine whether the geographical scope of the target task exists within the active area to obtain a scope determination result; Check if the target object exists in the executable object to obtain the object search result; Determine whether the executable task type contains the target task type to obtain a task type determination result; Based on the tool search results, the range judgment results, the object search results, and the task type judgment results, a feasibility analysis is performed on the target task to obtain the task analysis results.

[0140] In one possible implementation, the task execution module 330 can specifically be used for: Based on the target task, the current environmental information of the embodied intelligent agent, and the sub-tasks already executed, plan the i-th step sub-task to be executed, where i is a positive integer greater than or equal to 1, the i-th step sub-task is the sub-task to be executed to complete the target task, and the executed sub-tasks are the sub-tasks executed before the i-th step. During the execution of the i-th step subtask, the contextual events of the environment in which the embodied intelligent agent is located are detected in real time, wherein the contextual events include user interaction events and physical environment events; If it is determined that the situational event is related to the target task. Based on the content of the situational event, determine the new task objective corresponding to the situational event; The target task is updated based on the new task objective to obtain the updated target task; Based on the updated target task and executed subtasks, the subtasks that need to be executed at the current moment are re-determined, wherein the executed subtasks are the subtasks that were already executed when the target task was executed; Execute the redefined subtasks.

[0141] In one possible implementation, the task execution module 330 can specifically be used for: Determine the priority of the new task objective and the i-th step sub-task; If the priority of the new task objective is greater than the priority of the i-th step sub-task, then stop executing the i-th step sub-task; Based on the updated target task and the executed subtasks, determine the (i+1)th subtask that needs to be executed at the current moment.

[0142] In one possible implementation, the task execution module 330 can specifically be used for: If the priority of the new task objective is lower than the priority of the i-th step sub-task, then continue to execute the i-th step sub-task; After the i-th step subtask is completed, the (i+1)-th step subtask to be executed is determined based on the updated target task and the executed subtasks.

[0143] In one possible implementation, the task termination module 340 can specifically be used for: If the task analysis results indicate that the task is not feasible, a prompt message is output, which is used to indicate the reason why the user command cannot be executed and the user command cannot be completed. Terminate the execution of the user instruction.

[0144] In one possible implementation, the task execution module 330 can specifically be used for: If it is determined that the situational event is not related to the target task, the situational event is ignored, and the i-th step subtask is executed.

[0145] In one possible implementation, the embodied intelligent agent 300 further includes: The task reflection module is used to evaluate the execution result of the i-th sub-task after the i-th sub-task is completed, based on the usage status of the target tool used to execute the i-th sub-task, and obtain the task execution evaluation result. The replanning module is used to replan the i-th step subtask based on the task execution evaluation result if the task execution evaluation result indicates that the execution result of the i-th step subtask does not meet the preset requirements, thereby obtaining the task planning result of the i-th step subtask. The replanning includes execution path planning and / or navigation target planning. The execution module is used to execute the i-th step subtask according to the task planning result.

[0146] In one possible implementation, the execution module can specifically be used for: If the task execution evaluation result indicates that the execution result of the i-th step sub-task meets the preset requirements, then based on the target task and the current environment information of the embodied intelligent agent, the i+1-th step sub-task to be executed is planned.

[0147] It should be noted that the information interaction and execution process between the above-mentioned devices / units are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, and they will not be repeated here.

[0148] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0149] This application also provides an embodied intelligent agent, see [link to relevant documentation]. Figure 9 The embodied intelligent agent 400 may include: at least one processor 410, a memory 420, and a computer program stored in the memory 420 and executable on the at least one processor 410. When the processor 410 executes the computer program, it implements the steps in any of the above method embodiments, for example... Figure 3 Steps S101 to S104 in the illustrated embodiment. Alternatively, when the processor 410 executes the computer program, it implements the functions of each module / unit in the above-described device embodiments, for example... Figure 8 The functions of the task parsing module 310 to the task termination module 340 are shown.

[0150] For example, a computer program may be divided into one or more modules / units, one or more of which are stored in memory 420 and executed by processor 410 to complete this application. The one or more modules / units may be a series of computer program segments capable of performing specific functions, which describe the execution process of the computer program in the embodied intelligent agent 400.

[0151] Those skilled in the art will understand that Figure 9 This is merely an example of an embodied intelligent agent and does not constitute a limitation on embodied intelligent agents. It may include more or fewer components than illustrated, or combine certain components, or different components, such as input / output devices, network access devices, buses, etc.

[0152] The processor 410 can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor.

[0153] The memory 420 can be an internal storage unit of the embodied intelligent agent or an external storage device, such as a plug-in hard drive, a smart media card (SMC), a secure digital (SD) card, or a flash card. The memory 420 is used to store the computer program and other programs and data required by the embodied intelligent agent. The memory 420 can also be used to temporarily store data that has been output or will be output.

[0154] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0155] The method for dynamically expanding relational databases provided in this application can be applied to embodied intelligent agents such as computers, tablets, laptops, netbooks, and personal digital assistants (PDAs). This application does not impose any restrictions on the specific type of embodied intelligent agent.

[0156] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0157] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0158] In the embodiments provided in this application, it should be understood that the disclosed embodied intelligent agents, devices, and methods can be implemented in other ways. For example, the embodied intelligent agent embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0159] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0160] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0161] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by one or more processors, it can implement the steps of the various method embodiments described above.

[0162] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by one or more processors, it can implement the steps of the various method embodiments described above.

[0163] Similarly, as a computer program product, when the computer program product is run on an embodied intelligent agent, it enables the embodied intelligent agent to perform the steps in the above-described method embodiments.

[0164] The computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media may not include electrical carrier signals and telecommunication signals.

[0165] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A task execution method for an embodied intelligent agent, characterized in that, include: In response to receiving a user instruction, the user instruction is parsed to determine the target task to be executed and the task elements of the target task, wherein the task elements include at least one of the target tool to be used to execute the task, the target task object, the geographical scope of the target task, and the target task type; Based on the capabilities of the embodied intelligent agent and the task elements, a feasibility analysis is performed on the target task to obtain the task analysis results; If the task analysis results indicate that the task is feasible, the target task is executed based on the target task and the collected environmental information of the embodied intelligent agent. If the task analysis results indicate that the task is not feasible, the execution of the target task shall be terminated.

2. The task execution method of the embodied intelligent agent as described in claim 1, characterized in that, The capabilities of the embodied intelligent agent include the ability to invoke tools, the scope of its activities, the objects it can execute, and the types of tasks it can perform. When the task elements include the target tools required to perform the task, the target task object, the geographical scope of the target task, and the target task type, the feasibility analysis of the target task based on the embodied intelligent agent's own capabilities and the task elements is performed to obtain the task analysis results, including: The target tool is searched from the available tools to obtain the tool search results; Determine whether the geographical range of the target task exists within the active range, and obtain the range determination result; Check if the target object exists in the executable object to obtain the object search result; Determine whether the executable task type contains the target task type to obtain the task type determination result; Based on the tool search results, the range judgment results, the object search results, and the task type judgment results, a feasibility analysis is performed on the target task to obtain the task analysis results.

3. The task execution method of the embodied intelligent agent as described in claim 1, characterized in that, The step of executing the target task based on the target task and the collected current environment information of the embodied intelligent agent includes: Based on the target task, the current environmental information of the embodied intelligent agent, and the sub-tasks already executed, plan the i-th step sub-task to be executed, where i is a positive integer greater than or equal to 1, the i-th step sub-task is the sub-task to be executed to complete the target task, and the executed sub-tasks are the sub-tasks executed before the i-th step. During the execution of the i-th step subtask, the contextual events of the environment in which the embodied intelligent agent is located are detected in real time, wherein the contextual events include user interaction events and physical environment events; If it is determined that the situational event is related to the target task, a new task target corresponding to the situational event is determined; The target task is updated based on the new task objective to obtain the updated target task; Based on the updated target task and executed subtasks, the subtasks that need to be executed at the current moment are re-determined, wherein the executed subtasks are the subtasks that were already executed when the target task was executed; Execute the redefined subtasks.

4. The task execution method for an embodied intelligent agent as described in claim 3, characterized in that, The step of redetermining the subtasks to be executed at the current moment based on the updated target task and executed subtasks includes: Determine the priority of the new task objective and the i-th step sub-task; If the priority of the new task objective is greater than the priority of the i-th step sub-task, then stop executing the i-th step sub-task; Based on the updated target task and the executed subtasks, determine the (i+1)th subtask that needs to be executed at the current moment.

5. The task execution method for an embodied intelligent agent as described in claim 4, characterized in that, After determining the priority of the new task objective and the i-th step sub-task, the method further includes: If the priority of the new task objective is lower than the priority of the i-th step sub-task, then continue to execute the i-th step sub-task; After the i-th step subtask is completed, the (i+1)-th step subtask to be executed is determined based on the updated target task and the executed subtasks.

6. The task execution method for an embodied intelligent agent as described in claim 1, characterized in that, The termination of the execution of the target task includes: If the task analysis results indicate that the task is not feasible, a prompt message is output, which is used to indicate the reason why the user command cannot be executed and the user command cannot be completed. Terminate the execution of the target task.

7. The task execution method for an embodied intelligent agent as described in claim 3, characterized in that, After real-time detection of situational events in the environment where the embodied agent is located during the execution of the i-th step sub-task, the method further includes: If it is determined that the situational event is not related to the target task, the situational event is ignored, and the i-th step subtask is executed.

8. The task execution method of the embodied intelligent agent as described in claim 5 or 7, characterized in that, The method further includes: After the i-th subtask is completed, the execution result of the i-th subtask is evaluated based on the usage status of the target tool used to execute the i-th subtask, and the task execution evaluation result is obtained. If the task execution evaluation result indicates that the execution result of the i-th step sub-task does not meet the preset requirements, the i-th step sub-task is re-planned according to the task execution evaluation result to obtain the task planning result of the i-th step sub-task, wherein the re-planning includes execution path planning and / or navigation target planning; Execute the i-th subtask based on the task planning results.

9. The task execution method of the embodied intelligent agent as described in claim 8, characterized in that, After the i-th subtask is completed, the execution result of the i-th subtask is evaluated based on the usage status of the target tool used to execute the i-th subtask. After obtaining the task execution evaluation result, the method further includes: If the task execution evaluation result indicates that the execution result of the i-th step sub-task meets the preset requirements, then based on the target task and the current environment information of the embodied intelligent agent, the i+1-th step sub-task to be executed is planned.

10. An embodied intelligent agent, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the task execution method of the embodied intelligent agent as described in any one of claims 1 to 9.

Images